WO2022032861A1 - 一种交通灯识别的方法和装置 - Google Patents
一种交通灯识别的方法和装置 Download PDFInfo
- Publication number
- WO2022032861A1 WO2022032861A1 PCT/CN2020/121562 CN2020121562W WO2022032861A1 WO 2022032861 A1 WO2022032861 A1 WO 2022032861A1 CN 2020121562 W CN2020121562 W CN 2020121562W WO 2022032861 A1 WO2022032861 A1 WO 2022032861A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- camera
- traffic light
- image data
- cameras
- determined
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/809—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of classification results, e.g. where the classifiers operate on the same input data
- G06V10/811—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of classification results, e.g. where the classifiers operate on the same input data the classifiers operating on different input data, e.g. multi-modal recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/09—Arrangements for giving variable traffic instructions
- G08G1/095—Traffic lights
Definitions
- the present invention relates to the field of image technology, and in particular, to a method and device for identifying traffic lights.
- the picture based on the single front-view camera is usually used for traffic light recognition.
- the relationship between the focal length of the camera and the angle of view cannot meet the appropriate shooting conditions, it is easy to cause traffic accidents. For example, when the traffic light exceeds the angle of view of the camera, it will cause the visual perception recognition system to mistakenly think that there is no traffic light; or when the traffic light is small in the angle of view of the camera , which will cause the visual perception recognition system to fail to clearly identify the traffic lights.
- a method for traffic light recognition involving multiple cameras, the fields of view of any two cameras have an overlay relationship, and the method includes:
- the target camera According to the plurality of cameras to collect image data, determine the target camera
- Traffic light recognition is performed using the target camera.
- the determining of the target camera by collecting image data according to the multiple cameras includes:
- the second camera is determined to be a target camera.
- the first camera is determined to be a target camera.
- the judging whether the position of the traffic light matches the second camera includes:
- the method before determining the position of the traffic light in the first image data collected by the first camera, the method further includes:
- the determining the position of the traffic light in the first image data collected by the first camera is performed.
- the second camera is used as the first camera, and the process of determining whether a traffic light is detected in the first image data collected by the first camera is returned to;
- the plurality of cameras are sorted according to the size of the field of view, and the first camera and the second camera are cameras that are adjacent to each other.
- a device for traffic light recognition involving multiple cameras, the fields of view of any two cameras have an overlay relationship, and the device includes:
- an image data acquisition module configured to acquire the image data collected by the plurality of cameras
- a target camera determination module configured to collect image data according to the plurality of cameras to determine the target camera
- the traffic light recognition module is used for using the target camera to perform traffic light recognition.
- An electronic device comprising a processor, a memory, and a computer program stored on the memory and capable of running on the processor, the computer program implementing the above-mentioned traffic light recognition when executed by the processor method.
- a computer-readable storage medium storing a computer program on the computer-readable storage medium, when the computer program is executed by a processor, implements the above-mentioned method for identifying a traffic light.
- Image data is collected by acquiring multiple cameras, and then image data is collected according to multiple cameras to determine the target camera, and then use the target camera to conduct traffic.
- Light recognition realizes multi-camera fusion visual perception recognition for traffic lights, collects image data through multiple cameras to filter out the most suitable camera for traffic light recognition, avoids the omission of traffic lights caused by the limited viewing angle, and improves traffic The accuracy and reliability of light recognition can effectively reduce the occurrence of traffic accidents.
- FIG. 1 is a flowchart of steps of a method for identifying a traffic light provided by an embodiment of the present invention
- FIG. 2 is a flowchart of steps of another traffic light identification method provided by an embodiment of the present invention.
- FIG. 3 is a schematic diagram of an example of traffic light identification provided by an embodiment of the present invention.
- FIG. 4 is a flowchart of steps of another traffic light identification method provided by an embodiment of the present invention.
- 5a is a schematic diagram of a view relationship of multiple cameras according to an embodiment of the present invention.
- FIG. 5b is a schematic diagram of an example of a priority area provided by an embodiment of the present invention.
- 5c is a schematic diagram of an example of a priority area determination provided by an embodiment of the present invention.
- 5d is a schematic diagram of another example of priority area determination provided by an embodiment of the present invention.
- FIG. 6 is a schematic structural diagram of an apparatus for identifying a traffic light according to an embodiment of the present invention.
- FIG. 1 a flow chart of steps of a method for traffic light recognition provided by an embodiment of the present invention is shown.
- the method may involve multiple cameras, and the fields of view of any two cameras may have an overlay relationship, which may specifically include the following steps :
- Step 101 acquiring image data collected by the multiple cameras
- the image data may be scene image data for the current driving situation of the vehicle, and the scene image data may be collected by a vehicle-mounted camera to be identified by a user's traffic lights. For example, pictures may be taken by a plurality of preset vehicle-mounted cameras.
- the vehicle can be equipped with multiple on-board cameras, and multiple on-board cameras can take pictures respectively, and then can obtain image data collected by multiple cameras in the current driving situation to further identify the traffic lights.
- multi-eye cameras with different focal lengths and different viewing angles may be used, wherein the fields of view of any two cameras in the multi-eye camera may have an overlay relationship.
- the visual perception recognition method is used for traffic light recognition.
- cameras with different viewing angles in a multi-camera camera can cover a variety of situations such as traffic lights or far or near.
- a trinocular camera can be configured in an automatic driving system, and the fields of view of any two cameras can have an overlay relationship.
- the field of view of the camera 1 can be completely covered by the field of view of the camera 2, and the field of view of the camera 2 can be completely covered by the field of view of the camera 3.
- the same method can be used to configure a binocular camera and a quad camera, which is not limited in the present invention.
- Step 102 collecting image data according to the multiple cameras, and determining a target camera
- the target camera can be the most suitable camera selected for traffic light recognition.
- the traffic light visual perception recognition system can switch multiple cameras to select the most suitable camera as the target camera.
- the image data After acquiring the image data collected by the multiple cameras, the image data can be collected according to the multiple cameras, and the most suitable camera for traffic light recognition selected from the multiple cameras is used as the target camera.
- the traffic light visual perception recognition system can seamlessly switch the pictures taken by the multi-camera cameras, and then can select the most suitable camera according to the current driving scene pictures collected by multiple cameras, so as to further use the camera to identify traffic. lamp.
- Step 103 using the target camera to identify traffic lights.
- the target camera After the target camera is determined, the target camera can be determined as the current camera for traffic light recognition, and then the target camera can be used for traffic light recognition.
- the traffic light visual perception recognition system is based on the picture taken by the monocular camera, if the focal length of the monocular camera is long and the viewing distance is far, but the increase of the focal length will cause the viewing angle to narrow. It will cause the visual perception and recognition system to mistakenly think that no traffic light pictures have been taken, resulting in traffic accidents; while using a camera with a short focal length and a wide viewing angle, the camera will not be able to see far, causing the traffic light to be relatively small in the camera's viewing angle, making it difficult for the camera to see.
- the visual perception recognition system cannot clearly identify the traffic lights in the picture, and it is also easy to cause traffic accidents.
- the visual perception and recognition method of traffic lights using multi-eye cameras by integrating cameras with different focal lengths and different perspectives, can cover the traffic lights or take pictures in a variety of situations far or near, thus solving the problem of monocular camera perspective.
- the problem of missing or unclear traffic lights caused by limitations improves the reliability of traffic light recognition.
- Image data is collected by acquiring multiple cameras, and then image data is collected according to multiple cameras to determine the target camera, and then use the target camera to conduct traffic.
- Light recognition realizes multi-camera fusion visual perception recognition for traffic lights, collects image data through multiple cameras to filter out the most suitable camera for traffic light recognition, avoids the omission of traffic lights caused by the limited viewing angle, and improves traffic The accuracy and reliability of light recognition can effectively reduce the occurrence of traffic accidents.
- the method may involve multiple cameras, and the fields of view of any two cameras may have an overlay relationship, which may specifically include the following step:
- Step 201 acquiring image data collected by the multiple cameras
- the vehicle can be equipped with multiple on-board cameras, and multiple on-board cameras can take pictures respectively, and then can obtain image data collected by multiple cameras in the current driving situation to further identify the traffic lights.
- the plurality of cameras may be sorted according to the size of the field of view, and the plurality of cameras may be trinocular cameras, and the trinocular camera may include a fisheye camera, a main camera, and a narrow camera, wherein the sorting according to the size of the field of view may be a fisheye camera.
- the field of view can cover the main camera field of view, and the main camera field of view can cover the narrow camera field of view.
- image data can be collected from cameras with different viewing angles of the fisheye camera, the main camera, and the narrow camera.
- Step 202 judging whether a traffic light is detected in the first image data collected by the first camera
- the first camera may be a camera with a larger field of view among the two cameras under the current judgment condition
- the current judgment condition may be two cameras that are ranked adjacent to each other with respect to the coverage relationship of the field of view.
- the first image data may be image data collected by the first camera, for example, a picture taken by a fisheye camera for the current driving scene.
- the first camera After acquiring the image data collected by the multiple cameras, the first camera may be determined from the multiple cameras, and further detection may be performed on the first image data collected by the first camera to determine whether a traffic light is detected.
- the plurality of cameras may be trinocular cameras, and the trinocular cameras may include a fisheye camera, a main camera, and a narrow camera.
- the fisheye camera with the largest field of view may be the first one. The camera can then determine whether a traffic light is detected in the picture captured by the fisheye camera.
- traffic light detection may be performed on a picture captured by a trinocular camera through a traffic light detection network, so as to determine whether a traffic light is detected in the picture captured by the trinocular camera.
- Step 203 when it is determined that a traffic light is detected in the first image data collected by the first camera, determine the location of the traffic light in the first image data collected by the first camera;
- the regional location of the traffic light in the first image data can be determined as the location of the traffic light in the first image data, for example, a picture captured by the camera can be used When a traffic light is detected in the image, the geometric region position of the traffic light rendering area in the picture is further determined.
- Step 204 judging whether the position of the traffic light matches the second camera; wherein, the field of view of the first camera covers the field of view of the second camera;
- the first camera and the second camera may be cameras that are arranged adjacent to each other.
- the first camera and the second camera may be two cameras that are arranged according to the size of the field of view and whose field of view covers the relationship of adjacent cameras.
- the location of the traffic light After the location of the traffic light is determined, it can be determined whether the location of the traffic light in the first image data collected by the first camera matches that of the second camera through the preset geometrical position relationship between the cameras based on the image matching method.
- the field of view of the camera may cover the field of view of the second camera.
- the position of the main camera scene with a smaller viewing angle in the fisheye camera scene with a larger viewing angle can be determined based on the method of picture matching, and then Whether the traffic light is in the scene of the main camera can be detected according to the geometric position relationship determined between the cameras, so as to determine whether the position of the traffic light matches the second camera.
- Step 205 when it is determined that the position of the traffic light matches the second camera, determine that the second camera is a target camera;
- the second camera may be determined as the most suitable camera for traffic light recognition selected from the plurality of cameras as the target camera.
- Step 206 using the target camera to identify traffic lights.
- the target camera After the target camera is determined, the target camera can be determined as the current camera for traffic light recognition, and then the target camera can be used for traffic light recognition.
- the first camera is determined to be a target camera.
- the first camera may be determined as the most suitable camera for traffic light recognition selected from the plurality of cameras as the target camera.
- the second camera determines whether the second camera is the camera with the smallest field of view; when it is determined that the second camera is not the camera with the smallest field of view, the The second camera is used as the first camera, and returns to execute the judging whether a traffic light is detected in the first image data collected by the first camera; when it is determined that the second camera is the camera with the smallest field of view, the output is not detected Information about traffic lights.
- the second camera can be determined to determine whether the second camera is the camera with the smallest field of view among the current multi-camera cameras, and then the second camera is determined When it is not the camera with the smallest field of view, the second camera can be used as the first camera, and the step of judging whether a traffic light is detected in the first image data collected by the first camera is returned to further filter out the most traffic lights from the plurality of cameras.
- Target camera suitable for traffic light recognition.
- prompt information that no traffic lights are detected may be output to indicate that the traffic light visual perception method using multi-camera fusion is used for current driving. The scene does not detect traffic lights.
- a trinocular camera can be used to take pictures at the same time.
- the trinocular camera can include a fisheye camera, a main camera, and a narrow camera.
- the fisheye After obtaining the pictures taken by the trinocular camera, due to the sorting according to the size of the field of view, the fisheye The eye camera is the camera with the largest field of view, and then the traffic light detection network can be used to determine whether a traffic light is detected in the picture collected by the fisheye camera.
- the main camera can be used for traffic light recognition; if the traffic light does not exist in the scene of the main camera, the fisheye camera can be used for traffic light recognition.
- a picture matching algorithm can be used to determine the position of the scene of the main camera in the fisheye camera, and then based on the predetermined geometric position between the main camera and the fisheye camera relationship, to determine whether the traffic light exists in the scene of the main camera.
- the main camera adjacent to the fisheye camera's field of view is not the camera with the smallest field of view among the currently used trinocular cameras, it can be further determined that the main camera captures Whether a traffic light is detected in the picture of .
- the main camera and the narrow camera are cameras with adjacent fields of view, and the field of view of the main camera covers the field of view of the narrow camera, it can be judged whether the traffic light exists in the narrow camera.
- the narrow camera can be used for traffic light recognition; if the traffic light does not exist in the narrow camera scene, the main camera can be used for traffic light recognition.
- a picture matching algorithm can also be used to determine the position of the scene with a narrow camera in the main camera.
- the narrow camera adjacent to the main camera's field of view is the camera with the smallest field of view among the currently used trinocular cameras, it can output a prompt that no traffic light is detected. information.
- multiple cameras are involved, and the fields of view of any two cameras have an overlay relationship.
- a traffic light is detected in the first image data collected by the first camera.
- the position of the traffic light is determined in the first image data collected by the first camera, and then it is determined whether the position of the traffic light matches the second camera, and the first camera
- the field of view covers the field of view of the second camera, and when it is determined that the position of the traffic light matches the second camera, the second camera is determined as the target camera, and then the target camera is used for traffic light recognition, realizing multi-camera fusion vision for traffic lights Perceptual recognition, by judging whether traffic lights are detected in the image data collected by multiple cameras, and then determining the most suitable camera for traffic light recognition, avoiding the omission of traffic lights caused by the limited viewing angle, and improving the accuracy of traffic light recognition reliability and reliability, which can effectively reduce the occurrence of traffic accidents.
- the method may involve multiple cameras, and the fields of view of any two cameras may have an overlay relationship, which may specifically include the following step:
- Step 401 acquiring image data collected by the multiple cameras
- the vehicle can be equipped with multiple on-board cameras, and multiple on-board cameras can take pictures respectively, and then can obtain image data collected by multiple cameras in the current driving situation to further identify the traffic lights.
- Step 402 in the first image data collected by the first camera, determine the location of the traffic light
- the first camera After acquiring the image data collected by the multiple cameras, the first camera can be determined from the multiple cameras, and then the first image data collected by the first camera can be detected, and the detection can be performed in determining the first image data collected by the first camera.
- the area position of the traffic light in the first image data is further determined, as the traffic light position in the first image data, for example, when the traffic light is detected in the picture captured by the camera, the traffic light presentation area is further determined. The location of the geometric region in the picture.
- Step 403 obtaining the relative positional relationship between the first camera and the second camera
- the relative positional relationship between the first camera and the second camera can be obtained based on the image matching method, so as to further determine whether the location of the traffic light matches the second camera through the relative positional relationship.
- the trinocular camera may include a fisheye camera, a main camera, and a narrow camera, wherein the order according to the size of the field of view can be that the field of view of the fisheye camera can cover the main camera.
- Camera field of view, the main camera field of view can cover the narrow camera field of view.
- feature point extraction and matching algorithms such as SIFT algorithm (Scale-invariant feature transform), ORB algorithm (Oriented Fast and Rotated Brief) can be used to obtain the relative relationship between the cameras in the trinocular camera. Since the relative position of the multi-cameras configured for the same vehicle is fixed, the relative positional relationship of the field of view of each camera can remain unchanged for a long time, and then the relative positional relationship of the field of view between the cameras can be calculated offline and stored. In the process of identifying traffic lights, the relative positional relationship between the cameras can be directly called without multiple calculations.
- SIFT algorithm Scale-invariant feature transform
- ORB algorithm Oriented Fast and Rotated Brief
- Step 404 determine a priority image area in the first image data
- a priority image area for the second camera may be determined in the first image data according to the relative positional relationship to detect the priority image area to further determine whether the traffic light position matches the second camera.
- Fig. 5b by performing offline calculation on the relative positional relationship of the field of view between cameras, in the field of view of the first camera with a larger field of view, it is possible to determine the size of the camera's field of view that is adjacent to (and smaller than the camera)
- the geometric position of the second camera is used as the priority image area for the second camera (such as priority area B), and then the remaining areas in the field of view of the first camera can be used as the area outside the field of view of the second camera (such as priority area A, priority area B) area C).
- Step 405 detecting whether the priority image area includes the traffic light position
- the priority image area After the priority image area is determined, it may be detected whether the priority image area contains the traffic light position to determine whether the traffic light position matches the second camera.
- the priority image area may be the geometric position of the second camera in the field of view of the first camera, after the priority image area is determined, it may be detected whether the priority image area contains traffic lights in the first image data collected for the first camera Location.
- Step 406 when it is detected that the priority image area includes the traffic light position, determine that the traffic light position matches the second camera;
- the traffic light position matches the second camera.
- the traffic light position may be included in the first image data collected by the first camera in the priority image area for the second camera. , determine that the traffic light is in the scene of the second camera.
- the priority area A can be It is the area in the field of view of the fisheye camera for the area outside the field of view of the main camera, so that when it is determined that the traffic light exists in the priority area A, it can be determined that the traffic light exists in the scene of the fisheye camera; when it is determined that the traffic light does not exist in the priority area A , it can be determined whether the traffic light exists in the priority area B, and the priority area B can be the area in the fisheye camera field of view for the main camera field of view, and then when it is determined that the traffic light exists in the priority area B, it can be determined that the traffic light exists in the main camera. in the scene.
- the priority area A can be The main camera's field of view is aimed at the area outside the narrow camera's field of view, so that when it is determined that the traffic light exists in the priority area A, it can be determined that the traffic light exists in the scene of the main camera; when it is determined that the traffic light does not exist in the priority area A, you can Determine whether the traffic light exists in the priority area B, the priority area B can be the area in the main camera field of view for the narrow camera field of view, and then when it is determined that the traffic light exists in the priority area B, the traffic light exists in the narrow camera scene. .
- Step 407 when it is detected that the priority image area does not contain the location of the traffic light, determine that the location of the traffic light does not match the second camera;
- the priority image area does not contain the traffic light position
- the priority image area for the second camera may not contain the first image data collected by the first camera.
- the traffic light is located, it is determined that the traffic light does not exist in the scene of the second camera.
- Step 408 when it is determined that the position of the traffic light matches the second camera, determine that the second camera is a target camera;
- the second camera may be determined as the most suitable camera for traffic light recognition selected from the plurality of cameras as the target camera.
- Step 409 using the target camera to identify traffic lights.
- the target camera After the target camera is determined, the target camera can be determined as the current camera for traffic light recognition, and then the target camera can be used for traffic light recognition.
- the traffic light visual perception recognition system can seamlessly switch the pictures taken by the multi-camera camera, and then can select the most suitable camera for traffic light recognition, thus avoiding the problem of the monocular camera's perspective itself.
- the problem of missing or unclear traffic lights caused by defects improves the reliability of traffic light recognition.
- the algorithm compared to an algorithm using machine learning, whenever the data changes, the algorithm needs to be retrained, the data requirements are high, and the algorithm results are unstable.
- the image matching algorithm can be used to obtain the geometric position relationship between the cameras, and then the cameras can be switched accurately to select the most suitable camera for traffic light recognition. interpretability.
- Image data is collected by acquiring multiple cameras.
- the location of the traffic light is determined, and then the The relative positional relationship between the first camera and the second camera, according to the relative positional relationship, determine the priority image area in the first image data, detect whether the priority image area contains the traffic light position, when it is detected that the priority image area contains the traffic light position, It is determined that the traffic light position matches the second camera, and when it is detected that the priority image area does not contain the traffic light position, it is determined that the traffic light position does not match the second camera, and when it is determined that the traffic light position matches the second camera,
- the second camera is determined as the target camera, and the target camera is used for traffic light recognition, which realizes multi-camera fusion visual perception recognition for traffic lights, collects image data through multiple cameras, and detects whether the priority image area contains the traffic light position, so as to filter
- the most suitable camera is used for traffic light recognition,
- FIG. 6 a schematic structural diagram of a traffic light recognition device provided by an embodiment of the present invention is shown.
- the device may involve multiple cameras, and the fields of view of any two cameras may have an overlay relationship, and may specifically include the following modules:
- An image data acquisition module 601, configured to acquire image data collected by the plurality of cameras
- a target camera determination module 602 configured to collect image data according to the multiple cameras to determine a target camera
- the traffic light recognition module 603 is configured to use the target camera to perform traffic light recognition.
- the target camera determination module 602 includes:
- a traffic light position determination submodule configured to determine the traffic light position in the first image data collected by the first camera
- a judging sub-module for judging whether the position of the traffic light matches the second camera; wherein, the field of view of the first camera covers the field of view of the second camera;
- the target camera determination sub-module is configured to determine that the second camera is a target camera when it is determined that the position of the traffic light matches the second camera.
- the device further includes:
- a determination module configured to determine that the first camera is a target camera when it is determined that the position of the traffic light does not match the second camera.
- the judging submodule includes:
- a relative positional relationship acquiring unit configured to acquire the relative positional relationship between the first camera and the second camera
- a priority image area determination unit configured to determine a priority image area in the first image data according to the relative positional relationship
- a detection unit configured to detect whether the priority image area includes the traffic light position
- a first determining unit configured to determine that the traffic light position matches the second camera when it is detected that the priority image area includes the traffic light position
- a second determining unit configured to determine that the traffic light position does not match the second camera when it is detected that the priority image area does not contain the traffic light position.
- the device further includes:
- a first judging module for judging whether a traffic light is detected in the first image data collected by the first camera
- the second judging module is configured to call the traffic light position determination sub-module when it is judged that a traffic light is detected in the first image data collected by the first camera.
- the device further includes:
- a second camera determination module configured to determine whether the second camera is a camera with the smallest field of view when it is determined that no traffic lights are detected in the first image data collected by the first camera;
- the second camera determination module is configured to use the second camera as the first camera when determining that the second camera is not the camera with the smallest field of view, and return to call the first determination module;
- the prompt information output module is configured to output prompt information that no traffic light is detected when it is determined that the second camera is the camera with the smallest field of view.
- the plurality of cameras are ordered according to the size of the field of view, and the first camera and the second camera are cameras that are adjacent in order.
- Image data is collected by acquiring multiple cameras, and then image data is collected according to multiple cameras to determine the target camera, and then use the target camera to conduct traffic.
- Light recognition realizes multi-camera fusion visual perception recognition for traffic lights, collects image data through multiple cameras to filter out the most suitable camera for traffic light recognition, avoids the omission of traffic lights caused by the limited viewing angle, and improves traffic The accuracy and reliability of light recognition can effectively reduce the occurrence of traffic accidents.
- An embodiment of the present invention also provides an electronic device, which may include a processor, a memory, and a computer program stored in the memory and capable of running on the processor.
- a computer program stored in the memory and capable of running on the processor.
- An embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the above method for identifying a traffic light is implemented.
- embodiments of the present invention may be provided as a method, an apparatus, or a computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media having computer-usable program code embodied therein, including but not limited to disk storage, CD-ROM, optical storage, and the like.
- Embodiments of the present invention are described with reference to flowcharts and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present invention. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing terminal equipment to produce a machine that causes the instructions to be executed by the processor of the computer or other programmable data processing terminal equipment Means are created for implementing the functions specified in the flow or flows of the flowcharts and/or the blocks or blocks of the block diagrams.
- These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal equipment to operate in a particular manner, such that the instructions stored in the computer readable memory result in an article of manufacture comprising instruction means, the The instruction means implement the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
- Studio Devices (AREA)
Abstract
一种交通灯识别的方法和装置,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,方法包括:获取多个摄像头采集图像数据(101,201,401);根据多个摄像头采集的图像数据,确定目标摄像头(102);采用目标摄像头进行交通灯识别(103,206,409)。有益效果:通过多个摄像头采集图像数据,以筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
Description
本发明涉及图像技术领域,特别是涉及一种交通灯识别的方法和装置。
在交通灯视觉感知识别***中,通常采用的是基于单目前视摄像头拍摄的图片,以进行交通灯识别。
由于摄像头焦距与视角的关系无法达到合适拍摄条件,容易引发交通事故,如当交通灯超出摄像头视角时,将导致视觉感知识别***错误地认为当前没有交通灯;或在摄像头视角中交通灯较小时,将导致视觉感知识别***无法清楚识别出交通灯。
发明内容
鉴于上述问题,提出了以便提供克服上述问题或者至少部分地解决上述问题的一种交通灯识别的方法和装置,包括:
一种交通灯识别的方法,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,所述方法包括:
获取所述多个摄像头采集图像数据;
根据所述多个摄像头采集图像数据,确定目标摄像头;
采用所述目标摄像头进行交通灯识别。
可选地,所述根据所述多个摄像头采集图像数据,确定目标摄像头,包括:
在第一摄像头采集的第一图像数据中,确定交通灯位置;
判断所述交通灯位置是否与第二摄像头相匹配;其中,所述第一摄像头的视野覆盖所述第二摄像头的视野;
在判定所述交通灯位置与所述第二摄像头相匹配时,确定所述第二摄像头为目标摄像头。
可选地,还包括:
在判定所述交通灯位置与所述第二摄像头不相匹配时,确定所述第一摄像头为目标摄像头。
可选地,所述判断所述交通灯位置是否与第二摄像头相匹配,包括:
获取所述第一摄像头和所述第二摄像头的相对位置关系;
根据所述相对位置关系,在所述第一图像数据中确定优先图像区域;
检测所述优先图像区域是否包含所述交通灯位置;
在检测到所述优先图像区域包含所述交通灯位置时,判定所述交通灯位置与第二摄像头相匹配;
在检测到所述优先图像区域未包含所述交通灯位置时,判定所述交通灯位置与第二摄像头不相匹配。
可选地,在所述在第一摄像头采集的第一图像数据中,确定交通灯位置之前,还包括:
判断第一摄像头采集的第一图像数据中是否检测到交通灯;
在判定所述第一摄像头采集的第一图像数据中检测到交通灯时,执行所述在第一摄像头采集的第一图像数据中,确定交通灯位置。
可选地,还包括:
在判定所述第一摄像头采集的第一图像数据中未检测到交通灯时,判断第二摄像头是否为视野最小的摄像头;
在判定所述第二摄像头不为视野最小的摄像头时,将所述第二摄像头作为第一摄像头,并返回执行所述判断所述第一摄像头采集的第一图像数据中是否检测到交通灯;
在判定所述第二摄像头为视野最小的摄像头时,输出未检测到交通灯的提示信息。
可选地,所述多个摄像头按照视野大小排序,所述第一摄像头和所述第二摄像头为排序相邻的摄像头。
一种交通灯识别的装置,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,所述装置包括:
图像数据获取模块,用于获取所述多个摄像头采集图像数据;
目标摄像头确定模块,用于根据所述多个摄像头采集图像数据,确定目标摄像头;
交通灯识别模块,用于采用所述目标摄像头进行交通灯识别。
一种电子设备,包括处理器、存储器及存储在所述存储器上并能够在所述处理器上运行的计算机程序,所述计算机程序被所述处理器执行时实现如上所述的交通灯识别的方法。
一种计算机可读存储介质,所述计算机可读存储介质上存储计算机程序,所述计算机程序被处理器执行时实现如上所述的交通灯识别的方法。
本发明实施例具有以下优点:
在本发明实施例中,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,通过获取多个摄像头采集图像数据,然后根据多个摄像头采集图像数据,确定目标摄像头,进而采用目标摄像头进行交通灯识别,实现了针对交通灯的多摄像头融合视觉感知识别,通过多个摄像头采集图像数据,以筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
为了更清楚地说明本发明的技术方案,下面将对本发明的描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1是本发明一实施例提供的一种交通灯识别的方法的步骤流程图;
图2是本发明一实施例提供的另一种交通灯识别的方法的步骤流程图;
图3是本发明一实施例提供的一种交通灯识别示例的示意图;
图4是本发明一实施例提供的另一种交通灯识别的方法的步骤流程图;
图5a是本发明一实施例提供的一种多个摄像头视野关系的示意图;
图5b是本发明一实施例提供的一种优先区域示例的示意图;
图5c是本发明一实施例提供的一种优先区域判断示例的示意图;
图5d是本发明一实施例提供的另一种优先区域判断示例的示意图;
图6是本发明一实施例提供的一种交通灯识别的装置的结构示意图。
为使本发明的上述目的、特征和优点能够更加明显易懂,下面结合附图和具体实施方式对本发明作进一步详细的说明。显然,所描述的实施例是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。
参照图1,示出了本发明一实施例提供的一种交通灯识别的方法的步骤流程图,该方法可以涉及多个摄像头,任意两个摄像头的视野可以存在覆盖关系,具体可以包括如下步骤:
步骤101,获取所述多个摄像头采集图像数据;
作为一示例,图像数据可以为针对车辆当前行车情况下的场景图像数据,该场景图像数据可以由车载摄像头采集,以用户交通灯识别,例如,可以通过预置的多个车载摄像头分别拍摄图片。
在识别交通灯的过程中,车辆可以设置多个车载摄像头,多个车载摄像头可以分别拍摄图片,进而可以获取当前行车情况下的多个摄像头采集图像数据,以进一步针对交通灯进行识别。
在一示例中,可以采用多目不同焦距、不同视角的摄像头,其中,多目摄像头中任意两个摄像头的视野可以存在覆盖关系,通过交通灯视觉感知识别***,可以基于融合多目不同焦距摄像头的视觉感知识别方法,进行交通灯识别。
在又一示例中,多目摄像头中不同视角的摄像头可以覆盖交通灯或远或近的多种情况,例如,自动驾驶***中可以配置三目摄像头,任意两个摄像头的视野可以存在覆盖关系,例如,摄像头1的视野可以完全被摄像头2的视野覆盖,摄像头2的视野可以完全被摄像头3的视野覆盖,可以采用同样的方法配置双目摄像头、四目摄像头,本发明对此不做限制。
步骤102,根据所述多个摄像头采集图像数据,确定目标摄像头;
其中,目标摄像头可以为针对交通灯识别筛选出的最合适摄像头,例如,交通灯视觉感知识别***可以切换多目摄像头,以选择最合适的摄像头作为目标摄像头。
在获取多个摄像头采集图像数据后,可以根据多个摄像头采集图像数据,从多个摄像头中筛选出的最合适用于交通灯识别的摄像头,作为目标摄像头。
例如,针对当前行车情况,交通灯视觉感知识别***可以无缝切换多目摄像头拍摄的图片,进而可以根据多个摄像头采集的当前行车场景图片,选择最合适的摄像头,以进一步采用该摄像头识别交通灯。
步骤103,采用所述目标摄像头进行交通灯识别。
在确定目标摄像头后,可以将该目标摄像头确定为针对交通灯识别的当前摄像头,进而可以采用目标摄像头进行交通灯识别。
由于在交通灯视觉感知识别***基于单目摄像头拍摄的图片进行识别时,若单目摄像头焦距长、看的远,但焦距的增加会造成视角变窄,在交通灯超出摄像头视角的情况下,将导致视觉感知识别***错误地认为未拍摄到交通灯图片,从而引发交通事故;而采用短焦距、宽视角的摄像头时,将导致摄像头看不远, 造成交通灯在摄像头视角中比较小,使视觉感知识别***无法清楚地识别出图片中交通灯,同样容易引发交通事故。
在一示例中,采用多目摄像头的交通灯视觉感知识别方法,通过融合不同焦距、不同视角摄像头,可以覆盖交通灯或远或近的多种情况拍摄图片,从而解决了因单目摄像头视角的局限性导致的交通灯遗漏或看不清的问题,提升了交通灯识别的可靠性。
在本发明实施例中,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,通过获取多个摄像头采集图像数据,然后根据多个摄像头采集图像数据,确定目标摄像头,进而采用目标摄像头进行交通灯识别,实现了针对交通灯的多摄像头融合视觉感知识别,通过多个摄像头采集图像数据,以筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
参照图2,示出了本发明一实施例提供的另一种交通灯识别的方法的步骤流程图,该方法可以涉及多个摄像头,任意两个摄像头的视野可以存在覆盖关系,具体可以包括如下步骤:
步骤201,获取所述多个摄像头采集图像数据;
在识别交通灯的过程中,车辆可以设置多个车载摄像头,多个车载摄像头可以分别拍摄图片,进而可以获取当前行车情况下的多个摄像头采集图像数据,以进一步针对交通灯进行识别。
在一示例中,多个摄像头可以按照视野大小排序,多个摄像头可以为三目摄像头,该三目摄像头可以包括鱼眼摄像头、主摄像头、窄摄像头,其中,按照视野大小排序可以为鱼眼摄像头视野可以覆盖主摄像头视野,主摄像头视野可以覆盖窄摄像头视野。在识别交通灯的过程中,可以获取鱼眼摄像头、主摄像头、窄摄像头三目不同视角的摄像头采集图像数据。
步骤202,判断第一摄像头采集的第一图像数据中是否检测到交通灯;
其中,第一摄像头可以为针对当前判断条件下两个摄像头中视野范围较大的摄像头,当前判断条件可以为针对视野覆盖关系排序相邻的两个摄像头。
作为一示例,第一图像数据可以为第一摄像头采集图像数据,例如,鱼眼摄像头针对当前行车场景拍摄的图片。
在获取多个摄像头采集图像数据后,可以从多个摄像头中确定第一摄像头,进而可以针对第一摄像头采集的第一图像数据进行检测,以判断是否检测到交通灯。
例如,多个摄像头可以为三目摄像头,该三目摄像头可以包括鱼眼摄像头、主摄像头、窄摄像头,在获取三目摄像头采集图像数据后,可以将视野排序中最大的鱼眼摄像头作为第一摄像头,进而可以判断鱼眼摄像头拍摄的图片中是否检测到交通灯。
在一示例中,可以通过交通灯检测网络对三目摄像头拍摄的图片进行交通灯检测,以判断在三目摄像头拍摄的图片中是否检测到交通灯。
步骤203,在判定所述第一摄像头采集的第一图像数据中检测到交通灯时,在第一摄像头采集的第一图像数据中,确定交通灯位置;
在判定第一摄像头采集的第一图像数据中检测到交通灯时,可以确定交通灯在第一图像数据中的区域 位置,作为第一图像数据中交通灯位置,例如,可以在摄像头拍摄的图片中检测到交通灯时,进一步确定交通灯呈现区域在图片中的几何区域位置。
步骤204,判断所述交通灯位置是否与第二摄像头相匹配;其中,所述第一摄像头的视野覆盖所述第二摄像头的视野;
作为一示例,第一摄像头和第二摄像头可以为排序相邻的摄像头,例如,第一摄像头和第二摄像头可以为按照视野大小排序中,视野覆盖关系相邻的两个摄像头。
在确定交通灯位置后,可以基于图片匹配的方法,通过预设的摄像头之间几何位置关系,判断第一摄像头采集的第一图像数据中交通灯位置是否与第二摄像头相匹配,该第一摄像头的视野可以覆盖第二摄像头的视野。
例如,在第一摄像头为鱼眼摄像头,第二摄像头为主摄像头的情况下,可以基于图片匹配的方法,确定较小视角的主摄像头场景在较大视角的鱼眼摄像头场景中的位置,进而可以根据摄像头之间确定的几何位置关系,检测交通灯是否在主摄像头场景中,以判断交通灯位置是否与第二摄像头相匹配。
步骤205,在判定所述交通灯位置与所述第二摄像头相匹配时,确定所述第二摄像头为目标摄像头;
在判定交通灯位置与第二摄像头相匹配时,可以将第二摄像头确定为从多个摄像头中筛选出的最合适用于交通灯识别的摄像头,以作为目标摄像头。
步骤206,采用所述目标摄像头进行交通灯识别。
在确定目标摄像头后,可以将该目标摄像头确定为针对交通灯识别的当前摄像头,进而可以采用目标摄像头进行交通灯识别。
在本发明一实施例中,还可以包括如下步骤:
在判定所述交通灯位置与所述第二摄像头不相匹配时,确定所述第一摄像头为目标摄像头。
在判定交通灯位置与第二摄像头不相匹配时,可以将第一摄像头确定为从多个摄像头中筛选出的最合适用于交通灯识别的摄像头,以作为目标摄像头。
在本发明一实施例中,还可以包括如下步骤:
在判定所述第一摄像头采集的第一图像数据中未检测到交通灯时,判断第二摄像头是否为视野最小的摄像头;在判定所述第二摄像头不为视野最小的摄像头时,将所述第二摄像头作为第一摄像头,并返回执行所述判断所述第一摄像头采集的第一图像数据中是否检测到交通灯;在判定所述第二摄像头为视野最小的摄像头时,输出未检测到交通灯的提示信息。
在判定第一摄像头采集的第一图像数据中未检测到交通灯时,可以针对第二摄像头进行判断,以确定第二摄像头是否为当前多目摄像头中视野最小的摄像头,进而在判定第二摄像头不为视野最小的摄像头时,可以将第二摄像头作为第一摄像头,并返回执行判断第一摄像头采集的第一图像数据中是否检测到交通灯的步骤,以进一步从多个摄像头中筛选出最合适用于交通灯识别的目标摄像头。
在一示例中,可以在确定第二摄像头为当前多目摄像头中视野最小的摄像头时,输出未检测到交通灯的提示信息,以表示采用多目摄像头融合的交通灯视觉感知方法,针对当前行车场景未检测到交通灯。
为了使本领域技术人员能够更好地理解上述步骤,以下结合图3对本发明实施例加以示例性说明,但 应当理解的是,本发明实施例并不限于此。
在识别交通灯的过程中,可以采用三目摄像头同时拍摄图片,该三目摄像头可以包括鱼眼摄像头、主摄像头、窄摄像头,在获取三目摄像头拍摄的图片后,由于按照视野大小排序,鱼眼摄像头为视野最大的摄像头,进而可以通过交通灯检测网络判断鱼眼摄像头采集的图片中是否检测到交通灯。
在判定鱼眼摄像头采集的图片中检测到交通灯时,由于鱼眼摄像头和主摄像头为视野排序相邻的摄像头,进而可以判断交通灯是否存在于主摄像头的场景中,若交通灯存在于主摄像头的场景中,可以采用主摄像头进行交通灯识别;若交通灯未存在于主摄像头的场景中,则可以采用鱼眼摄像头进行交通灯识别。
针对判断交通灯是否存在于主摄像头的场景中,可以采用图片匹配算法,通过确定主摄像头的场景在鱼眼摄像头中的位置,进而可以基于预先确定的主摄像头与鱼眼摄像头之间的几何位置关系,判断交通灯是否存在于主摄像头的场景中。
在判定鱼眼摄像头采集的图片中未检测到交通灯时,由于与鱼眼摄像头视野排序相邻的主摄像头在当前采用的三目摄像头中不为视野最小的摄像头,进而可以进一步判断主摄像头采集的图片中是否检测到交通灯。
在判定主摄像头采集的图片中检测到交通灯时,由于主摄像头和窄摄像头为视野排序相邻的摄像头,且主摄像头的视野覆盖窄摄像头的视野,进而可以判断交通灯是否存在于窄摄像头的场景中,若交通灯存在于窄摄像头的场景中,可以采用窄摄像头进行交通灯识别;若交通灯未存在于窄摄像头的场景中,则可以采用主摄像头进行交通灯识别。
针对判断交通灯是否存在于窄摄像头的场景中,同样可以采用图片匹配算法,通过确定窄摄像头的场景在主摄像头中的位置以进行判断。
在判定主摄像头采集的图片中未检测到交通灯时,由于与主摄像头视野排序相邻的窄摄像头在当前采用的三目摄像头中为视野最小的摄像头,进而可以输出未检测到交通灯的提示信息。
在本发明实施例中,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,通过获取多个摄像头采集图像数据,判断第一摄像头采集的第一图像数据中是否检测到交通灯,在判定第一摄像头采集的第一图像数据中检测到交通灯时,在第一摄像头采集的第一图像数据中,确定交通灯位置,然后判断交通灯位置是否与第二摄像头相匹配,第一摄像头的视野覆盖第二摄像头的视野,在判定交通灯位置与第二摄像头相匹配时,确定第二摄像头为目标摄像头,进而采用所述目标摄像头进行交通灯识别,实现了针对交通灯的多摄像头融合视觉感知识别,通过对多个摄像头采集图像数据中判断是否检测到交通灯,进而确定筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
参照图4,示出了本发明一实施例提供的另一种交通灯识别的方法的步骤流程图,该方法可以涉及多个摄像头,任意两个摄像头的视野可以存在覆盖关系,具体可以包括如下步骤:
步骤401,获取所述多个摄像头采集图像数据;
在识别交通灯的过程中,车辆可以设置多个车载摄像头,多个车载摄像头可以分别拍摄图片,进而可以获取当前行车情况下的多个摄像头采集图像数据,以进一步针对交通灯进行识别。
步骤402,在第一摄像头采集的第一图像数据中,确定交通灯位置;
在获取多个摄像头采集图像数据后,可以从多个摄像头中确定第一摄像头,进而可以针对第一摄像头采集的第一图像数据进行检测,可以在判定第一摄像头采集的第一图像数据中检测到交通灯时,进一步确定交通灯在第一图像数据中的区域位置,作为第一图像数据中交通灯位置,例如,可以在摄像头拍摄的图片中检测到交通灯时,进一步确定交通灯呈现区域在图片中的几何区域位置。
步骤403,获取所述第一摄像头和所述第二摄像头的相对位置关系;
在确定交通灯位置后,可以基于图片匹配的方法,获取第一摄像头和第二摄像头的相对位置关系,以通过该相对位置关系进一步确定交通灯位置是否与第二摄像头相匹配。
例如,如图5a所示,采用三目摄像头识别交通灯的过程中,该三目摄像头可以包括鱼眼摄像头、主摄像头、窄摄像头,其中,按照视野大小排序可以为鱼眼摄像头视野可以覆盖主摄像头视野,主摄像头视野可以覆盖窄摄像头视野。
在一示例中,基于图片匹配的方法,可以使用SIFT算法(Scale-invariant feature transform)、ORB算法(Oriented Fast and Rotated Brief)等特征点提取及匹配算法,得到三目摄像头中摄像头之间的相对位置关系,由于针对同一车辆,配置的多目摄像头相对位置固定,则各摄像头的视野相对位置关系可以长期保持不变,进而可以将各摄像头之间的视野相对位置关系进行离线计算并存储,以在识别交通灯的过程中能够直接调用摄像头之间的相对位置关系,不需多次计算。
步骤404,根据所述相对位置关系,在所述第一图像数据中确定优先图像区域;
在获取相对位置关系后,可以在第一图像数据中根据相对位置关系确定针对第二摄像头的优先图像区域,以检测优先图像区域进一步确定交通灯位置是否与第二摄像头相匹配。
例如,如图5b所示,通过针对摄像头之间的视野相对位置关系进行离线计算,可以在视野较大的第一摄像头视野中,确定针对该摄像头视野大小排序相邻(且小于该摄像头)的第二摄像头的几何位置,作为针对第二摄像头的优先图像区域(如优先区域B),进而可以将第一摄像头视野中的其余区域作为针对第二摄像头视野外的区域(如优先区域A、优先区域C)。
步骤405,检测所述优先图像区域是否包含所述交通灯位置;
在确定优先图像区域后,可以检测优先图像区域是否包含交通灯位置,以确定交通灯位置是否与第二摄像头相匹配。
例如,由于优先图像区域可以为第二摄像头在第一摄像头视野中的几何位置,进而可以在确定优先图像区域后,检测优先图像区域中是否包含针对第一摄像头采集的第一图像数据中交通灯位置。
步骤406,在检测到所述优先图像区域包含所述交通灯位置时,判定所述交通灯位置与第二摄像头相匹配;
在检测到优先图像区域包含交通灯位置时,可以判定交通灯位置与第二摄像头相匹配,例如,可以在针对第二摄像头的优先图像区域包含第一摄像头采集的第一图像数据中交通灯位置时,确定交通灯在第二摄像头的场景中。
例如,如图5c所示,针对交通灯是否存在于主摄像头的场景中进行判断时,可以在鱼眼摄像头中检测 到交通灯后,判断交通灯是否存在于优先区域A,该优先区域A可以为鱼眼摄像头视野中针对主摄像头视野外的区域,进而可以在判定交通灯存在于优先区域A时,确定交通灯存在于鱼眼摄像头的场景中;在判定交通灯未存在于优先区域A时,可以判断交通灯是否存在于优先区域B,该优先区域B可以为鱼眼摄像头视野中针对主摄像头视野的区域,进而可以在判定交通灯存在于优先区域B时,确定交通灯存在于主摄像头的场景中。
又如,如图5d所示,针对交通灯是否存在于窄摄像头的场景中进行判断时,可以在主摄像头中检测到交通灯后,判断交通灯是否存在于优先区域A,该优先区域A可以为主摄像头视野中针对窄摄像头视野外的区域,进而可以在判定交通灯存在于优先区域A时,确定交通灯存在于主摄像头的场景中;在判定交通灯未存在于优先区域A时,可以判断交通灯是否存在于优先区域B,该优先区域B可以为主摄像头视野中针对窄摄像头视野的区域,进而可以在判定交通灯存在于优先区域B时,确定交通灯存在于窄摄像头的场景中。
步骤407,在检测到所述优先图像区域未包含所述交通灯位置时,判定所述交通灯位置与第二摄像头不相匹配;
在检测到优先图像区域未包含交通灯位置时,可以判定交通灯位置与第二摄像头不相匹配,例如,可以在针对第二摄像头的优先图像区域未包含第一摄像头采集的第一图像数据中交通灯位置时,确定交通灯不存在于第二摄像头的场景中。
步骤408,在判定所述交通灯位置与所述第二摄像头相匹配时,确定所述第二摄像头为目标摄像头;
在判定交通灯位置与第二摄像头相匹配时,可以将第二摄像头确定为从多个摄像头中筛选出的最合适用于交通灯识别的摄像头,以作为目标摄像头。
步骤409,采用所述目标摄像头进行交通灯识别。
在确定目标摄像头后,可以将该目标摄像头确定为针对交通灯识别的当前摄像头,进而可以采用目标摄像头进行交通灯识别。
在一示例中,通过针对当前行车情况,交通灯视觉感知识别***可以无缝切换多目摄像头拍摄的图片,进而可以选择最合适的摄像头进行交通灯识别,从而避免了因单目摄像头视角的本身缺陷导致的交通灯遗漏或看不清的问题,提升了交通灯识别的可靠性。
在又一示例中,对比于采用机器学习的算法,每当数据发生变化时,需要重新训练算法,对数据要求高且算法结果不稳定。采用图片匹配算法可以得到摄像头之间的几何位置关系,进而可以准确地切换摄像头,以选择最合适的摄像头进行交通灯识别,鲁棒性更强,且与端到端的机器学习模型相比,具有可解释性。
在本发明实施例中,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,通过获取多个摄像头采集图像数据,在第一摄像头采集的第一图像数据中,确定交通灯位置,然后获取第一摄像头和第二摄像头的相对位置关系,根据相对位置关系,在第一图像数据中确定优先图像区域,检测优先图像区域是否包含交通灯位置,在检测到优先图像区域包含交通灯位置时,判定交通灯位置与第二摄像头相匹配,在检测到优先图像区域未包含交通灯位置时,判定交通灯位置与第二摄像头不相匹配,进而在判定交通灯位置与第二摄像头相匹配时,确定第二摄像头为目标摄像头,采用目标摄像头进行交通灯识别,实现了针对交通灯 的多摄像头融合视觉感知识别,通过多个摄像头采集图像数据,并检测优先图像区域是否包含交通灯位置,以筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
需要说明的是,对于方法实施例,为了简单描述,故将其都表述为一系列的动作组合,但是本领域技术人员应该知悉,本发明实施例并不受所描述的动作顺序的限制,因为依据本发明实施例,某些步骤可以采用其他顺序或者同时进行。其次,本领域技术人员也应该知悉,说明书中所描述的实施例均属于优选实施例,所涉及的动作并不一定是本发明实施例所必须的。
参照图6,示出了本发明一实施例提供的一种交通灯识别的装置的结构示意图,该装置可以涉及多个摄像头,任意两个摄像头的视野可以存在覆盖关系,具体可以包括如下模块:
图像数据获取模块601,用于获取所述多个摄像头采集图像数据;
目标摄像头确定模块602,用于根据所述多个摄像头采集图像数据,确定目标摄像头;
交通灯识别模块603,用于采用所述目标摄像头进行交通灯识别。
在本发明一实施例中,所述目标摄像头确定模块602包括:
交通灯位置确定子模块,用于在第一摄像头采集的第一图像数据中,确定交通灯位置;
判断子模块,用于判断所述交通灯位置是否与第二摄像头相匹配;其中,所述第一摄像头的视野覆盖所述第二摄像头的视野;
目标摄像头确定子模块,用于在判定所述交通灯位置与所述第二摄像头相匹配时,确定所述第二摄像头为目标摄像头。
在本发明一实施例中,所述装置还包括:
判定模块,用于在判定所述交通灯位置与所述第二摄像头不相匹配时,确定所述第一摄像头为目标摄像头。
在本发明一实施例中,所述判断子模块包括:
相对位置关系获取单元,用于获取所述第一摄像头和所述第二摄像头的相对位置关系;
优先图像区域确定单元,用于根据所述相对位置关系,在所述第一图像数据中确定优先图像区域;
检测单元,用于检测所述优先图像区域是否包含所述交通灯位置;
第一判定单元,用于在检测到所述优先图像区域包含所述交通灯位置时,判定所述交通灯位置与第二摄像头相匹配;
第二判定单元,用于在检测到所述优先图像区域未包含所述交通灯位置时,判定所述交通灯位置与第二摄像头不相匹配。
在本发明一实施例中,所述装置还包括:
第一判断模块,用于判断第一摄像头采集的第一图像数据中是否检测到交通灯;
第二判断模块,用于在判定所述第一摄像头采集的第一图像数据中检测到交通灯时,调用所述交通灯位置确定子模块。
在本发明一实施例中,所述装置还包括:
第二摄像头判断模块,用于在判定所述第一摄像头采集的第一图像数据中未检测到交通灯时,判断第二摄像头是否为视野最小的摄像头;
第二摄像头判定模块,用于在判定所述第二摄像头不为视野最小的摄像头时,将所述第二摄像头作为第一摄像头,并返回调用所述第一判断模块;
提示信息输出模块,用于在判定所述第二摄像头为视野最小的摄像头时,输出未检测到交通灯的提示信息。
在本发明一实施例中,所述多个摄像头按照视野大小排序,所述第一摄像头和所述第二摄像头为排序相邻的摄像头。
在本发明实施例中,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,通过获取多个摄像头采集图像数据,然后根据多个摄像头采集图像数据,确定目标摄像头,进而采用目标摄像头进行交通灯识别,实现了针对交通灯的多摄像头融合视觉感知识别,通过多个摄像头采集图像数据,以筛选出最合适的摄像头进行交通灯识别,避免了视角限制导致的交通灯遗漏情况,提升了交通灯识别的准确性和可靠性,能够有效减少交通事故的发生。
本发明一实施例还提供了一种电子设备,可以包括处理器、存储器及存储在存储器上并能够在处理器上运行的计算机程序,计算机程序被处理器执行时实现如上交通灯识别的方法。
本发明一实施例还提供了一种计算机可读存储介质,计算机可读存储介质上存储计算机程序,计算机程序被处理器执行时实现如上交通灯识别的方法。
对于装置实施例而言,由于其与方法实施例基本相似,所以描述的比较简单,相关之处参见方法实施例的部分说明即可。
本说明书中的各个实施例均采用递进的方式描述,每个实施例重点说明的都是与其他实施例的不同之处,各个实施例之间相同相似的部分互相参见即可。
本领域内的技术人员应明白,本发明实施例可提供为方法、装置、或计算机程序产品。因此,本发明实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且,本发明实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明实施例是参照根据本发明实施例的方法、终端设备(***)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理终端设备的处理器以产生一个机器,使得通过计算机或其他可编程数据处理终端设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理终端设备以特定方式工作的计算机可读存储器中,使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品,该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理终端设备上,使得在计算机或其他可编程终端设备上执行一系列操作步骤以产生计算机实现的处理,从而在计算机或其他可编程终端设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
尽管已描述了本发明实施例的优选实施例,但本领域内的技术人员一旦得知了基本创造性概念,则可对这些实施例做出另外的变更和修改。所以,所附权利要求意欲解释为包括优选实施例以及落入本发明实施例范围的所有变更和修改。
最后,还需要说明的是,在本文中,诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来,而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且,术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者终端设备不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者终端设备所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括所述要素的过程、方法、物品或者终端设备中还存在另外的相同要素。
以上对所提供的一种交通灯识别的方法和装置,进行了详细介绍,本文中应用了具体个例对本发明的原理及实施方式进行了阐述,以上实施例的说明只是用于帮助理解本发明的方法及其核心思想;同时,对于本领域的一般技术人员,依据本发明的思想,在具体实施方式及应用范围上均会有改变之处,综上所述,本说明书内容不应理解为对本发明的限制。
Claims (10)
- 一种交通灯识别的方法,其特征在于,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,所述方法包括:获取所述多个摄像头采集图像数据;根据所述多个摄像头采集图像数据,确定目标摄像头;采用所述目标摄像头进行交通灯识别。
- 根据权利要求1所述的方法,其特征在于,所述根据所述多个摄像头采集图像数据,确定目标摄像头,包括:在第一摄像头采集的第一图像数据中,确定交通灯位置;判断所述交通灯位置是否与第二摄像头相匹配;其中,所述第一摄像头的视野覆盖所述第二摄像头的视野;在判定所述交通灯位置与所述第二摄像头相匹配时,确定所述第二摄像头为目标摄像头。
- 根据权利要求2所述的方法,其特征在于,还包括:在判定所述交通灯位置与所述第二摄像头不相匹配时,确定所述第一摄像头为目标摄像头。
- 根据权利要求2或3所述的方法,其特征在于,所述判断所述交通灯位置是否与第二摄像头相匹配,包括:获取所述第一摄像头和所述第二摄像头的相对位置关系;根据所述相对位置关系,在所述第一图像数据中确定优先图像区域;检测所述优先图像区域是否包含所述交通灯位置;在检测到所述优先图像区域包含所述交通灯位置时,判定所述交通灯位置与第二摄像头相匹配;在检测到所述优先图像区域未包含所述交通灯位置时,判定所述交通灯位置与第二摄像头不相匹配。
- 根据权利要求2所述的方法,其特征在于,在所述在第一摄像头采集的第一图像数据中,确定交通灯位置之前,还包括:判断第一摄像头采集的第一图像数据中是否检测到交通灯;在判定所述第一摄像头采集的第一图像数据中检测到交通灯时,执行所述在第一摄像头采集的第一图像数据中,确定交通灯位置。
- 根据权利要求5所述的方法,其特征在于,还包括:在判定所述第一摄像头采集的第一图像数据中未检测到交通灯时,判断第二摄像头是否为视野最小的摄像头;在判定所述第二摄像头不为视野最小的摄像头时,将所述第二摄像头作为第一摄像头,并返回执行所述判断所述第一摄像头采集的第一图像数据中是否检测到交通灯;在判定所述第二摄像头为视野最小的摄像头时,输出未检测到交通灯的提示信息。
- 根据权利要求2所述的方法,其特征在于,所述多个摄像头按照视野大小排序,所述第一摄像头和所述第二摄像头为排序相邻的摄像头。
- 一种交通灯识别的装置,其特征在于,涉及多个摄像头,任意两个摄像头的视野存在覆盖关系,所述装置包括:图像数据获取模块,用于获取所述多个摄像头采集图像数据;目标摄像头确定模块,用于根据所述多个摄像头采集图像数据,确定目标摄像头;交通灯识别模块,用于采用所述目标摄像头进行交通灯识别。
- 一种电子设备,其特征在于,包括处理器、存储器及存储在所述存储器上并能够在所述处理器上运行的计算机程序,所述计算机程序被所述处理器执行时实现如权利要求1至7中任一项所述的交通灯识别的方法。
- 一种计算机可读存储介质,其特征在于,所述计算机可读存储介质上存储计算机程序,所述计算机程序被处理器执行时实现如权利要求1至7中任一项所述的交通灯识别的方法。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP20900756.6A EP3985554B1 (en) | 2020-08-12 | 2020-10-16 | Traffic light recognition method and apparatus |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010809487.3 | 2020-08-12 | ||
CN202010809487.3A CN111967370B (zh) | 2020-08-12 | 2020-08-12 | 一种交通灯识别的方法和装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022032861A1 true WO2022032861A1 (zh) | 2022-02-17 |
Family
ID=73364320
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/121562 WO2022032861A1 (zh) | 2020-08-12 | 2020-10-16 | 一种交通灯识别的方法和装置 |
Country Status (3)
Country | Link |
---|---|
EP (1) | EP3985554B1 (zh) |
CN (1) | CN111967370B (zh) |
WO (1) | WO2022032861A1 (zh) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271892A (zh) * | 2018-08-30 | 2019-01-25 | 百度在线网络技术(北京)有限公司 | 一种物体识别方法、装置、设备、车辆和介质 |
CN109492507A (zh) * | 2017-09-12 | 2019-03-19 | 百度在线网络技术(北京)有限公司 | 红绿灯状态的识别方法及装置、计算机设备及可读介质 |
US10339400B1 (en) * | 2018-12-18 | 2019-07-02 | Chongqing Jinkang New Energy Automobile Co., Ltd. | Traffic light detection using multiple cameras |
US10467487B1 (en) * | 2018-12-11 | 2019-11-05 | Chongqing Jinkang New Energy Automobile Co., Ltd. | Fusion-based traffic light recognition for autonomous driving |
CN111491093A (zh) * | 2019-01-26 | 2020-08-04 | 初速度(苏州)科技有限公司 | 一种摄像头视场角的调整方法和装置 |
Family Cites Families (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10497109B2 (en) * | 2015-10-07 | 2019-12-03 | Fuchs Consulting, Inc. | Time-lapse infrared thermography system and method for damage detection in large-scale objects |
CN107886033B (zh) * | 2016-09-30 | 2021-04-20 | 比亚迪股份有限公司 | 识别圆形交通灯的方法、装置及车辆 |
US10740627B2 (en) * | 2017-05-10 | 2020-08-11 | Fotonation Limited | Multi-camera vision system and method of monitoring |
CN110889314B (zh) * | 2018-09-10 | 2022-09-13 | 北京市商汤科技开发有限公司 | 图像处理方法、装置、电子设备、服务器及*** |
CN109325967B (zh) * | 2018-09-14 | 2023-04-07 | 腾讯科技(深圳)有限公司 | 目标跟踪方法、装置、介质以及设备 |
CN109326129B (zh) * | 2018-10-24 | 2022-07-15 | 东软集团股份有限公司 | 交通灯控制方法、装置、交通灯网关及存储介质 |
CN109598743B (zh) * | 2018-11-20 | 2021-09-03 | 北京京东尚科信息技术有限公司 | 行人目标跟踪方法、装置及设备 |
CN109583415B (zh) * | 2018-12-11 | 2022-09-30 | 兰州大学 | 一种基于激光雷达与摄像机融合的交通灯检测与识别方法 |
CN110008849A (zh) * | 2019-03-13 | 2019-07-12 | 北京小马智行科技有限公司 | 信号灯的识别方法、装置、存储介质和处理器 |
CN110737266B (zh) * | 2019-09-17 | 2022-11-18 | 中国第一汽车股份有限公司 | 一种自动驾驶控制方法、装置、车辆和存储介质 |
CN110619307B (zh) * | 2019-09-18 | 2020-09-18 | 广州文远知行科技有限公司 | 交通灯状态确定方法、装置、设备和存储介质 |
CN110567475B (zh) * | 2019-09-19 | 2023-09-29 | 北京地平线机器人技术研发有限公司 | 导航方法、装置、计算机可读存储介质及电子设备 |
CN110706282B (zh) * | 2019-10-31 | 2020-09-15 | 镁佳(北京)科技有限公司 | 全景***自动标定方法、装置、可读存储介质及电子设备 |
CN110853339A (zh) * | 2019-11-25 | 2020-02-28 | 华录易云科技有限公司 | 一种基于视频检测的行人过街自适应***及方法 |
CN111521186A (zh) * | 2020-05-27 | 2020-08-11 | 广州小鹏车联网科技有限公司 | 一种车辆定位方法及装置、车辆、存储介质 |
-
2020
- 2020-08-12 CN CN202010809487.3A patent/CN111967370B/zh active Active
- 2020-10-16 EP EP20900756.6A patent/EP3985554B1/en active Active
- 2020-10-16 WO PCT/CN2020/121562 patent/WO2022032861A1/zh unknown
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109492507A (zh) * | 2017-09-12 | 2019-03-19 | 百度在线网络技术(北京)有限公司 | 红绿灯状态的识别方法及装置、计算机设备及可读介质 |
CN109271892A (zh) * | 2018-08-30 | 2019-01-25 | 百度在线网络技术(北京)有限公司 | 一种物体识别方法、装置、设备、车辆和介质 |
US10467487B1 (en) * | 2018-12-11 | 2019-11-05 | Chongqing Jinkang New Energy Automobile Co., Ltd. | Fusion-based traffic light recognition for autonomous driving |
US10339400B1 (en) * | 2018-12-18 | 2019-07-02 | Chongqing Jinkang New Energy Automobile Co., Ltd. | Traffic light detection using multiple cameras |
CN111491093A (zh) * | 2019-01-26 | 2020-08-04 | 初速度(苏州)科技有限公司 | 一种摄像头视场角的调整方法和装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP3985554A4 * |
Also Published As
Publication number | Publication date |
---|---|
CN111967370A (zh) | 2020-11-20 |
EP3985554C0 (en) | 2023-12-06 |
EP3985554A1 (en) | 2022-04-20 |
EP3985554B1 (en) | 2023-12-06 |
EP3985554A4 (en) | 2022-06-01 |
CN111967370B (zh) | 2021-12-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10620005B2 (en) | Building height calculation method, device, and storage medium | |
US8861802B2 (en) | Face image prioritization based on face quality analysis | |
CN111222395A (zh) | 目标检测方法、装置与电子设备 | |
US9948869B2 (en) | Image fusion method for multiple lenses and device thereof | |
US20210174092A1 (en) | Image processing apparatus and method for feature extraction | |
JP2008158958A (ja) | 路面判別方法および路面判別装置。 | |
CN111507145A (zh) | 嵌入式车载环视***库位障碍物检测方法、***及装置 | |
CN111386530B (zh) | 车辆检测方法和设备 | |
CN110659547A (zh) | 物体识别方法、装置、车辆和计算机可读存储介质 | |
JP6516012B2 (ja) | 画像処理装置、物体認識装置、機器制御システム、画像処理方法およびプログラム | |
KR101236223B1 (ko) | 차선 검출 방법 | |
JP6635621B2 (ja) | 自動車用視覚システム及び視覚システムを制御する方法 | |
CN109740526B (zh) | 信号灯识别方法、装置、设备及介质 | |
US20120162412A1 (en) | Image matting apparatus using multiple cameras and method of generating alpha maps | |
JP2008004989A (ja) | 車両周囲表示装置 | |
WO2022032861A1 (zh) | 一种交通灯识别的方法和装置 | |
JP2005156199A (ja) | 車両検知方法及び車両検知装置 | |
JP2008165595A (ja) | 障害物検出方法、障害物検出装置、障害物検出システム | |
JP6089698B2 (ja) | 情報処理装置及び方法 | |
KR102188164B1 (ko) | 3차원 데이터를 이용한 도로 인식 방법 | |
JP2011120076A (ja) | 画像検出装置及び画像検出方法 | |
CN110309844B (zh) | 用于对光斑进行聚类的装置和方法 | |
JP2001116545A (ja) | 距離画像算出装置 | |
JP2014067320A (ja) | ステレオカメラ装置 | |
CN113066121A (zh) | 图像分析***和识别重复细胞的方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2020900756 Country of ref document: EP Effective date: 20210625 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |