WO2022143802A1 - 换电站排队车辆的数量识别方法、***、设备及介质 - Google Patents

换电站排队车辆的数量识别方法、***、设备及介质 Download PDF

Info

Publication number
WO2022143802A1
WO2022143802A1 PCT/CN2021/142657 CN2021142657W WO2022143802A1 WO 2022143802 A1 WO2022143802 A1 WO 2022143802A1 CN 2021142657 W CN2021142657 W CN 2021142657W WO 2022143802 A1 WO2022143802 A1 WO 2022143802A1
Authority
WO
WIPO (PCT)
Prior art keywords
queuing
vehicle
station
image
vehicles
Prior art date
Application number
PCT/CN2021/142657
Other languages
English (en)
French (fr)
Inventor
王昊杰
Original Assignee
奥动新能源汽车科技有限公司
上海电巴新能源科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202011618777.6A external-priority patent/CN114694053A/zh
Priority claimed from CN202011618807.3A external-priority patent/CN114694084A/zh
Application filed by 奥动新能源汽车科技有限公司, 上海电巴新能源科技有限公司 filed Critical 奥动新能源汽车科技有限公司
Publication of WO2022143802A1 publication Critical patent/WO2022143802A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition

Definitions

  • the present application relates to the technical field of power exchange, and in particular, to a method, system, device and medium for identifying the number of vehicles queuing in a power exchange station.
  • Swap station is a place to quickly and efficiently replenish electric energy for new energy vehicles.
  • the swap station can not only save the requirements of new energy vehicles on charging stations/piles and other equipment, but also improve the utilization rate of equipment and meet the needs of new energy drivers to the greatest extent. need.
  • Fully automatic power exchange and fast charging services can be provided for electric vehicles in the power exchange station.
  • the number of power exchange users has gradually increased, especially taxis.
  • more efficient power exchange services are preferred.
  • the number of queuing vehicles for battery swapping is generally counted manually by the staff of the swapping station, which leads to a waste of human resources, increases the cost of the swapping station, and has low work efficiency.
  • the technical problem to be solved by this application is to provide a method, system, equipment and method for identifying the number of vehicles queuing in a battery swap station, in order to overcome the defects in the prior art that the number of queuing vehicles for battery swapping is manually counted, which leads to waste of human resources and reduces efficiency. medium.
  • a method for identifying the number of vehicles queued at a power exchange station including:
  • the queue number of the swap station is determined according to the identified queued vehicles.
  • the identification efficiency of the queuing quantity can be effectively improved, and it can also effectively It can save the human resource consumption of the power station and reduce the cost.
  • the identifying the queued vehicles included in the environment image includes:
  • Queued vehicle identification is performed on the blurred environment image.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • other non-exchange vehicles in the invalid area can effectively eliminate the interference to the identification of the vehicle.
  • the recognition accuracy of queuing vehicles can also avoid the workload of identifying invalid areas and improve the recognition efficiency of queuing vehicles.
  • the acquiring an environmental image of the vehicle queuing area of the swap station includes:
  • the identifying the queued vehicles included in the environment image includes:
  • the queuing vehicle identification is performed.
  • multiple cameras are used to collect video streams from different viewing angles to ensure that the environment image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the queuing vehicle identification is performed, including:
  • the environment images are cropped and merged in the repeated area
  • Queued vehicle identification is performed on the target image obtained after compression and/or encoding.
  • the environment images with repeated areas are cropped and merged, the effective and non-repetitive areas are retained, the image content that needs to be recognized is reduced as much as possible, and the recognition efficiency of queuing vehicles is improved;
  • the image processing of compression and/or encoding of the environmental image can reduce the data amount of the image to be identified, and can further improve the identification efficiency of the queued vehicles.
  • the acquiring an environmental image of a vehicle queuing area at the power swap station includes:
  • the obtaining of the environmental image of the vehicle queuing area of the swap station includes:
  • An environment image of the vehicle queuing area of the battery swap station reported by the battery swap station is received.
  • image recognition when applied to a power station, image recognition can be performed through the local server of the power station, which can reduce bandwidth pressure and cloud computing pressure; when applied to the cloud, image recognition can be performed through the cloud server to facilitate unified management and scheduling of data , which can effectively avoid data tampering and so on.
  • the method when applied to a power exchange station, the method further includes:
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting to be swapped can easily select swap stations with fewer queues, thereby saving user time and improving power swapping. efficiency.
  • the queuing quantity determined according to the identified queuing vehicles is checked by using the queuing quantity obtained according to the positioning information of the vehicles entering the network.
  • the number of queues obtained through the positioning information of vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the identifying the queued vehicles included in the environment image includes:
  • the environment image is input into the queuing vehicle recognition model, and the queuing vehicle recognition model is used to identify the queuing vehicle in the environment image;
  • the queuing vehicle recognition model is trained according to the environmental image of the vehicle queuing area of the swap station and the corresponding training label get;
  • the determining of the queued number of the swapping station according to the identified queued vehicles includes:
  • the queuing quantity of the swapping station is determined by the recognition result output by the queuing vehicle recognition model.
  • the historical environment images are used to train the queuing vehicle recognition model in advance, so that the queuing vehicle recognition model can learn the ability to recognize queuing vehicles based on the environmental image, and then the trained model can identify the queuing vehicles in the environmental image, and further improve the recognition accuracy and efficiency.
  • the training steps of the queuing vehicle identification model include:
  • the queuing vehicle identification model is trained according to the model training samples and corresponding training labels.
  • multiple cameras are used to collect video streams from different perspectives to capture environmental images from them, and de-overlap them and use them as model training samples, so that the environmental images can fully cover the vehicle queuing area as much as possible, thereby improving model recognition. accuracy, and improve the stability of the training model.
  • model training samples including:
  • the blurred intermediate image is used as a model training sample.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • other non-exchange vehicles in the invalid area can effectively eliminate the interference to the identification of the vehicle.
  • the recognition accuracy of queuing vehicles can also avoid the workload of identifying invalid areas and improve the recognition efficiency of queuing vehicles.
  • the training of the queuing vehicle recognition model according to the model training samples and corresponding training labels includes:
  • the model parameters of the queuing vehicle recognition model are optimized.
  • the corresponding image processing is performed on the model training samples, so as to facilitate the model training and reduce the amount of data transmitted, thereby improving the stability of the model training, effectively saving the transmission network bandwidth, reducing the cost, and also Effectively ensure data security.
  • the updating step of the queuing vehicle identification model includes:
  • the queued vehicle identification model is updated.
  • the training step is performed by a dedicated server; after the training of the queuing vehicle identification model is completed, the dedicated server sends the queuing vehicle identification model to each of the battery swap stations and/or the battery swap cloud, so that The power exchange station and/or the power exchange cloud use the queued vehicle identification model to identify the environment image and output the number of queues.
  • the trained model is delivered to each power exchange station and/or power exchange cloud, so that each power exchange station and/or power exchange cloud can perform model recognition, so as to arrange the model training process and use process in an orderly manner, effectively It saves costs and improves the stability of model recognition.
  • a system for identifying the number of vehicles queuing at a power exchange station including:
  • an environmental image acquisition module configured to acquire an environmental image of the vehicle queuing area of the swap station
  • a queued vehicle identification module configured to identify queued vehicles included in the environmental image
  • the queuing quantity determining module is configured to determine the queuing quantity of the swap station according to the identified queuing vehicles.
  • the queuing quantity of the swapping station can be conveniently, timely and accurately determined by the real-time collected environmental images of the vehicle queuing area at the swapping station, so that the queuing time, sequence and other power swap strategies can be determined based on the queuing quantity. It can be visually displayed to the user to be replaced, thereby effectively saving human resources, reducing the cost of the power exchange station, and improving work efficiency.
  • the queuing vehicle identification module is configured to:
  • Queued vehicle identification is performed on the blurred environment image.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • other non-exchange vehicles in the invalid area can effectively eliminate the interference to the identification of the vehicle.
  • the recognition accuracy of queuing vehicles can also avoid the workload of identifying invalid areas and improve the recognition efficiency of queuing vehicles.
  • the environmental image acquisition module is configured to:
  • the queued vehicle identification module is configured to:
  • the queuing vehicle identification is performed.
  • multiple cameras are used to collect video streams from different viewing angles to ensure that the environment image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the queuing vehicle identification module is further configured to:
  • the environment images are cropped and merged in the repeated area
  • Queued vehicle identification is performed on the target image obtained after compression and/or encoding.
  • the environment images with repeated areas are cropped and merged, the effective and non-repetitive areas are retained, the image content that needs to be recognized is reduced as much as possible, and the recognition efficiency of queuing vehicles is improved;
  • the image processing of compression and/or encoding of the environmental image can reduce the data amount of the image to be identified, and can further improve the identification efficiency of the queued vehicles.
  • the environmental image acquisition module is configured to:
  • the environmental image acquisition module is configured to:
  • An environment image of the vehicle queuing area of the battery swap station reported by the battery swap station is received.
  • image recognition when applied to a power station, image recognition can be performed through the local server of the power station, which can reduce bandwidth pressure and cloud computing pressure; when applied to the cloud, image recognition can be performed through the cloud server to facilitate unified management and scheduling of data , which can effectively avoid data tampering and so on.
  • the queuing quantity determining module is further configured to:
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting to be swapped can easily select swap stations with fewer queues, thereby saving user time and improving power swapping. efficiency.
  • the system also includes a queuing quantity verification module
  • the queue quantity verification module is configured to:
  • the queuing quantity determined according to the identified queuing vehicles is checked by using the queuing quantity obtained according to the positioning information of the vehicles entering the network.
  • the number of queues obtained through the positioning information of vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the queuing vehicle identification module is configured to: input the environment image into a queuing vehicle identification model, and identify the queuing vehicle in the environment image through the queuing vehicle identification model; The environmental images and corresponding training labels of the vehicle queuing area of the power station are trained;
  • the queuing quantity determination module is configured to: determine the queuing quantity of the swapping station through the identification result output by the queuing vehicle identification model.
  • the historical environment images are used to train the queuing vehicle recognition model in advance, so that the queuing vehicle recognition model can learn the ability to recognize queuing vehicles based on the environmental image, and then the trained model can identify the queuing vehicles in the environmental image, and further improve the recognition accuracy and efficiency.
  • the system also includes a model training module
  • the model training module is configured to:
  • the queuing vehicle identification model is trained according to the model training samples and corresponding training labels.
  • multiple cameras are used to collect video streams from different perspectives to capture environmental images from them, and de-overlap them and use them as model training samples, so that the environmental images can fully cover the vehicle queuing area as much as possible, thereby improving model recognition. accuracy, and improve the stability of the training model.
  • model training module is also configured to:
  • the blurred intermediate image is used as a model training sample.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • other non-exchange vehicles in the invalid area can effectively eliminate the interference to the identification of the vehicle.
  • the recognition accuracy of queuing vehicles can also avoid the workload of identifying invalid areas and improve the recognition efficiency of queuing vehicles.
  • model training module is also configured to:
  • the model parameters of the queuing vehicle recognition model are optimized.
  • the corresponding image processing is performed on the model training samples, so as to facilitate the model training and reduce the amount of data transmitted, thereby improving the stability of the model training, effectively saving the transmission network bandwidth, reducing the cost, and also Effectively ensure data security.
  • model training module is also configured to:
  • the queued vehicle identification model is updated.
  • the model training module is configured on a dedicated server, and after the training of the queuing vehicle identification model is completed, the dedicated server is configured to deliver the queuing vehicle identification model to each power exchange station and/or power exchange cloud, so that the power exchange can be changed.
  • the power station and/or the power exchange cloud adopts the queuing vehicle recognition model to identify the number of queues in the environment image output.
  • an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, when the processor executes the computer program, the above-mentioned power exchange station is implemented A method for identifying the number of queued vehicles.
  • an electronic device which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, when the processor executes the computer program, the above-mentioned queuing at a power exchange station is implemented.
  • a computer-readable medium on which computer instructions are stored, and when executed by a processor, the computer instructions implement the steps of the above-mentioned method for identifying the number of vehicles queued at a power exchange station.
  • the present application automatically acquires the environmental image of the vehicle queuing area of the swapping station, and identifies the queued vehicles from the environmental image to determine the number of queues at the swapping station, which can effectively improve the identification efficiency of the number of queues, and can also effectively save the replacement cost.
  • the human resources of the power station are consumed and the cost is reduced.
  • FIG. 1 is a schematic flowchart of a method for identifying the number of vehicles queued at a power exchange station according to Embodiment 1 of the present application.
  • FIG. 2 is a schematic flowchart of a method for identifying the number of vehicles queued at a swapping station according to Embodiment 2 of the present application.
  • FIG. 3 is a schematic structural diagram of a module of a system for identifying the number of vehicles queuing at a power exchange station according to Embodiment 3 of the present application.
  • FIG. 4 is a schematic structural diagram of a module of a system for identifying the number of vehicles queuing at a power exchange station according to Embodiment 4 of the present application.
  • FIG. 5 is a schematic structural diagram of an electronic device for implementing a method for identifying the number of vehicles queuing at a power exchange station according to Embodiment 5 of the present application.
  • the present embodiment provides a method for identifying the number of vehicles queued at a swap station, including: acquiring an environmental image of a vehicle queuing area at the swap station; identifying queuing vehicles included in the environmental image; The vehicle determines the number of queues at the swap station.
  • the method can be applied to a power station or the cloud, but its application scenarios are not specifically limited, and corresponding selection and adjustment can be made according to actual needs.
  • the identification efficiency of the queuing quantity can be effectively improved, and the It can effectively save the human resource consumption of the swap station and reduce the cost.
  • the method for identifying the number of vehicles queuing at a power exchange station mainly includes the following steps:
  • Step 101 Obtain an environmental image of the vehicle queuing area of the swap station.
  • the vehicle queuing area may be a predetermined area in the power exchange station for vehicle queuing, or may be an area formed by the natural queuing of vehicles.
  • the local server of the swapping station invokes the camera of the swapping station to collect the video stream of the vehicle queuing area of the swapping station in real time, and converts the video stream into an image to capture the vehicles of the swapping station therefrom. Environmental image of the queuing area.
  • image recognition can be performed through the local server of the power exchange station, which can reduce bandwidth pressure and cloud computing pressure.
  • multiple cameras at different positions of the battery swap station are called to collect real-time video streams of the vehicle queuing area of the battery swap station from multiple perspectives, and the video streams are respectively converted into images to be intercepted from them.
  • Video streams from different perspectives are collected by multiple cameras to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • a camera can be used to change the viewing angle in real time to collect video streams from different viewing angles, so as to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the exchange station when the method is applied to the cloud, the exchange station will report the collected video stream to the cloud server, and may also report the captured environment image of the vehicle queuing area of the exchange station to the cloud server, and the cloud server will receive the video Stream to get the environment image or receive the environment image directly.
  • image recognition can be performed through the cloud server to facilitate the unified management and scheduling of data, and can effectively avoid data tampering and other situations.
  • Step 102 Identify the queued vehicles included in the environment image.
  • image processing is performed on the obtained environment image, and the queued vehicles are identified from the image-processed environment image through image recognition technology.
  • This embodiment does not specifically limit the image recognition method, as long as the corresponding functions can be realized, corresponding selections can be made according to actual needs.
  • the image processing includes at least one of compression processing, encoding processing and encryption processing, for example, compressing the environmental image, and converting the compressed image into Base64 (the most common on the network for transmitting 8Bit bytes One of the encoding methods of the code), and then compress and encrypt the Base64 encoding.
  • this embodiment does not specifically limit the image processing mode, and corresponding selections can be made according to actual needs.
  • Corresponding image processing is performed on the obtained environmental image, so as to reduce the amount of transmitted data, thereby effectively saving transmission network bandwidth, reducing costs, and effectively ensuring data security.
  • the environmental image after image processing can be reported to the cloud server, and the cloud server can choose to call a third-party image recognition service for vehicle recognition, thereby effectively saving the transmission network bandwidth, cut costs.
  • the queuing vehicles are identified after the acquired environment image is deduplicated.
  • the environment images are cropped and merged, and the merged environment images are compressed and/or encoded, and then the compressed and/or encoded environment images are compressed and/or encoded.
  • the obtained target image is used for queuing vehicle identification.
  • the environment images with repeated areas are cropped and merged, the effective and non-repetitive areas are retained, the image content that needs to be recognized is reduced as much as possible, and the recognition efficiency of queuing vehicles is improved.
  • the vehicle queuing area is divided in the environment image, the areas in the environment image except the vehicle queuing area are blurred, and the queuing vehicle identification is performed on the blurred environment image.
  • the vehicle queuing area is a predetermined area of the power exchange station for vehicle queuing.
  • the vehicle queuing area can be divided from the environment image based on the template image, and the queuing path in the environment image can also be identified to divide the vehicle queuing area.
  • the vehicle queuing area is an area formed by the natural queuing of vehicles.
  • the vehicle queuing area can be identified based on the positional relationship in which the heads of multiple vehicles are connected.
  • the non-queuing area in the environmental image is blurred, so as to effectively exclude other non-exchange vehicles, and further improve the accuracy of identifying the number of queues.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • it can effectively eliminate the interference of other non-swap vehicles in the invalid area to the recognition of the swapped vehicles, and improve the recognition accuracy of the queuing vehicles.
  • it can avoid the workload of identifying invalid areas and improve the identification efficiency of queuing vehicles.
  • Step 103 Determine the queued number of the swap station according to the identified queued vehicles.
  • the number of vehicles is counted to determine the number of queuing vehicles at the swap station, wherein whether the identified vehicles are queuing vehicles can be further determined by a preset method, for example, it can be excluded Vehicles that have left after battery replacement or vehicles that are currently being accurately replaced, etc.
  • Step 104 Verify the determined number of queues.
  • the queuing quantity obtained according to the positioning information of the vehicles entering the network is obtained, and the queuing quantity determined according to the identified queuing vehicles is checked by using the queuing quantity obtained according to the positioning information of the vehicles entering the network.
  • the positioning information of the vehicles entering the network determine the number of vehicles whose current position is in the vehicle queuing area and use it as the queuing quantity, and judge whether the queuing quantity is consistent with the queuing quantity output in step 103, if yes, the verification is successful, if not, Prompt validation failed, and choose which data to use as the queue amount.
  • the number of queues obtained from the positioning information of the vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the number of queues determined by the environment image can also be verified through the number of queues obtained by manual statistics, so as to further improve the accuracy of the data, and which data to use can be further selected.
  • Step 105 recommending a swap station according to the verified number of queues.
  • the determined number of queues is uploaded to the cloud, so that when the cloud receives a request for finding a replacement station, the cloud recommends the replacement station according to the number of queues reported by each replacement station.
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting for power swap can easily select swap stations with fewer queues, thereby saving user time and improving power swap efficiency.
  • the waiting time for battery swapping can also be determined according to the identified number of queues, and the number of queues and/or the waiting time for battery swapping is output to the display terminal or user terminal of the battery swapping station.
  • Queuing users can choose a suitable power exchange station for power exchange according to the waiting time for power exchange and the number of queues, thus improving the user experience.
  • the battery swap station when the method is applied to a battery swap station, can upload the identified queue number of the battery swap station to the cloud according to a time period, so that the cloud can report to the driver user according to the queue number of the battery swap station when necessary. Swap stations are recommended.
  • the swapping station can upload the new queued number to the cloud when it recognizes that the number of queues at the swapping station changes, so that the cloud can recommend the swapping station to the driving user according to the number of queues at the swapping station when necessary.
  • the method for identifying the number of queued vehicles at the swapping station automatically acquires an environmental image of the vehicle queuing area of the swapping station, identifies the queuing vehicles from the environmental image to determine the number of queued vehicles at the swapping station, and further verifies the number of queues. At the same time, it can also recommend suitable power exchange stations to users based on the number of queues, which can effectively improve the identification efficiency of the number of queues, and can also effectively save the human resource consumption of power exchange stations and reduce costs.
  • the present embodiment provides a method for identifying the number of vehicles queued at a swap station, including: acquiring an environmental image of a vehicle queuing area at the swap station; identifying queuing vehicles included in the environmental image; The vehicle determines the number of queues at the swap station.
  • the method can be applied to a power station or the cloud, but its application scenarios are not specifically limited, and corresponding selection and adjustment can be made according to actual needs.
  • the identification efficiency of the queuing quantity can be effectively improved, and the It can effectively save the human resource consumption of the swap station and reduce the cost.
  • the method for identifying the number of vehicles queuing at a power exchange station mainly includes the following steps:
  • Step 101 Obtain an environmental image of the vehicle queuing area of the swap station.
  • the vehicle queuing area may be a predetermined area in the power exchange station for vehicle queuing, or may be an area formed by the natural queuing of vehicles.
  • the local server of the swapping station invokes the camera of the swapping station to collect the video stream of the vehicle queuing area of the swapping station in real time, and converts the video stream into an image to capture the vehicles of the swapping station therefrom. Environmental image of the queuing area.
  • image recognition can be performed through the local server of the power exchange station, which can reduce bandwidth pressure and cloud computing pressure.
  • multiple cameras at different positions of the battery swap station are called to collect real-time video streams of the vehicle queuing area of the battery swap station from multiple perspectives, and the video streams are respectively converted into images for intercepting from them.
  • Video streams from different perspectives are collected by multiple cameras to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • a camera can be used to change the viewing angle in real time to collect video streams from different viewing angles, so as to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the power exchange station when the method is applied to the cloud, the power exchange station will report the collected video stream to the cloud server, and may also report the captured environment image of the vehicle queuing area of the power exchange station to the cloud server, and the cloud server will receive the video Stream to get the environment image or receive the environment image directly.
  • image recognition can be performed through the cloud server to facilitate the unified management and scheduling of data, and can effectively avoid data tampering and other situations.
  • Step 102' input the environment image into the trained queuing vehicle recognition model to output the queuing vehicle in the environment image.
  • the environment image is input into the queuing vehicle identification model, and the queuing vehicle identification model is used to identify the queuing vehicles in the environment image.
  • the queuing vehicle identification model is trained according to the environment image of the vehicle queuing area of the swap station and the corresponding training labels.
  • the initialized queuing vehicle recognition model which can be a neural network model for image recognition.
  • this embodiment does not specifically limit the model type, and corresponding Select and adjust.
  • image processing is further performed on the obtained environment image, that is, the model training sample.
  • the image processing includes at least one of compression processing, encoding processing and encryption processing, for example, compressing the environmental image, and converting the compressed image into Base64 (the most common on the network for transmitting 8Bit bytes One of the encoding methods of the code), and then compress and encrypt the Base64 encoding.
  • this embodiment does not specifically limit the image processing mode, and corresponding selections can be made according to actual needs.
  • the environment images from multiple viewing angles are de-overlapped to obtain an intermediate image
  • the vehicle queuing area is divided in the intermediate image
  • the areas in the intermediate image except the vehicle queuing area are blurred
  • the blurred area is
  • the processed intermediate images are used as model training samples.
  • the environment images are cropped and merged.
  • the environment images with repeated regions are cropped and merged to improve the comprehensiveness of model training.
  • the vehicle queuing area is a predetermined area of the power exchange station for vehicle queuing.
  • the vehicle queuing area can be divided from the environment image based on the template image, and the queuing path in the environment image can also be identified to divide the vehicle queuing area.
  • the vehicle queuing area is an area formed by the natural queuing of vehicles.
  • the vehicle queuing area can be identified based on the positional relationship in which the heads of multiple vehicles are connected.
  • the non-queuing area in the model training sample is blurred, so as to effectively exclude other non-battery-swappable vehicles, and further improve the accuracy of model recognition.
  • the queuing vehicle recognition model is trained according to the model training samples and the corresponding training labels.
  • model training samples are input into the queuing vehicle recognition model, the predicted recognition results output by the queuing vehicle recognition model are obtained, and the model parameters of the queuing vehicle recognition model are optimized according to the difference between the predicted recognition results and the training labels corresponding to the model training samples.
  • the queuing vehicle identification model when the accuracy rate of the queuing vehicle identification model is lower than a preset threshold, the queuing vehicle identification model is updated.
  • the number of queues obtained according to the positioning information of vehicles entering the network or the number of queues identified manually can also be obtained, and the number of queues obtained according to the positioning information of vehicles entering the network or the number of queues manually identified can be used to verify the queues determined according to the identified vehicles in queue. number to obtain the accuracy of the queuing vehicle recognition model.
  • This embodiment does not specifically limit the preset threshold, and can be set accordingly according to actual needs.
  • the queuing vehicle identification model when the usage duration of the queuing vehicle identification model is longer than a preset duration, the queuing vehicle identification model is updated.
  • This embodiment does not specifically limit the preset duration, and can be set according to actual needs.
  • the model is updated in time, and the actual demand is followed up in real time, thereby further improving the accuracy and efficiency of model recognition.
  • the step of training the model is performed by a dedicated server.
  • the dedicated server sends the queuing vehicle identification model to each power exchange station and/or power exchange cloud, so that the power exchange station and / or the power exchange cloud uses the queuing vehicle identification model to identify the environment image and output the number of queues.
  • the trained model is sent to each power exchange station and/or power exchange cloud, so that each power exchange station and/or power exchange cloud can perform model identification, so as to arrange the model training process and use process in an orderly manner, effectively saving costs, Improved the stability of model recognition.
  • the model deployed on the local server of the swap station can quickly and efficiently identify the queuing vehicles, which saves operating costs, and can also be implemented through the local LAN when the Internet is abnormal. Vehicle identification, thus ensuring identification stability and reliability.
  • Step 103' determine the number of queues at the power exchange station through the identification result output by the queued vehicle identification model.
  • the number of queues at the swapping station is determined by the recognition result after model recognition, and whether the identified vehicles are queued vehicles can be further determined by a preset method. Vehicles with accurate battery replacement, etc.
  • Step 104' verify the determined number of queues.
  • the queuing quantity obtained according to the positioning information of the vehicles entering the network is obtained, and the queuing quantity determined according to the identified queuing vehicles is checked by using the queuing quantity obtained according to the positioning information of the vehicles entering the network.
  • the positioning information of the vehicles entering the network determine the number of vehicles whose current position is in the vehicle queuing area and use it as the queuing quantity, and judge whether the queuing quantity is consistent with the queuing quantity output in step 103, if yes, the verification is successful, if not, Prompt validation failed, and choose which data to use as the queue amount.
  • the number of queues obtained from the positioning information of the vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the number of queues determined by the environment image can also be verified through the number of queues obtained by manual statistics, so as to further improve the accuracy of the data, and which data to use can be further selected.
  • Step 105' recommending the swap station according to the number of queues after verification.
  • the determined number of queues is uploaded to the cloud, so that when the cloud receives a request for finding a replacement station, the cloud recommends the replacement station according to the number of queues reported by each replacement station.
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting for power swap can easily select swap stations with fewer queues, thereby saving user time and improving power swap efficiency.
  • the waiting time for battery swapping can also be determined according to the identified number of queues, and the number of queues and/or the waiting time for battery swapping is output to the display terminal or user terminal of the battery swapping station.
  • Queuing users can choose a suitable power exchange station for power exchange according to the waiting time for power exchange and the number of queues, thus improving the user experience.
  • the battery swap station when the method is applied to a battery swap station, can upload the identified queue number of the battery swap station to the cloud according to a time period, so that the cloud can report to the driver user according to the queue number of the battery swap station when necessary. Swap stations are recommended.
  • the swapping station can upload the new queued number to the cloud when it recognizes that the number of queues at the swapping station changes, so that the cloud can recommend the swapping station to the driving user according to the number of queues at the swapping station when necessary.
  • the method for identifying the number of queued vehicles at the swapping station automatically acquires an environmental image of the vehicle queuing area of the swapping station, identifies the queuing vehicles from the environmental image to determine the number of queued vehicles at the swapping station, and further verifies the number of queues. At the same time, it can also recommend suitable power exchange stations to users based on the number of queues, which can effectively improve the identification efficiency of the number of queues, and can also effectively save the human resource consumption of power exchange stations and reduce costs.
  • the present embodiment provides a system for identifying the number of vehicles queued at a swap station, including: an environmental image acquisition module configured to acquire an environmental image of the vehicle queuing area of the swap station; a queued vehicle identification module, which is The queuing number determination module is configured to identify the queuing vehicles included in the environment image;
  • the system for identifying the number of vehicles queued at a power exchange station mainly includes an environmental image acquisition module 21 , a queue vehicle identification module 22 , a queue number determination module 23 , and a queue number determination module 23 .
  • the verification module 24 and the swap station recommendation module 25, the system utilizes the method for identifying the number of vehicles queued at the swap station as in the above-mentioned Embodiment 1.
  • the environmental image acquisition module 21 is configured to call the camera of the swapping station to acquire the video stream of the vehicle queuing area of the swapping station in real time, and convert the video stream into an image to intercept the vehicle queuing of the swapping station therefrom. Environment image of the area.
  • image recognition can be performed through the local server of the power exchange station, which can reduce bandwidth pressure and cloud computing pressure.
  • the environmental image acquisition module 21 is further configured to call a plurality of cameras in different positions of the battery swap station to obtain real-time video streams of the vehicle queuing area of the battery swap station from multiple perspectives, and convert the video streams into The image is used to extract the environment image including the vehicle queuing area from multiple viewing angles.
  • Video streams from different perspectives are collected by multiple cameras to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the environmental image acquisition module 21 is configured to acquire the video stream reported by the swapping station, and can also acquire the environmental image of the vehicle queuing area reported by the swapping station.
  • image recognition can be performed through the cloud server to facilitate the unified management and scheduling of data, which can effectively avoid data tampering and other situations.
  • the queuing vehicle identification module 22 is configured to perform image processing on the acquired environment image, and identify the queuing vehicle from the image-processed environment image through image recognition technology. This embodiment does not specifically limit the image recognition method, as long as the corresponding functions can be realized, corresponding selections can be made according to actual needs.
  • the image processing includes at least one of compression processing, encoding processing and encryption processing.
  • the environmental image is compressed, and the compressed image is converted into Base64 encoding, and then the Base64 encoding is compressed and encrypted.
  • this embodiment does not specifically limit the image processing mode, and corresponding selections can be made according to actual needs.
  • Corresponding image processing is performed on the obtained environmental image, so as to reduce the amount of transmitted data, thereby effectively saving transmission network bandwidth, reducing costs, and effectively ensuring data security.
  • the queuing vehicle identification module 22 is configured to selectively call a third-party image identification service for vehicle identification, thereby effectively saving transmission network bandwidth and reducing costs.
  • the queuing vehicle identification module 22 is configured to identify the queuing vehicles after deduplicating the acquired environment image.
  • the queuing vehicle identification module 22 is configured to cut the repeated areas and merge the environmental images, and compress and/or encode the combined environmental images, The queuing vehicle identification is then performed on the target image obtained after compression and/or encoding.
  • the environment images with repeated areas are cropped and merged, the effective and non-repetitive areas are retained, the image content that needs to be recognized is reduced as much as possible, and the recognition efficiency of queuing vehicles is improved.
  • the queuing vehicle identification module 22 is configured to divide the vehicle queuing area in the environment image, blur the area except the vehicle queuing area in the environment image, and perform the queuing vehicle on the blurred environment image. identify.
  • the vehicle queuing area is a predetermined area of the power exchange station for vehicle queuing.
  • the vehicle queuing area can be divided from the environment image based on the template image, and the queuing path in the environment image can also be identified to divide the vehicle queuing area.
  • the vehicle queuing area is an area formed by the natural queuing of vehicles.
  • the vehicle queuing area can be identified based on the positional relationship in which the heads of multiple vehicles are connected.
  • the non-queuing area in the environmental image is blurred, so as to effectively exclude other non-exchange vehicles, and further improve the accuracy of identifying the number of queues.
  • the vehicle queuing area is divided in the environmental image, and the non-queuing area in the environmental image is blurred.
  • it can effectively eliminate the interference of other non-swap vehicles in the invalid area to the recognition of the swapped vehicles, and improve the recognition accuracy of the queuing vehicles.
  • it can avoid the workload of identifying invalid areas and improve the identification efficiency of queuing vehicles.
  • the queuing quantity determination module 23 is configured to, after identifying the queuing vehicles from the environment image, perform statistics on the number of vehicles to determine the queuing quantity of the swap station, wherein whether the identified vehicles are queuing vehicles can be further determined by a preset method, for example , which can exclude vehicles that have left after battery replacement or vehicles that are currently accurately battery replacement.
  • the queuing quantity verification module 24 is configured to obtain the queuing quantity obtained according to the positioning information of the vehicles entering the network, and verify the queuing quantity determined according to the identified queuing vehicles using the queuing quantity obtained according to the positioning information of the entering vehicles.
  • the queuing quantity verification module 24 is configured to, according to the positioning information of the vehicles entering the network, determine the number of vehicles whose current position is in the vehicle queuing area as the queuing quantity, and judge whether the queuing quantity is the same as the queuing quantity outputted by the queuing quantity determining module 23 If it is consistent, if it is, the verification is successful, if not, it prompts the verification to fail, and choose which data to use as the queue quantity.
  • the number of queues obtained from the positioning information of the vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the number of queues determined by the environment image can also be verified through the number of queues obtained by manual statistics, so as to further improve the accuracy of the data, and which data to use can be further selected.
  • the swapping station recommending module 25 is configured to receive the determined queuing quantity, and recommend swapping stations according to the queuing quantity reported by each swapping station when receiving the swapping station searching request.
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting for power swap can easily select swap stations with fewer queues, thereby saving user time and improving power swap efficiency.
  • the system can also determine the waiting time for battery swapping according to the identified number of queues, and output the number of queues and/or the waiting time for battery swapping to the display terminal or user terminal of the battery swapping station.
  • Queuing users can choose a suitable power exchange station for power exchange according to the waiting time for power exchange and the number of queues, thus improving the user experience.
  • the system for identifying the number of vehicles queuing at the swapping station provided by this embodiment automatically acquires an environmental image of the vehicle queuing area of the swapping station to identify the queuing vehicles from the environmental image to determine the number of queuing vehicles at the swapping station, and further verifies the queuing quantity.
  • it can also recommend suitable power exchange stations to users based on the number of queues, which can effectively improve the identification efficiency of the number of queues, and can also effectively save the human resource consumption of power exchange stations and reduce costs.
  • the present embodiment provides a system for identifying the number of vehicles queued at a swap station, including: an environmental image acquisition module configured to acquire an environmental image of the vehicle queuing area of the swap station; a queued vehicle identification module, which is The queuing number determination module is configured to identify the queuing vehicles included in the environment image;
  • the system for identifying the number of vehicles queuing at a power exchange station mainly includes an environmental image acquisition module 21 , a queuing vehicle identification module 22 , a queuing quantity determining module 23 , and a queuing module.
  • the number verification module 24, the swap station recommendation module 25 and the model training module 26, the system uses the method for identifying the number of vehicles queued at the swap station as described in the above-mentioned Embodiment 2.
  • the environmental image acquisition module 21 is configured to call the camera of the swapping station to acquire the video stream of the vehicle queuing area of the swapping station in real time, and convert the video stream into an image to intercept the vehicle queuing of the swapping station therefrom. Environment image of the area.
  • image recognition can be performed through the local server of the power exchange station, which can reduce bandwidth pressure and cloud computing pressure.
  • the environmental image acquisition module 21 is further configured to call multiple cameras of the swap station at different positions to acquire in real time the video streams of the vehicle queuing area of the swap station from multiple perspectives, and convert the video streams into The image can be cut out of the environment image including the vehicle queuing area from multiple viewing angles.
  • Video streams from different perspectives are collected by multiple cameras to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • a camera can be used to change the viewing angle in real time to collect video streams from different viewing angles, so as to ensure that the environmental image can fully cover the vehicle queuing area as much as possible, thereby further improving the accuracy of identifying the number of queues.
  • the environmental image acquisition module 21 is configured to acquire the video stream reported by the swapping station, and can also acquire the environmental image of the vehicle queuing area reported by the swapping station.
  • image recognition can be performed through the cloud server to facilitate the unified management and scheduling of data, which can effectively avoid data tampering and other situations.
  • the queuing vehicle identification module 22 is configured to input the environment image into the queuing vehicle identification model, identify the queuing vehicle in the environment image through the queuing vehicle identification model, and the queuing vehicle identification model calls the model training module 26 according to the environment of the vehicle queuing area of the swap station. Images and corresponding training labels are trained.
  • the model training module 26 is configured to obtain an initialized queuing vehicle recognition model, which can be a neural network model for image recognition. As long as the vehicle recognition function can be realized, this embodiment does not specifically limit the model type, and can be based on Select and adjust accordingly according to actual needs.
  • the model training module 26 is further configured to collect environmental images including vehicle queuing areas from multiple perspectives captured based on video streams captured from multiple perspectives, and de-overlap the environmental images from multiple perspectives and use them as model training samples. , and obtain the training labels corresponding to the training samples of each model.
  • model training module 26 is further configured to further perform image processing on the acquired environment image, that is, the model training sample.
  • the image processing includes at least one of compression processing, encoding processing and encryption processing, for example, compressing the environmental image, and converting the compressed image into Base64 (the most common on the network for transmitting 8Bit bytes One of the encoding methods of the code), and then compress and encrypt the Base64 encoding.
  • this embodiment does not specifically limit the image processing mode, and corresponding selections can be made according to actual needs.
  • the model training module 26 is further configured to de-overlap the environmental images from multiple perspectives to obtain an intermediate image, divide the vehicle queuing area in the intermediate image, and divide the intermediate image except for the vehicle queuing area.
  • the outer region is blurred, and the blurred intermediate image is used as a model training sample.
  • the vehicle queuing area may be a predetermined area in the power exchange station for vehicle queuing, or may be an area formed by the natural queuing of vehicles.
  • the environment images are cropped and merged.
  • the environment images with repeated regions are cropped and merged to improve the comprehensiveness of model training.
  • the vehicle queuing area is a predetermined area of the power exchange station for vehicle queuing.
  • the vehicle queuing area can be divided from the environment image based on the template image, and the queuing path in the environment image can also be identified to divide the vehicle queuing area.
  • the vehicle queuing area is an area formed by the natural queuing of vehicles.
  • the vehicle queuing area can be identified based on the positional relationship in which the heads of multiple vehicles are connected.
  • the non-queuing area in the model training sample is blurred, so as to effectively exclude other non-battery-swappable vehicles, and further improve the accuracy of model recognition.
  • the model training module 26 is also configured to train the queuing vehicle recognition model based on the model training samples and the corresponding training labels.
  • model training samples are input into the queuing vehicle recognition model, the predicted recognition results output by the queuing vehicle recognition model are obtained, and the model parameters of the queuing vehicle recognition model are optimized according to the difference between the predicted recognition results and the corresponding training labels of the model training samples.
  • the model training module 26 is further configured to update the queuing vehicle recognition model when the accuracy of the queuing vehicle recognition model is lower than a preset threshold.
  • a preset threshold the number of queues obtained according to the positioning information of vehicles entering the network or the number of queues manually identified can also be obtained, and the number of queues obtained according to the positioning information of vehicles entering the network or the number of queues manually identified can be used to check the queues determined according to the identified vehicles in queue. number to obtain the accuracy of the queuing vehicle recognition model.
  • This embodiment does not specifically limit the preset threshold, and can be set accordingly according to actual needs.
  • the model training module 26 is further configured to update the queuing vehicle identification model when the usage duration of the queuing vehicle identification model is longer than a preset duration. needs to be set accordingly.
  • the model is updated in time, and the actual demand is followed up in real time, thereby further improving the accuracy and efficiency of model recognition.
  • the model training module 26 can be configured on a dedicated server. After the training of the queuing vehicle identification model is completed, the dedicated server will deliver the queuing vehicle identification model to each power exchange station and/or power exchange cloud, so that the exchange The power station and/or the power exchange cloud adopts the queuing vehicle recognition model to identify the number of queues in the environment image output.
  • the trained model is sent to each power exchange station and/or power exchange cloud, so that each power exchange station and/or power exchange cloud can perform model identification, so as to arrange the model training process and use process in an orderly manner, effectively saving costs, Improved the stability of model recognition.
  • the model deployed on the local server of the swap station can quickly and efficiently identify the queuing vehicles, which saves operating costs, and can also be implemented through the local LAN when the Internet is abnormal. Vehicle identification, thus ensuring identification stability and reliability.
  • the queuing quantity determination module 23 is configured to determine the queuing quantity of the swapping station through the recognition result of the vehicle recognition model, and can further determine whether the identified vehicle is a queuing vehicle by a preset method, for example, it can be excluded that the vehicle leaves after changing the battery. vehicles or vehicles that are currently being accurately replaced, etc.
  • the queuing quantity verification module 24 is configured to obtain the queuing quantity obtained according to the positioning information of the vehicles entering the network, and verify the queuing quantity determined according to the identified queuing vehicles using the queuing quantity obtained according to the positioning information of the entering vehicles.
  • the queuing quantity verification module 24 is configured to, according to the positioning information of the vehicles entering the network, determine the number of vehicles whose current position is in the vehicle queuing area as the queuing quantity, and judge whether the queuing quantity is the same as the queuing quantity outputted by the queuing quantity determining module 23 If it is consistent, if it is, the verification is successful, if not, it prompts the verification to fail, and choose which data to use as the queue quantity.
  • the number of queues obtained from the positioning information of the vehicles entering the network is verified by the number of queues determined by the environmental image, thereby further improving the accuracy of the data.
  • the number of queues determined by the environment image can also be verified through the number of queues obtained by manual statistics, so as to further improve the accuracy of the data, and which data to use can be further selected.
  • the swapping station recommending module 25 is configured to receive the determined queuing quantity, and recommend swapping stations according to the queuing quantity reported by each swapping station when receiving the swapping station searching request.
  • the number of queues determined by the environment image can be used to recommend swap stations to users, so that users in queue or users waiting for power swap can easily select swap stations with fewer queues, thereby saving user time and improving power swap efficiency.
  • the system can also determine the waiting time for battery swapping according to the identified number of queues, and output the number of queues and/or the waiting time for battery swapping to the display terminal or user terminal of the swapping station.
  • Queuing users can choose a suitable power exchange station for power exchange according to the waiting time for power exchange and the number of queues, thus improving the user experience.
  • the battery swap station when the system is applied to a battery swap station, can upload the identified queue number of the battery swap station to the cloud according to a time period, so that the cloud can report to the driver user according to the queue number of the battery swap station when necessary. Swap stations are recommended.
  • the swapping station can upload the new queued number to the cloud when it recognizes that the number of queues at the swapping station changes, so that the cloud can recommend the swapping station to the driving user according to the number of queues at the swapping station when necessary.
  • the system for identifying the number of vehicles queuing at the swapping station provided by this embodiment automatically acquires an environmental image of the vehicle queuing area of the swapping station to identify the queuing vehicles from the environmental image to determine the number of queuing vehicles at the swapping station, and further verifies the queuing quantity.
  • it can also recommend suitable power exchange stations to users based on the number of queues, which can effectively improve the identification efficiency of the number of queues, and can also effectively save the human resource consumption of power exchange stations and reduce costs.
  • FIG. 5 is a schematic structural diagram of an electronic device provided according to this embodiment.
  • the electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor.
  • the processor executes the program, the method for identifying the number of vehicles queuing at the power exchange station in Embodiment 1 or Embodiment 2 is implemented.
  • the electronic device 30 shown in FIG. 5 is only an example, and should not impose any limitations on the functions and scope of use of the embodiments of the present application.
  • the electronic device 30 may take the form of a general-purpose computing device, for example, it may be a server device.
  • Components of the electronic device 30 may include, but are not limited to, the above-mentioned at least one processor 31 , the above-mentioned at least one memory 32 , and a bus 33 connecting different system components (including the memory 32 and the processor 31 ).
  • the bus 33 includes a data bus, an address bus and a control bus.
  • Memory 32 may include volatile memory, such as random access memory (RAM) 321 and/or cache memory 322 , and may further include read only memory (ROM) 323 .
  • RAM random access memory
  • ROM read only memory
  • the memory 32 may also include a program/utility 325 having a set (at least one) of program modules 324 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, which An implementation of a network environment may be included in each or some combination of the examples.
  • the processor 31 executes various functional applications and data processing by running the computer program stored in the memory 32, such as the method for identifying the number of vehicles queuing at the swap station in the above embodiment of the present application.
  • the electronic device 30 may also communicate with one or more external devices 34 (eg, keyboards, pointing devices, etc.). Such communication may take place through input/output (I/O) interface 35 .
  • the model-generating device 30 may also communicate with one or more networks, such as a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet, through a network adapter 36. As shown in FIG. 5 , the network adapter 36 communicates with the other modules of the model generation device 30 via the bus 33 .
  • LAN local area network
  • WAN wide area network
  • public network such as the Internet
  • model-generated device 30 may be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk) array) systems, tape drives, and data backup storage systems.
  • This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the steps in the method for identifying the number of vehicles queuing at the swap station in the above-mentioned Embodiment 1 or Embodiment 2 .
  • the readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any of the above suitable combination.
  • the present application can also be implemented in the form of a program product, which includes program codes.
  • the program product runs on a terminal device, the program code is used to cause the terminal device to execute the conversion in the above embodiment. Steps in a method for identifying the number of vehicles queued at a power station.
  • the program code for executing the present application can be written in any combination of one or more programming languages, and the program code can be completely executed on the user equipment, partially executed on the user equipment, as an independent software
  • the package executes, partly on the user device, partly on the remote device, or entirely on the remote device.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Traffic Control Systems (AREA)

Abstract

本申请公开了一种换电站排队车辆的数量识别方法、***、设备及介质，该识别方法包括：获取换电站的车辆排队区域的环境图像；识别所述环境图像中包括的排队车辆；根据识别出的排队车辆确定所述换电站的排队数量。本申请通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。

Description

换电站排队车辆的数量识别方法、***、设备及介质
本申请要求申请日为2020/12/31的中国专利申请2020116188073和中国专利申请2020116187776的优先权。本申请引用上述中国专利申请的全文。
技术领域
本申请涉及换电技术领域，尤其涉及一种换电站排队车辆的数量识别方法、***、设备及介质。
背景技术
电动汽车等新能源车辆能够有效的缓解空气污染，也是未来汽车发展的主要方向。换电站是一种快速高效为新能源车辆补充电能的场所，换电站不仅能够节省新能源车辆对充电站/桩等设备的要求，而且能够提高设备利用率，能够最大限度的满足新能源司机的需求。
在换电站内可为电动汽车提供全自动换电和快速充电服务，随着日益增长的电动车需求使得换电用户逐步增多，特别是出租车，为了节省时间成本，更倾向于高效换电服务，由于换电站数量以及电池数量有限，就可能需要换电用户在换电站内排队一定的时间才能为电动车换电。
目前，换电车辆排队数量一般由换电站的员工人工进行统计，导致浪费人力资源，增加换电站的成本，而且工作效率低。
申请内容
本申请要解决的技术问题是为了克服现有技术中换电车辆排队数量采用人工统计方式，导致浪费人力资源，降低效率的缺陷，提供一种换电站排队车辆的数量识别方法、***、设备及介质。
本申请是通过下述技术方案来解决上述技术问题：
根据本申请的一实施方式，提供一种换电站排队车辆的数量识别方法，包括：
获取换电站的车辆排队区域的环境图像；
识别所述环境图像中包括的排队车辆；
根据识别出的排队车辆确定所述换电站的排队数量。
在本方案中，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中 识别出排队车辆来确定换电站的排队数量，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
可选地，所述识别所述环境图像中包括的排队车辆，包括：
在所述环境图像中划分出车辆排队区域；
将所述环境图像中除所述车辆排队区域之外的区域模糊处理；
对模糊处理后的所述环境图像进行排队车辆识别。
在本方案中，在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
可选地，所述获取换电站的车辆排队区域的环境图像，包括：
获取从多个视角拍摄换电站的车辆排队区域的视频流；
从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
所述识别所述环境图像中包括的排队车辆，包括：
在对所述环境图像进行去重后，进行排队车辆识别。
在本方案中，通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
可选地，所述在对所述环境图像进行去重后，进行排队车辆识别，包括：
当多个视角下的环境图像之间存在重复区域时，对所述环境图像进行重复区域裁剪后合并；
对合并后的环境图像进行压缩和/或编码；
对压缩和/或编码后得到的目标图像进行排队车辆识别。
在本方案中，对存在重复区域的环境图像进行重复区域裁剪后合并，保留有效且没有重复的区域，尽可能减少需要进行识别的图像内容，提高排队车辆的识别效率；并且通过对获取到的环境图像进行压缩和/或编码的图像处理，可减少进行识别的图像的数据量，能够进一步提高排队车辆的识别效率。
可选地，当应用于换电站时，所述获取换电站的车辆排队区域的环境图像，包括：
获取换电站摄像头采集的视频流；
从所述视频流中截取出包括车辆排队区域的环境图像；
当应用于云端时，所述获取换电站的车辆排队区域的环境图像，包括：
接收换电站上报的所述换电站的车辆排队区域的环境图像。
在本方案中，应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力；应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
可选地，当应用于换电站时，所述方法还包括：
将所述排队数量上传至所述云端，以使所述云端在接收到换电站查找请求时，根据各所述换电站上报的排队数量推荐换电站。
在本方案中，通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
可选地，还包括：
获取根据入网车辆的定位信息得到的排队数量；
采用根据入网车辆的定位信息得到的排队数量校验所述根据识别出的排队车辆确定的排队数量。
在本方案中，通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
可选地，所述识别所述环境图像中包括的排队车辆，包括：
将所述环境图像输入排队车辆识别模型，通过所述排队车辆识别模型识别所述环境图像中的排队车辆；所述排队车辆识别模型根据换电站的车辆排队区域的环境图像和相应的训练标签训练得到；
所述根据识别出的排队车辆确定所述换电站的排队数量，包括：
通过所述排队车辆识别模型输出的识别结果确定所述换电站的排队数量。
在本方案中，事先采用历史的环境图像训练排队车辆识别模型，使得排队车辆识别模型学会基于环境图像进行排队车辆识别的能力，再通过训练得到的模型识别出环境图像中的排队车辆，进一步提升了识别准确性和效率。
可选地，所述排队车辆识别模型的训练步骤包括：
获取初始化的排队车辆识别模型；
获取从多个视角拍摄换电站的车辆排队区域的视频流；
从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
对多个视角下的环境图像进行去重合并后用作模型训练样本；
获取各所述模型训练样本相应的训练标签；
根据所述模型训练样本和相应的训练标签，训练所述排队车辆识别模型。
在本方案中，通过多个摄像头采集不同视角下的视频流以从中截取出环境图像，并去重合并后用作模型训练样本，以使环境图像能够尽量全面覆盖车辆排队区域，从而提升模型识别的准确性，而且提升了训练模型的稳定性。
可选地，所述对多个视角下的环境图像进行去重合并后用作模型训练样本，包括：
对多个视角下的环境图像进行去重合并得到中间图像；
在所述中间图像中划分出车辆排队区域；
将所述中间图像中除所述车辆排队区域之外的区域模糊处理；
将模糊处理后的所述中间图像用作模型训练样本。
在本方案中，在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
可选地，所述根据所述模型训练样本和相应的训练标签，训练所述排队车辆识别模型，包括：
将所述模型训练样本压缩和/或编码后输入所述排队车辆识别模型；
获取所述排队车辆识别模型输出的预测识别结果；
根据所述预测识别结果和所述模型训练样本相应的训练标签之间的差异，优化所述排队车辆识别模型的模型参数。
在本方案中，对模型训练样本进行相应的图像处理，以使得便于模型训练，而且减小传输的数据量，从而提升模型训练的稳定性，有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
可选地，所述排队车辆识别模型的更新步骤包括：
在所述排队车辆识别模型的准确率低于预设阈值时，更新所述排队车辆识别模型；
或，
在所述排队车辆识别模型的使用时长高于预设时长时，更新所述排队车辆识别模型。
在本步骤中，当符合更新条件时，及时进行模型更新，实时跟进实际需求，从而进一步提升模型识别的准确性和效率。
可选地，所述训练步骤通过专用服务器执行；所述排队车辆识别模型在训练完成后，由专用服务器将所述排队车辆识别模型下发至各所述换电站和/或换电云端，以使所述换电站和/或换电云端采用所述排队车辆识别模型识别所述环境图像输出排队数量。
在本方案中，将训练好的模型下发至各换电站和/或换电云端，以使得各换电站和/或 换电云端执行模型识别，从而有序安排模型训练过程和使用过程，有效地节省了成本，提升了模型识别的稳定性。
根据本申请的一实施方式，提供一种换电站排队车辆的数量识别***，包括：
环境图像获取模块，被配置为获取换电站的车辆排队区域的环境图像；
排队车辆识别模块，被配置为识别所述环境图像中包括的排队车辆；
排队数量确定模块，被配置为根据识别出的排队车辆确定所述换电站的排队数量。
在本方案中，通过实时采集到的换电站车辆排队区域的环境图像来方便、及时且准确地确定出换电站的排队数量，以使得基于排队数量确定排队时间、顺序等换电策略，同时还可以直观地展示给待换电的用户，从而有效地节省了人力资源，降低了换电站的成本，提升了工作效率。
可选地，所述排队车辆识别模块被配置为：
在所述环境图像中划分出车辆排队区域；
将所述环境图像中除所述车辆排队区域之外的区域模糊处理；
对模糊处理后的所述环境图像进行排队车辆识别。
在本方案中，在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
可选地，所述环境图像获取模块被配置为：
获取从多个视角拍摄换电站的车辆排队区域的视频流；
从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
所述排队车辆识别模块被配置为：
在对所述环境图像进行去重后，进行排队车辆识别。
在本方案中，通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
可选地，所述排队车辆识别模块还被配置为：
当多个视角下的环境图像之间存在重复区域时，对所述环境图像进行重复区域裁剪后合并；
对合并后的环境图像进行压缩和/或编码；
对压缩和/或编码后得到的目标图像进行排队车辆识别。
在本方案中，对存在重复区域的环境图像进行重复区域裁剪后合并，保留有效且没 有重复的区域，尽可能减少需要进行识别的图像内容，提高排队车辆的识别效率；并且通过对获取到的环境图像进行压缩和/或编码的图像处理，可减少进行识别的图像的数据量，能够进一步提高排队车辆的识别效率。
可选地，当该***应用于换电站时，所述环境图像获取模块被配置为：
获取换电站摄像头采集的视频流；
从所述视频流中截取出包括车辆排队区域的环境图像；
当该***应用于云端时，所述环境图像获取模块被配置为：
接收换电站上报的所述换电站的车辆排队区域的环境图像。
在本方案中，应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力；应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
可选地，当该***应用于换电站时，所述排队数量确定模块还被配置为：
将所述排队数量上传至所述云端，以使所述云端在接收到换电站查找请求时，根据各所述换电站上报的排队数量推荐换电站。
在本方案中，通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
可选地，该***还包括排队数量验证模块；
所述排队数量验证模块被配置为：
获取根据入网车辆的定位信息得到的排队数量；
采用根据入网车辆的定位信息得到的排队数量校验所述根据识别出的排队车辆确定的排队数量。
在本方案中，通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
可选地，所述排队车辆识别模块被配置为：将所述环境图像输入排队车辆识别模型，通过所述排队车辆识别模型识别所述环境图像中的排队车辆；所述排队车辆识别模型根据换电站的车辆排队区域的环境图像和相应的训练标签训练得到；
所述排队数量确定模块被配置为：通过所述排队车辆识别模型输出的识别结果确定所述换电站的排队数量。
在本方案中，事先采用历史的环境图像训练排队车辆识别模型，使得排队车辆识别模型学会基于环境图像进行排队车辆识别的能力，再通过训练得到的模型识别出环境图 像中的排队车辆，进一步提升了识别准确性和效率。
可选地，该***还包括模型训练模块；
所述模型训练模块被配置为：
获取初始化的排队车辆识别模型；
获取从多个视角拍摄换电站的车辆排队区域的视频流；
从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
对多个视角下的环境图像进行去重合并后用作模型训练样本；
获取各所述模型训练样本相应的训练标签；
根据所述模型训练样本和相应的训练标签，训练所述排队车辆识别模型。
在本方案中，通过多个摄像头采集不同视角下的视频流以从中截取出环境图像，并去重合并后用作模型训练样本，以使环境图像能够尽量全面覆盖车辆排队区域，从而提升模型识别的准确性，而且提升了训练模型的稳定性。
可选地，所述模型训练模块还被配置为：
对多个视角下的环境图像进行去重合并得到中间图像；
在所述中间图像中划分出车辆排队区域；
将所述中间图像中除所述车辆排队区域之外的区域模糊处理；
将模糊处理后的所述中间图像用作模型训练样本。
在本方案中，在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
可选地，所述模型训练模块还被配置为：
将所述模型训练样本压缩和/或编码后输入所述排队车辆识别模型；
获取所述排队车辆识别模型输出的预测识别结果；
根据所述预测识别结果和所述模型训练样本相应的训练标签之间的差异，优化所述排队车辆识别模型的模型参数。
在本方案中，对模型训练样本进行相应的图像处理，以使得便于模型训练，而且减小传输的数据量，从而提升模型训练的稳定性，有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
可选地，所述模型训练模块还被配置为：
在所述排队车辆识别模型的准确率低于预设阈值时，更新所述排队车辆识别模型；
或，
在所述排队车辆识别模型的使用时长高于预设时长时，更新所述排队车辆识别模型。
在本方案中，当符合更新条件时，及时进行模型更新，实时跟进实际需求，从而进一步提升模型识别的准确性和效率。
可选地，所述模型训练模块配置在专用服务器上，排队车辆识别模型在训练完成后，专用服务器被配置为将排队车辆识别模型下发至各换电站和/或换电云端，以使换电站和/或换电云端采用排队车辆识别模型识别环境图像输出排队数量。
在本方案中，将训练好的模型下发至各换电站和/或换电云端，以使得各换电站和/或换电云端执行模型识别，从而有序安排模型训练过程和使用过程，有效地节省了成本，提升了模型识别的稳定性。根据本申请的另一实施方式，提供一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行计算机程序时实现如上述的换电站排队车辆的数量识别方法。
根据本申请的一实施方式，提供一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行计算机程序时实现如上述的换电站排队车辆的数量识别方法。
根据本申请的一实施方式，提供一种计算机可读介质，其上存储有计算机指令，所述计算机指令在由处理器执行时实现如上述的换电站排队车辆的数量识别方法的步骤。
在符合本领域常识的基础上，所述各优选条件，可任意组合，即得本申请各较佳实施例。
本申请的积极进步效果在于：
本申请通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
附图说明
图1为根据本申请实施例1的换电站排队车辆的数量识别方法的流程示意图。
图2为根据本申请实施例2的换电站排队车辆的数量识别方法的流程示意图。
图3为根据本申请实施例3的换电站排队车辆的数量识别***的模块结构示意图。
图4为根据本申请实施例4的换电站排队车辆的数量识别***的模块结构示意图。
图5为根据本申请实施例5的实现换电站排队车辆的数量识别方法的电子设备的结构示意图。
具体实施方式
下面通过实施例的方式进一步说明本申请，但并不因此将本申请限制在所述的实施例范围之中。
实施例1
为了克服目前存在的上述缺陷，本实施例提供一种换电站排队车辆的数量识别方法，包括：获取换电站的车辆排队区域的环境图像；识别环境图像中包括的排队车辆；根据识别出的排队车辆确定换电站的排队数量。
在本实施例中，该方法可应用于换电站或云端，但并不具体限定其应用场景，可根据实际需求进行相应的选择及调整。
在本实施例中，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
具体地，作为一实施方式，如图1所示，本实施例提供的换电站排队车辆的数量识别方法，主要包括以下步骤：
步骤101、获取换电站的车辆排队区域的环境图像。
其中，车辆排队区域可以是换电站中既定的用于车辆排队的区域，也可以是车辆自然排队所形成的区域。
在本步骤中，当该方法应用于换电站时，换电站本地服务器调用换电站摄像头以实时采集换电站的车辆排队区域的视频流，并且将视频流转换为图像以从中截取出换电站的车辆排队区域的环境图像。
将该方法应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力。
作为一优选实施例，在本步骤中，调用多个不同位置上的换电站摄像头以实时采集从多个视角拍摄换电站的车辆排队区域的视频流，并且将视频流分别转换为图像以从中截取出多个视角下包括车辆排队区域的环境图像。
通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
在一种可实施的方式中，可以通过一个摄像头实时变化视角以采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
在本步骤中，当该方法应用于云端时，换电站将采集到的视频流上报至云端服务器，也可将截取到的换电站的车辆排队区域的环境图像上报至云端服务器，云端服务器接收视频流以获取环境图像或直接接收环境图像。
将该方法应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
步骤102、识别环境图像中包括的排队车辆。
在本步骤中，对获取到的环境图像进行图像处理，并且通过图像识别技术从图像处理后的环境图像中识别出排队车辆。本实施例并不具体限定图像识别方式，只要能够实现相应的功能，可根据实际需求进行相应的选择。
具体地，图像处理包括压缩处理、编码处理及加密处理中的至少一种，例如，对环境图像进行压缩处理，并且将压缩后的图像转换成Base64(网络上最常见的用于传输8Bit字节码的编码方式之一)编码，再对Base64编码进行压缩和加密处理。但本实施例并不具体限定图像处理方式，可根据实际需求进行相应的选择。
通过对获取到的环境图像进行相应的图像处理，以使得减小传输的数据量，从而有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
在本步骤中，当该方法应用于云端时，可将图像处理后的环境图像上报至云端服务器，云端服务器可选择调用第三方图像识别服务等来进行车辆识别，从而有效地节省传输网络带宽，降低成本。
在本步骤中，当采用多个摄像头采集方式时，在对获取到的环境图像进行去重后，进行排队车辆识别。
具体地，当多个视角下的环境图像之间存在重复区域时，对环境图像进行重复区域裁剪后合并，并且对合并后的环境图像进行压缩和/或编码，再对压缩和/或编码后得到的目标图像进行排队车辆识别。
对存在重复区域的环境图像进行重复区域裁剪后合并，保留有效且没有重复的区域，尽可能减少需要进行识别的图像内容，提高排队车辆的识别效率。
作为一优选实施例，在本步骤中，在环境图像中划分出车辆排队区域，将环境图像中除车辆排队区域之外的区域模糊处理，并且对模糊处理后的环境图像进行排队车辆识别。
在一种可实施的方式中，车辆排队区域是换电站既定的用于车辆排队的区域。在本实施例中，可基于模板图像从环境图像中划分出车辆排队区域，也可识别环境图像中的排队路径以划分出车辆排队区域。
在一种可实施的方式中，车辆排队区域是车辆自然排队所形成的区域。在本实施例中，可基于多个车辆首位相连的位置关系识别出车辆排队区域。
上述可实施的方式中，将环境图像中的非排队区域模糊处理，从而有效地排除其他非换电车辆，进一步提升了识别排队数量的精度。
在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
步骤103、根据识别出的排队车辆确定换电站的排队数量。
在本步骤中，从环境图像中识别出排队车辆后，对车辆数量进行统计以确定换电站的排队数量，其中，可通过预设方式进一步确定识别出的车辆是否为排队车辆，例如，可排除换电后离开的车辆或目前正准确换电的车辆等。
步骤104、对确定出的排队数量进行验证。
在本步骤中，获取根据入网车辆的定位信息得到的排队数量，采用根据入网车辆的定位信息得到的排队数量校验根据识别出的排队车辆确定的排队数量。
具体地，根据入网车辆的定位信息，确定当前位置在车辆排队区域内的车辆数量并作为排队数量，判断从该排队数量是否与步骤103中输出的排队数量一致，若是，验证成功，若否，提示验证失败，并且选择使用哪一个数据作为排队数量。
通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
作为另一实施例，还可通过人工统计方式获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性，还可进一步选择使用哪一个数据。
步骤105、根据验证后的排队数量推荐换电站。
在本步骤中，将确定后的排队数量上传至云端，以使云端在接收到换电站查找请求时，根据各换电站上报的排队数量推荐换电站。
通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
在本实施例中，还可根据识别出的排队数量确定出换电等待时间，并且将排队数量和/或换电等待时间输出至换电站的显示终端或用户终端。
排队用户可通过换电等待时间和排队数量能够选择合适的换电站进行换电，从而提升了用户体验度。
在一种可实施的方式中，该方法应用于换电站时，换电站可按照时间周期将识别的换电站的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。换电站可在识别换电站的排队数量发生变化时，将新的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。
本实施例提供的换电站排队车辆的数量识别方法，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，而且进一步验证排队数量的准确性，同时还可以基于排队数量向用户推荐合适的换电站，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
实施例2
为了克服目前存在的上述缺陷，本实施例提供一种换电站排队车辆的数量识别方法，包括：获取换电站的车辆排队区域的环境图像；识别环境图像中包括的排队车辆；根据识别出的排队车辆确定换电站的排队数量。
在本实施例中，该方法可应用于换电站或云端，但并不具体限定其应用场景，可根据实际需求进行相应的选择及调整。
在本实施例中，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
具体地，作为一实施方式，如图2所示，本实施例提供的换电站排队车辆的数量识别方法，主要包括以下步骤：
步骤101、获取换电站的车辆排队区域的环境图像。
其中，车辆排队区域可以是换电站中既定的用于车辆排队的区域，也可以是车辆自然排队所形成的区域。
在本步骤中，当该方法应用于换电站时，换电站本地服务器调用换电站摄像头以实时采集换电站的车辆排队区域的视频流，并且将视频流转换为图像以从中截取出换电站的车辆排队区域的环境图像。
将该方法应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力。
作为一优选实施例，在本步骤中，调用多个不同位置上的换电站摄像头以实时采集从多个视角拍摄换电站的车辆排队区域的视频流，并且将视频流分别转换为图像以从中截取出多个视角下包括车辆排队区域的环境图像。
通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
在一种可实施的方式中，可以通过一个摄像头实时变化视角以采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
在本步骤中，当该方法应用于云端时，换电站将采集到的视频流上报至云端服务器，也可将截取到的换电站的车辆排队区域的环境图像上报至云端服务器，云端服务器接收视频流以获取环境图像或直接接收环境图像。
将该方法应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
步骤102’、将环境图像输入至训练后的排队车辆识别模型以输出环境图像中的排队车辆。
在本步骤中，将环境图像输入排队车辆识别模型，通过排队车辆识别模型识别环境图像中的排队车辆，排队车辆识别模型根据换电站的车辆排队区域的环境图像和相应的训练标签训练得到。
以下具体说明排队车辆识别模型训练过程。
首先，获取初始化的排队车辆识别模型，该排队车辆识别模型可以为用于图像识别的神经网络模型，只要能够实现车辆识别功能，本实施例并不具体限定模型类型，可根据实际需求进行相应的选择及调整。
其次，收集基于多个视角拍摄的视频流截取到的多个视角下包括车辆排队区域的环境图像，对多个视角下的环境图像进行去重合并后用作模型训练样本，并且获取各模型训练样本相应的训练标签。
作为一优选实施例，对获取到的环境图像即模型训练样本进一步进行图像处理。
具体地，图像处理包括压缩处理、编码处理及加密处理中的至少一种，例如，对环境图像进行压缩处理，并且将压缩后的图像转换成Base64(网络上最常见的用于传输8Bit字节码的编码方式之一)编码，再对Base64编码进行压缩和加密处理。但本实施例并不具体限定图像处理方式，可根据实际需求进行相应的选择。
对模型训练样本进行相应的图像处理，以使得便于模型训练，而且减小传输的数据量，从而提升模型训练的稳定性，有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
作为另一优选实施例，对多个视角下的环境图像进行去重合并得到中间图像，在中 间图像中划分出车辆排队区域，将中间图像中除车辆排队区域之外的区域模糊处理，将模糊处理后的中间图像用作模型训练样本。
具体地，当多个视角下的环境图像之间存在重复区域时，对环境图像进行重复区域裁剪后合并。对存在重复区域的环境图像进行复区域裁剪后合并，以使得提升模型训练的全面性。
在一种可实施的方式中，车辆排队区域是换电站既定的用于车辆排队的区域。在本实施例中，可基于模板图像从环境图像中划分出车辆排队区域，也可识别环境图像中的排队路径以划分出车辆排队区域。
在一种可实施的方式中，车辆排队区域是车辆自然排队所形成的区域。在本实施例中，可基于多个车辆首位相连的位置关系识别出车辆排队区域。
上述可实施的方式中，将模型训练样本中的非排队区域模糊处理，从而有效地排除其他非换电车辆，进一步提升了模型识别的精度。
最后，根据模型训练样本和相应的训练标签，训练排队车辆识别模型。
具体地，将模型训练样本输入排队车辆识别模型，获取排队车辆识别模型输出的预测识别结果，根据预测识别结果和模型训练样本相应的训练标签之间的差异，优化排队车辆识别模型的模型参数。
作为一优选实施例，在排队车辆识别模型的准确率低于预设阈值时，更新排队车辆识别模型。其中，还可获取根据入网车辆的定位信息得到的排队数量或者人工识别的排队数量，采用根据入网车辆的定位信息得到的排队数量或者人工识别的排队数量校验根据识别出的排队车辆确定的排队数量，得到排队车辆识别模型的准确率。本实施例并不具体限定预设阈值，可根据实际需求进行相应的设定。
作为另一优选实施例，在排队车辆识别模型的使用时长高于预设时长时，更新排队车辆识别模型，本实施例并不具体限定预设时长，可根据实际需求进行相应的设定。
当符合更新条件时，及时进行模型更新，实时跟进实际需求，从而进一步提升模型识别的准确性和效率。
作为一优选实施例，训练模型的步骤通过专用服务器执行，排队车辆识别模型在训练完成后，由专用服务器将排队车辆识别模型下发至各换电站和/或换电云端，以使换电站和/或换电云端采用排队车辆识别模型识别环境图像输出排队数量。
将训练好的模型下发至各换电站和/或换电云端，以使得各换电站和/或换电云端执行模型识别，从而有序安排模型训练过程和使用过程，有效地节省了成本，提升了模型识别的稳定性。
当车辆识别模型部署在换电站时，通过部署在换电站本地服务器上的模型来快速高效的识别出排队车辆，节省了运营成本，而且在互联网产生异常的情况下，也可通过本地局域网来实现车辆识别，从而保证了识别稳定性和可靠性。
步骤103’、通过排队车辆识别模型输出的识别结果确定换电站的排队数量。
在本步骤中，通过模型识别后的识别结果确定出换电站的排队数量，而且可通过预设方式进一步确定识别出的车辆是否为排队车辆，例如，可排除换电后离开的车辆或目前正准确换电的车辆等。
步骤104’、对确定出的排队数量进行验证。
在本步骤中，获取根据入网车辆的定位信息得到的排队数量，采用根据入网车辆的定位信息得到的排队数量校验根据识别出的排队车辆确定的排队数量。
具体地，根据入网车辆的定位信息，确定当前位置在车辆排队区域内的车辆数量并作为排队数量，判断从该排队数量是否与步骤103中输出的排队数量一致，若是，验证成功，若否，提示验证失败，并且选择使用哪一个数据作为排队数量。
通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
作为另一实施例，还可通过人工统计方式获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性，还可进一步选择使用哪一个数据。
步骤105’、根据验证后的排队数量推荐换电站。
在本步骤中，将确定后的排队数量上传至云端，以使云端在接收到换电站查找请求时，根据各换电站上报的排队数量推荐换电站。
通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
在本实施例中，还可根据识别出的排队数量确定出换电等待时间，并且将排队数量和/或换电等待时间输出至换电站的显示终端或用户终端。
排队用户可通过换电等待时间和排队数量能够选择合适的换电站进行换电，从而提升了用户体验度。
在一种可实施的方式中，该方法应用于换电站时，换电站可按照时间周期将识别的换电站的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。换电站可在识别换电站的排队数量发生变化时，将新的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。
本实施例提供的换电站排队车辆的数量识别方法，通过自动获取换电站的车辆排队 区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，而且进一步验证排队数量的准确性，同时还可以基于排队数量向用户推荐合适的换电站，从而可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
实施例3
为了克服目前存在的上述缺陷，本实施例提供一种换电站排队车辆的数量识别***，包括：环境图像获取模块，被配置为获取换电站的车辆排队区域的环境图像；排队车辆识别模块，被配置为识别环境图像中包括的排队车辆；排队数量确定模块，被配置为根据识别出的排队车辆确定换电站的排队数量。
具体地，如图3所示，作为另一实施方式，本实施例提供的换电站排队车辆的数量识别***主要包括环境图像获取模块21、排队车辆识别模块22、排队数量确定模块23、排队数量验证模块24及换电站推荐模块25，该***利用如上述实施例1的换电站排队车辆的数量识别方法。
当该***应用于换电站时，环境图像获取模块21被配置为调用换电站摄像头以实时获取换电站的车辆排队区域的视频流，并且将视频流转换为图像以从中截取出换电站的车辆排队区域的环境图像。
将该***应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力。
作为一优选实施例，环境图像获取模块21还被配置为调用多个不同位置上的换电站摄像头以实时获取从多个视角拍摄换电站的车辆排队区域的视频流，并且将视频流分别转换为图像以从中截取出多个视角下包括车辆排队区域的环境图像。
通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
当该***应用于云端时，环境图像获取模块21被配置为获取换电站上报的视频流，也可获取换电站上报的车辆排队区域的环境图像。
将该***应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
排队车辆识别模块22被配置为对获取到的环境图像进行图像处理，并且通过图像识别技术从图像处理后的环境图像中识别出排队车辆。本实施例并不具体限定图像识别方式，只要能够实现相应的功能，可根据实际需求进行相应的选择。
具体地，图像处理包括压缩处理、编码处理及加密处理中的至少一种，例如，对环境 图像进行压缩处理，并且将压缩后的图像转换成Base64编码，再对Base64编码进行压缩和加密处理。但本实施例并不具体限定图像处理方式，可根据实际需求进行相应的选择。
通过对获取到的环境图像进行相应的图像处理，以使得减小传输的数据量，从而有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
当该***应用于云端时，排队车辆识别模块22被配置为可选择调用第三方图像识别服务等来进行车辆识别，从而有效地节省传输网络带宽，降低成本。
当采用多个摄像头采集方式时，排队车辆识别模块22被配置为在对获取到的环境图像进行去重后，进行排队车辆识别。
具体地，当多个视角下的环境图像之间存在重复区域时，排队车辆识别模块22被配置为对环境图像进行重复区域裁剪后合并，并且对合并后的环境图像进行压缩和/或编码，再对压缩和/或编码后得到的目标图像进行排队车辆识别。
对存在重复区域的环境图像进行重复区域裁剪后合并，保留有效且没有重复的区域，尽可能减少需要进行识别的图像内容，提高排队车辆的识别效率。
作为一优选实施例，排队车辆识别模块22被配置为在环境图像中划分出车辆排队区域，将环境图像中除车辆排队区域之外的区域模糊处理，并且对模糊处理后的环境图像进行排队车辆识别。
在一种可实施的方式中，车辆排队区域是换电站既定的用于车辆排队的区域。在本实施例中，可基于模板图像从环境图像中划分出车辆排队区域，也可识别环境图像中的排队路径以划分出车辆排队区域。
在一种可实施的方式中，车辆排队区域是车辆自然排队所形成的区域。在本实施例中，可基于多个车辆首位相连的位置关系识别出车辆排队区域。
上述可实施的方式中，将环境图像中的非排队区域模糊处理，从而有效地排除其他非换电车辆，进一步提升了识别排队数量的精度。
在环境图像中划分出车辆排队区域，将环境图像中的非排队区域模糊处理，一方面可以有效地排除无效区域的其他非换电车辆的对换电车辆识别的干扰，提高排队车辆的识别准确率，另一方面还可以避免对无效区域进行识别的工作量，提高排队车辆的识别效率。
排队数量确定模块23被配置为从环境图像中识别出排队车辆后，对车辆数量进行统计以确定换电站的排队数量，其中，可通过预设方式进一步确定识别出的车辆是否为排队车辆，例如，可排除换电后离开的车辆或目前正准确换电的车辆等。
排队数量验证模块24被配置为获取根据入网车辆的定位信息得到的排队数量，采用根据入网车辆的定位信息得到的排队数量校验根据识别出的排队车辆确定的排队数量。
具体地，排队数量验证模块24被配置为根据入网车辆的定位信息，确定当前位置在车辆排队区域内的车辆数量并作为排队数量，判断从该排队数量是否与排队数量确定模块23输出的排队数量一致，若是，验证成功，若否，提示验证失败，并且选择使用哪一个数据作为排队数量。
通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
作为另一实施例，还可通过人工统计方式获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性，还可进一步选择使用哪一个数据。
换电站推荐模块25被配置为接收确定后的排队数量，在接收到换电站查找请求时，根据各换电站上报的排队数量推荐换电站。
通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
在本实施例中，该***还可根据识别出的排队数量确定出换电等待时间，并且将排队数量和/或换电等待时间输出至换电站的显示终端或用户终端。
排队用户可通过换电等待时间和排队数量能够选择合适的换电站进行换电，从而提升了用户体验度。
本实施例提供的换电站排队车辆的数量识别***，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，而且进一步验证排队数量的准确性，同时还可以基于排队数量向用户推荐合适的换电站，可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
实施例4
为了克服目前存在的上述缺陷，本实施例提供一种换电站排队车辆的数量识别***，包括：环境图像获取模块，被配置为获取换电站的车辆排队区域的环境图像；排队车辆识别模块，被配置为识别环境图像中包括的排队车辆；排队数量确定模块，被配置为根据识别出的排队车辆确定换电站的排队数量。
具体地，如图4所示，作为另一实施实施方式，本实施例提供的换电站排队车辆的数量识别***主要包括环境图像获取模块21、排队车辆识别模块22、排队数量确定模块23、排队数量验证模块24、换电站推荐模块25及模型训练模块26，该***利用如上述 实施例2的换电站排队车辆的数量识别方法。
当该***应用于换电站时，环境图像获取模块21被配置为调用换电站摄像头以实时获取换电站的车辆排队区域的视频流，并且将视频流转换为图像以从中截取出换电站的车辆排队区域的环境图像。
将该***应用于换电站时，可通过换电站本地服务器进行图像识别，可以降低带宽压力和云端计算压力。
作为一优选实施例，环境图像获取模块21还被配置为调用多个不同位置上的换电站摄像头以实时获取从多个视角拍摄换电站的车辆排队区域的视频流，并且将视频流分别转换为图像以从中截取出多个视角下包括车辆排队区域的环境图像。
通过多个摄像头采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
在一种可实施的方式中，可以通过一个摄像头实时变化视角以采集不同视角下的视频流，以保证环境图像能够尽量全面覆盖车辆排队区域，从而进一步提升识别排队数量的准确性。
当该***应用于云端时，环境图像获取模块21被配置为获取换电站上报的视频流，也可获取换电站上报的车辆排队区域的环境图像。
将该***应用于云端时，可通过云端服务器进行图像识别，以方便数据的统一管理调度，能够有效地避免数据篡改等情况。
排队车辆识别模块22被配置为将环境图像输入排队车辆识别模型，通过排队车辆识别模型识别环境图像中的排队车辆，排队车辆识别模型通过调用模型训练模块26来根据换电站的车辆排队区域的环境图像和相应的训练标签训练得到。
以下具体说明调用模型训练模块26训练排队车辆识别模型训练过程。
模型训练模块26被配置为获取初始化的排队车辆识别模型，该排队车辆识别模型可以为用于图像识别的神经网络模型，只要能够实现车辆识别功能，本实施例并不具体限定模型类型，可根据实际需求进行相应的选择及调整。
模型训练模块26还被配置为收集基于多个视角拍摄的视频流截取到的多个视角下包括车辆排队区域的环境图像，对多个视角下的环境图像进行去重合并后用作模型训练样本，并且获取各模型训练样本相应的训练标签。
作为一优选实施例，模型训练模块26还被配置为对获取到的环境图像即模型训练样本进一步进行图像处理。
具体地，图像处理包括压缩处理、编码处理及加密处理中的至少一种，例如，对环境 图像进行压缩处理，并且将压缩后的图像转换成Base64(网络上最常见的用于传输8Bit字节码的编码方式之一)编码，再对Base64编码进行压缩和加密处理。但本实施例并不具体限定图像处理方式，可根据实际需求进行相应的选择。
对模型训练样本进行相应的图像处理，以使得便于模型训练，而且减小传输的数据量，从而提升模型训练的稳定性，有效地节省传输网络带宽，降低成本，而且还可以有效地保证数据安全性。
作为另一优选实施例，模型训练模块26还被配置为对多个视角下的环境图像进行去重合并得到中间图像，在中间图像中划分出车辆排队区域，将中间图像中除车辆排队区域之外的区域模糊处理，将模糊处理后的中间图像用作模型训练样本。其中，车辆排队区域可以是换电站中既定的用于车辆排队的区域，也可以是车辆自然排队所形成的区域。
具体地，当多个视角下的环境图像之间存在重复区域时，对环境图像进行重复区域裁剪后合并。对存在重复区域的环境图像进行复区域裁剪后合并，以使得提升模型训练的全面性。
在一种可实施的方式中，车辆排队区域是换电站既定的用于车辆排队的区域。在本实施例中，可基于模板图像从环境图像中划分出车辆排队区域，也可识别环境图像中的排队路径以划分出车辆排队区域。
在一种可实施的方式中，车辆排队区域是车辆自然排队所形成的区域。在本实施例中，可基于多个车辆首位相连的位置关系识别出车辆排队区域。
上述可实施的方式中，将模型训练样本中的非排队区域模糊处理，从而有效地排除其他非换电车辆，进一步提升了模型识别的精度。
模型训练模块26还被配置为根据模型训练样本和相应的训练标签，训练排队车辆识别模型。
具体地，将模型训练样本输入排队车辆识别模型，获取排队车辆识别模型输出的预测识别结果，根据预测识别结果和模型训练样本相应的训练标签之间的差异，优化排队车辆识别模型的模型参数。
作为一优选实施例，模型训练模块26还被配置为在排队车辆识别模型的准确率低于预设阈值时，更新排队车辆识别模型。其中，还可获取根据入网车辆的定位信息得到的排队数量或者人工识别的排队数量，采用根据入网车辆的定位信息得到的排队数量或者人工识别的排队数量校验根据识别出的排队车辆确定的排队数量，得到排队车辆识别模型的准确率。本实施例并不具体限定预设阈值，可根据实际需求进行相应的设定。
作为另一优选实施例，模型训练模块26还被配置为在排队车辆识别模型的使用时长 高于预设时长时，更新排队车辆识别模型，本实施例并不具体限定预设时长，可根据实际需求进行相应的设定。
当符合更新条件时，及时进行模型更新，实时跟进实际需求，从而进一步提升模型识别的准确性和效率。
作为一优选实施例，模型训练模块26可以配置在专用服务器上，排队车辆识别模型在训练完成后，由专用服务器将排队车辆识别模型下发至各换电站和/或换电云端，以使换电站和/或换电云端采用排队车辆识别模型识别环境图像输出排队数量。
将训练好的模型下发至各换电站和/或换电云端，以使得各换电站和/或换电云端执行模型识别，从而有序安排模型训练过程和使用过程，有效地节省了成本，提升了模型识别的稳定性。
当车辆识别模型部署在换电站时，通过部署在换电站本地服务器上的模型来快速高效的识别出排队车辆，节省了运营成本，而且在互联网产生异常的情况下，也可通过本地局域网来实现车辆识别，从而保证了识别稳定性和可靠性。
排队数量确定模块23被配置为通过车辆识别模型识别后的识别结果确定出换电站的排队数量，而且可通过预设方式进一步确定识别出的车辆是否为排队车辆，例如，可排除换电后离开的车辆或目前正准确换电的车辆等。
排队数量验证模块24被配置为获取根据入网车辆的定位信息得到的排队数量，采用根据入网车辆的定位信息得到的排队数量校验根据识别出的排队车辆确定的排队数量。
具体地，排队数量验证模块24被配置为根据入网车辆的定位信息，确定当前位置在车辆排队区域内的车辆数量并作为排队数量，判断从该排队数量是否与排队数量确定模块23输出的排队数量一致，若是，验证成功，若否，提示验证失败，并且选择使用哪一个数据作为排队数量。
通过入网车辆的定位信息获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性。
作为另一实施例，还可通过人工统计方式获取到的排队数量，对通过环境图像确定出的排队数量进行验证，从而进一步提升数据准确性，还可进一步选择使用哪一个数据。
换电站推荐模块25被配置为接收确定后的排队数量，在接收到换电站查找请求时，根据各换电站上报的排队数量推荐换电站。
通过环境图像确定出的排队数量可用于向用户推荐换电站，以使得排队用户或待换电用户可方便地选择排队数量较少的换电站，从而节省用户时间，提升换电效率。
在本实施例中，该***还可根据识别出的排队数量确定出换电等待时间，并且将排 队数量和/或换电等待时间输出至换电站的显示终端或用户终端。
排队用户可通过换电等待时间和排队数量能够选择合适的换电站进行换电，从而提升了用户体验度。
在一种可实施的方式中，该***应用于换电站时，换电站可按照时间周期将识别的换电站的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。换电站可在识别换电站的排队数量发生变化时，将新的排队数量上传至云端，供云端在有需要时根据换电站的排队数量向驾驶用户推荐换电站。
本实施例提供的换电站排队车辆的数量识别***，通过自动获取换电站的车辆排队区域的环境图像，以从该环境图像中识别出排队车辆来确定换电站的排队数量，而且进一步验证排队数量的准确性，同时还可以基于排队数量向用户推荐合适的换电站，从而可以有效地提高排队数量的识别效率，而且还能够有效地节省换电站的人力资源耗费，降低成本。
实施例5
图5为根据本实施例提供的一种电子设备的结构示意图。电子设备包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，处理器执行程序时实现如上实施例1或实施例2中的换电站排队车辆的数量识别方法。图5显示的电子设备30仅仅是一个示例，不应对本申请实施例的功能和使用范围带来任何限制。
如图5所示，电子设备30可以以通用计算设备的形式表现，例如其可以为服务器设备。电子设备30的组件可以包括但不限于：上述至少一个处理器31、上述至少一个存储器32、连接不同***组件(包括存储器32和处理器31)的总线33。
总线33包括数据总线、地址总线和控制总线。
存储器32可以包括易失性存储器，例如随机存取存储器(RAM)321和/或高速缓存存储器322，还可以进一步包括只读存储器(ROM)323。
存储器32还可以包括具有一组(至少一个)程序模块324的程序/实用工具325，这样的程序模块324包括但不限于：操作***、一个或者多个应用程序、其它程序模块以及程序数据，这些示例中的每一个或某种组合中可能包括网络环境的实现。
处理器31通过运行存储在存储器32中的计算机程序，从而执行各种功能应用以及数据处理，例如本申请如上实施例中的换电站排队车辆的数量识别方法。
电子设备30也可以与一个或多个外部设备34(例如键盘、指向设备等)通信。这种通信可以通过输入/输出(I/O)接口35进行。并且，模型生成的设备30还可以通过网络适配器36与一个或者多个网络(例如局域网(LAN)，广域网(WAN)和/或公共网络，例如因特 网)通信。如图5所示，网络适配器36通过总线33与模型生成的设备30的其它模块通信。应当明白，尽管图中未示出，可以结合模型生成的设备30使用其它硬件和/或软件模块，包括但不限于：微代码、设备驱动器、冗余处理器、外部磁盘驱动阵列、RAID(磁盘阵列)***、磁带驱动器以及数据备份存储***等。
应当注意，尽管在上文详细描述中提及了电子设备的若干单元/模块或子单元/模块，但是这种划分仅仅是示例性的并非强制性的。实际上，根据本申请的实施方式，上文描述的两个或更多单元/模块的特征和功能可以在一个单元/模块中具体化。反之，上文描述的一个单元/模块的特征和功能可以进一步划分为由多个单元/模块来具体化。
本实施例还提供了一种计算机可读存储介质，其上存储有计算机程序，程序被处理器执行时实现如上述实施例1或实施例2中的换电站排队车辆的数量识别方法中的步骤。
其中，可读存储介质可以采用的更具体可以包括但不限于：便携式盘、硬盘、随机存取存储器、只读存储器、可擦拭可编程只读存储器、光存储器件、磁存储器件或上述的任意合适的组合。
在可能的实施方式中，本申请还可以实现为一种程序产品的形式，其包括程序代码，当程序产品在终端设备上运行时，程序代码用于使终端设备执行实现如上实施例中的换电站排队车辆的数量识别方法中的步骤。
其中，可以以一种或多种程序设计语言的任意组合来编写用于执行本申请的程序代码，程序代码可以完全地在用户设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户设备上部分在远程设备上执行或完全在远程设备上执行。
虽然以上描述了本申请的具体实施方式，但是本领域的技术人员应当理解，这些仅是举例说明，在不背离本申请的原理和实质的前提下，可以对这些实施方式做出多种变更或修改。因此，本申请的保护范围由所附权利要求书限定。

Claims (16)

  1. 一种换电站排队车辆的数量识别方法，其特征在于，包括：
    获取换电站的车辆排队区域的环境图像；
    识别所述环境图像中包括的排队车辆；
    根据识别出的排队车辆确定所述换电站的排队数量。
  2. 根据权利要求1所述的方法，其特征在于，所述识别所述环境图像中包括的排队车辆，包括：
    在所述环境图像中划分出车辆排队区域；
    将所述环境图像中除所述车辆排队区域之外的区域模糊处理；
    对模糊处理后的所述环境图像进行排队车辆识别。
  3. 根据权利要求1所述的方法，其特征在于，所述获取换电站的车辆排队区域的环境图像，包括：
    获取从多个视角拍摄换电站的车辆排队区域的视频流；
    从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
    所述识别所述环境图像中包括的排队车辆，包括：
    在对所述环境图像进行去重后，进行排队车辆识别。
  4. 根据权利要求3所述的方法，其特征在于，所述在对所述环境图像进行去重后，进行排队车辆识别，包括：
    当多个视角下的环境图像之间存在重复区域时，对所述环境图像进行重复区域裁剪后合并；
    对合并后的环境图像进行压缩和/或编码；
    对压缩和/或编码后得到的目标图像进行排队车辆识别。
  5. 根据权利要求1所述的方法，其特征在于，当应用于换电站时，所述获取换电站的车辆排队区域的环境图像，包括：
    获取换电站摄像头采集的视频流；
    从所述视频流中截取出包括车辆排队区域的环境图像；
    当应用于云端时，所述获取换电站的车辆排队区域的环境图像，包括：
    接收换电站上报的所述换电站的车辆排队区域的环境图像。
  6. 根据权利要求5所述的方法，其特征在于，当应用于换电站时，所述方法还包括：
    将所述排队数量上传至所述云端，以使所述云端在接收到换电站查找请求时，根据 各所述换电站上报的排队数量推荐换电站。
  7. 根据权利要求1所述的方法，其特征在于，还包括：
    获取根据入网车辆的定位信息得到的排队数量；
    采用根据入网车辆的定位信息得到的排队数量校验所述根据识别出的排队车辆确定的排队数量。
  8. 根据权利要求1所述的方法，其特征在于，所述识别所述环境图像中包括的排队车辆，包括：
    将所述环境图像输入排队车辆识别模型，通过所述排队车辆识别模型识别所述环境图像中的排队车辆；所述排队车辆识别模型根据换电站的车辆排队区域的环境图像和相应的训练标签训练得到；
    所述根据识别出的排队车辆确定所述换电站的排队数量，包括：
    通过所述排队车辆识别模型输出的识别结果确定所述换电站的排队数量。
  9. 根据权利要求8所述的方法，其特征在于，所述排队车辆识别模型的训练步骤包括：
    获取初始化的排队车辆识别模型；
    获取从多个视角拍摄换电站的车辆排队区域的视频流；
    从所述视频流中截取出多个视角下包括车辆排队区域的环境图像；
    对多个视角下的环境图像进行去重合并后用作模型训练样本；
    获取各所述模型训练样本相应的训练标签；
    根据所述模型训练样本和相应的训练标签，训练所述排队车辆识别模型。
  10. 根据权利要求9所述的方法，其特征在于，所述对多个视角下的环境图像进行去重合并后用作模型训练样本，包括：
    对多个视角下的环境图像进行去重合并得到中间图像；
    在所述中间图像中划分出车辆排队区域；
    将所述中间图像中除所述车辆排队区域之外的区域模糊处理；
    将模糊处理后的所述中间图像用作模型训练样本。
  11. 根据权利要求9所述的方法，其特征在于，所述根据所述模型训练样本和相应的训练标签，训练所述排队车辆识别模型，包括：
    将所述模型训练样本压缩和/或编码后输入所述排队车辆识别模型；
    获取所述排队车辆识别模型输出的预测识别结果；
    根据所述预测识别结果和所述模型训练样本相应的训练标签之间的差异，优化所述 排队车辆识别模型的模型参数。
  12. 根据权利要求9所述的方法，其特征在于，所述排队车辆识别模型的更新步骤包括：
    在所述排队车辆识别模型的准确率低于预设阈值时，更新所述排队车辆识别模型；
    或，
    在所述排队车辆识别模型的使用时长高于预设时长时，更新所述排队车辆识别模型。
  13. 根据权利要求9所述的方法，其特征在于，所述训练步骤通过专用服务器执行；所述排队车辆识别模型在训练完成后，由专用服务器将所述排队车辆识别模型下发至各所述换电站和/或换电云端，以使所述换电站和/或换电云端采用所述排队车辆识别模型识别所述环境图像输出排队数量。
  14. 一种换电站排队车辆的数量识别***，其特征在于，包括：
    环境图像获取模块，被配置为获取换电站的车辆排队区域的环境图像；
    排队车辆识别模块，被配置为识别所述环境图像中包括的排队车辆；
    排队数量确定模块，被配置为根据识别出的排队车辆确定所述换电站的排队数量。
  15. 一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行计算机程序时实现如权利要求1～13中任意一项所述的换电站排队车辆的数量识别方法。
  16. 一种计算机可读介质，其上存储有计算机指令，其特征在于，所述计算机指令在由处理器执行时实现如权利要求1～13中任意一项所述的换电站排队车辆的数量识别方法。
PCT/CN2021/142657 2020-12-31 2021-12-29 换电站排队车辆的数量识别方法、***、设备及介质 WO2022143802A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202011618807.3 2020-12-31
CN202011618777.6A CN114694053A (zh) 2020-12-31 2020-12-31 换电站排队车辆的数量识别方法、***、设备及介质
CN202011618807.3A CN114694084A (zh) 2020-12-31 2020-12-31 换电站排队数量的识别方法、***、设备及介质
CN202011618777.6 2020-12-31

Publications (1)

Publication Number Publication Date
WO2022143802A1 true WO2022143802A1 (zh) 2022-07-07

Family

ID=82260251

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/142657 WO2022143802A1 (zh) 2020-12-31 2021-12-29 换电站排队车辆的数量识别方法、***、设备及介质

Country Status (1)

Country Link
WO (1) WO2022143802A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116542005A (zh) * 2023-07-06 2023-08-04 杭州宇谷科技股份有限公司 基于深度学习的换电柜网络布局方法、***、装置及介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105809956A (zh) * 2014-12-31 2016-07-27 大唐电信科技股份有限公司 获取车辆排队长度的方法和装置
CN108550258A (zh) * 2018-03-29 2018-09-18 东软集团股份有限公司 车辆排队长度检测方法、装置、存储介质和电子设备
US20200035101A1 (en) * 2018-07-27 2020-01-30 Walmart Apollo, Llc Systems and methods for allocating vehicle parking spaces
CN111189451A (zh) * 2019-11-26 2020-05-22 恒大智慧科技有限公司 社区充电区域路线引导方法、***、计算机设备及存储介质
CN111523722A (zh) * 2020-04-20 2020-08-11 武汉大学 一种基于深度强化学习的智能充电站优化选择***
CN111784092A (zh) * 2019-12-30 2020-10-16 蒋兴德 基于排队情景分析的电量判断***

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105809956A (zh) * 2014-12-31 2016-07-27 大唐电信科技股份有限公司 获取车辆排队长度的方法和装置
CN108550258A (zh) * 2018-03-29 2018-09-18 东软集团股份有限公司 车辆排队长度检测方法、装置、存储介质和电子设备
US20200035101A1 (en) * 2018-07-27 2020-01-30 Walmart Apollo, Llc Systems and methods for allocating vehicle parking spaces
CN111189451A (zh) * 2019-11-26 2020-05-22 恒大智慧科技有限公司 社区充电区域路线引导方法、***、计算机设备及存储介质
CN111784092A (zh) * 2019-12-30 2020-10-16 蒋兴德 基于排队情景分析的电量判断***
CN111523722A (zh) * 2020-04-20 2020-08-11 武汉大学 一种基于深度强化学习的智能充电站优化选择***

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116542005A (zh) * 2023-07-06 2023-08-04 杭州宇谷科技股份有限公司 基于深度学习的换电柜网络布局方法、***、装置及介质
CN116542005B (zh) * 2023-07-06 2023-10-10 杭州宇谷科技股份有限公司 基于深度学习的换电柜网络布局方法、***、装置及介质

Similar Documents

Publication Publication Date Title
CN110733038A (zh) 工业机器人远程监控及数据处理***
CN102222213B (zh) 一种基于开放式Web Service架构的分布式视觉计算方法
WO2022143802A1 (zh) 换电站排队车辆的数量识别方法、***、设备及介质
CN107146623B (zh) 基于人工智能的语音识别方法、装置和***
CN102546793B (zh) 一种基于云计算的移动终端自适应方法及其***
WO2022143879A1 (zh) 换电站的推荐方法、***、电子设备及存储介质
CN111510774A (zh) 一种结合边缘计算与深度学习的智能终端图像压缩算法
WO2022143872A1 (zh) 换电站的排队车辆识别方法、***、设备及介质
CN104268525A (zh) 一种基于人脸识别的车辆监控方法及***
CN109001530B (zh) 一种电能表识别分拣***
CN102523432A (zh) 一种基于3g的智能车载监控***
CN114694053A (zh) 换电站排队车辆的数量识别方法、***、设备及介质
US11265594B2 (en) Data processing method and device, storage medium, electronic device
CN114694085A (zh) 换电排队车辆的数量识别方法、***、设备及介质
EP4332826A1 (en) Image recognition method and electronic device
CN115018205B (zh) 一种基于物联网的智慧城市无人机管理方法与***
CN110855947A (zh) 一种图像抓拍处理方法及装置
WO2023045365A1 (zh) 视频质量评估方法、装置、电子设备及存储介质
CN110764931B (zh) Ota网站上传凭证的处理方法、***、设备和存储介质
CN114694084A (zh) 换电站排队数量的识别方法、***、设备及介质
CN110099116A (zh) 一种基于大数据的子网安全性评估方法
CN112667390B (zh) 一种结合5g技术和边缘计算的覆冰监测数据处理***及方法
CN112487850A (zh) 一种基于人工智能人脸识别的司法场所快速点名方法
CN114683909B (zh) 换电处理方法、***、设备及可读存储介质
CN116841772B (zh) 基于事件队列的跨线程信号处理方法和***

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21914534

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21914534

Country of ref document: EP

Kind code of ref document: A1