CN117956120B - Insurance remote video exploration method and system based on self-adaptive communication selection - Google Patents
Insurance remote video exploration method and system based on self-adaptive communication selection Download PDFInfo
- Publication number
- CN117956120B CN117956120B CN202410347596.6A CN202410347596A CN117956120B CN 117956120 B CN117956120 B CN 117956120B CN 202410347596 A CN202410347596 A CN 202410347596A CN 117956120 B CN117956120 B CN 117956120B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- image
- damage
- communication
- accident scene
- Prior art date
- Legal status (The legal status 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 status listed.)
- Active
Links
- 230000006854 communication Effects 0.000 title claims abstract description 205
- 238000004891 communication Methods 0.000 title claims abstract description 204
- 238000000034 method Methods 0.000 title claims abstract description 48
- 230000006378 damage Effects 0.000 claims abstract description 163
- 238000004088 simulation Methods 0.000 claims abstract description 37
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 230000005540 biological transmission Effects 0.000 claims description 22
- 230000003044 adaptive effect Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 11
- 238000011897 real-time detection Methods 0.000 claims description 6
- 238000011835 investigation Methods 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 description 17
- 238000004422 calculation algorithm Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 7
- 238000012549 training Methods 0.000 description 7
- 238000013473 artificial intelligence Methods 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 6
- 230000003902 lesion Effects 0.000 description 6
- 238000013256 Gubra-Amylin NASH model Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000008439 repair process Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000008485 antagonism Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010835 comparative analysis Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010191 image analysis Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003706 image smoothing Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000009525 mild injury Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000009526 moderate injury Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/094—Adversarial learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
-
- 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
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Remote Sensing (AREA)
- Astronomy & Astrophysics (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Technology Law (AREA)
- Signal Processing (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses an insurance remote video exploration method and system based on self-adaptive communication selection, comprising the following steps: s1: the vehicle-mounted communication terminal detects information such as satellite communication, base station communication received signal strength and the like in real time; s2: calculating a communication switching value parameter, if the communication switching value parameter is larger than a set threshold value, using a base station to communicate, otherwise, using satellite communication: s3: transmitting the collected video frame images of the accident scene of the vehicle to an insurance company processing module for preprocessing; s4: inputting the preprocessed video frame image into a trained depth generation countermeasure network model to generate a simulation image; Calculating the structural similarity index of the simulated image and the accident scene image so as to judge the severity of the damage of the vehicle; s5: the damage degree of the vehicle is displayed on an insurance company display screen. According to the method, the accuracy and the efficiency of judging the damage degree of the vehicle are greatly improved by calculating the structural similarity index of the image.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an insurance remote video exploration method and system based on self-adaptive communication selection.
Background
With the rapid development of mobile communication technology, satellite communication is a powerful complement to terrestrial communication, and exhibits unique advantages in specific situations, especially in remote areas or in situations where terrestrial base station communication coverage is not available. In the modern insurance industry, rapid and accurate assessment of the extent of damage to a vehicle accident is critical to improving customer satisfaction and reducing operating costs. However, conventional insurance claim procedures often require on-site surveyors to go to the accident site, which is not only time consuming and labor intensive, but also difficult to implement in some remote areas or under severe weather conditions. Therefore, how to improve the insurance remote video exploration method by using advanced communication technology and artificial intelligence algorithm becomes a problem to be solved in industry.
Currently, challenges commonly faced by insurance companies in handling vehicle accident claims include: the on-site investigation is inefficient, particularly in areas of inconvenient traffic or bad weather; the evaluation of the damage degree of the accident depends on personal experience and is easy to generate errors; the quality of transmission of live pictures or video material is limited by the communication conditions, especially in areas where the coverage of the base station is insufficient. In the prior art, the advantages of satellite communication and base station communication are not fully combined to perform selective communication, and the advantages of satellite communication are: wide area coverage, strong communication capability, high flexibility and strong anti-interference performance; disadvantages of satellite communications: signal delay, high cost and great weather effect; advantages of base station communication: low delay, low cost, stable signal and high reliability; drawbacks of base station communication: geographic limitations, limited capacity for base station communications, and long deployment times.
In recent years, with the rapid development of artificial intelligence technology, particularly satellite communication, deep learning and video image processing, an analysis method based on artificial intelligence provides a new idea for solving the problems. For example, convolutional Neural Networks (CNNs) have shown excellent performance in the fields of image recognition, natural language processing, etc., and satellite communications have been presented on a large number of terminal devices, providing conditions for communication mode selection, and these communicated data are pre-trained on large amounts of data by artificial intelligence, enabling rich feature representations to be learned, enabling rapid and accurate analysis and prediction on specific tasks.
Although satellite communication technology and artificial intelligence are successfully applied in many fields, in the existing vehicle accident damage degree judging process, only analysis of vehicle accident scene images or judgment by combining with longer historical images of vehicles are considered, and comparison judgment after restoration is not performed by combining with images of vehicles at the moment before the accident occurs, so that the vehicle damage degree judgment is inaccurate; and how to apply the advanced communication technology to the traffic accident insurance processing field, especially how to effectively combine the advantages and disadvantages of different communication modes and select a reasonable communication mode according to different communication conditions is not considered; in addition, how to comprehensively analyze and judge the damage degree of the vehicle by combining the video image data of the accident scene of the vehicle and the image data of the vehicle before the accident and how to design a reasonable algorithm frame to improve the accuracy and the efficiency of damage evaluation are still technical challenges to be solved at present. There is therefore an urgent need for a new solution to improve the efficiency of the insurance communication process, the accuracy of the impairment determination and the customer satisfaction.
Disclosure of Invention
In order to solve the above-mentioned problems in the prior art, the present invention provides a method and a system for insurance remote video exploration based on adaptive communication selection, which intelligently selects an optimal communication mode (satellite communication or ground base station communication) to transmit video frame images of a vehicle accident scene, and automatically generates a simulation image similar to a vehicle state before an accident by using a trained depth generation countermeasure network model. The damage severity of the vehicle can be accurately judged by calculating the Structural Similarity Index (SSIM) of the simulation image and the accident scene image, so that scientific and objective claim settlement basis is provided for insurance companies, the accuracy and efficiency of judging the damage degree of the vehicle are greatly improved, and the user experience is greatly improved.
The application provides an insurance remote video exploration method based on self-adaptive communication selection, which comprises the following steps:
s1: vehicle-mounted communication terminal real-time detection satellite communication received signal strength Transmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delay/>Bandwidth utilization/>；
S2: calculating communication handover value parametersIf the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
S3: the vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module, and preprocesses the video frame images of the vehicle accident scene;
S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image ; Calculating analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The proportion of the image damage area of the accident scene to the vehicle image;
s5: the damage degree of the vehicle is displayed on an insurance company display screen.
Preferably, the step S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
Preferably, the preprocessing of the video frame image of the vehicle accident scene comprises denoising processing by adopting Gaussian filtering.
Preferably, the step S5: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone.
Preferably, the index of similarity according to structureJudging the severity of damage to the vehicle, when/>The severity of the vehicle damage is slight when/>The severity of the vehicle damage is significant when/>When the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
The application also provides an insurance remote video exploration system based on self-adaptive communication selection, which comprises:
Real-time detection module of vehicle-mounted communication terminal for detecting satellite communication received signal strength in real time Transmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delay/>Bandwidth utilization/>；
Communication handover value parameterA calculation module for calculating the communication switching value parameter/>If the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
The vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module to preprocess the video frame images of the vehicle accident scene;
Vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image ; Calculating analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The proportion of the image damage area of the accident scene to the vehicle image;
And a display module: the damage degree of the vehicle is displayed on an insurance company display screen.
Preferably, the vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
Preferably, the preprocessing of the video frame image of the vehicle accident scene comprises denoising processing by adopting Gaussian filtering.
Preferably, the display module: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone.
Preferably, the index of similarity according to structureJudging the severity of damage to the vehicle, when/>The severity of the vehicle damage is slight when/>The severity of the vehicle damage is significant when/>When the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
The invention provides an insurance remote video exploration method and system based on self-adaptive communication selection, which can realize the following beneficial technical effects:
1. the invention intelligently selects the optimal communication mode (satellite communication or ground base station communication) to transmit the video frame image of the vehicle accident scene, and automatically generates a simulation image similar to the vehicle state before the accident by utilizing the trained depth generation countermeasure network model. By calculating the Structural Similarity Index (SSIM) of the simulation image and the accident scene image, the method can accurately judge the damage severity of the vehicle, thereby providing scientific and objective claim settlement basis for insurance companies, greatly improving the accuracy and efficiency of judging the damage severity of the vehicle and greatly increasing the user experience.
2. The invention adopts the calculation of the communication switching value parameterIf the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used: by considering the intensity weight of the received signal of the base station communication, the signal-to-noise ratio weight, the network delay weight, the bandwidth utilization ratio weight, the satellite communication received signal intensity weight, the transmission delay weight and the satellite antenna weight, the advantages and disadvantages of the satellite communication and the base station communication can be fully combined for selective communication, the communication cost is greatly reduced, and the stability and the accuracy of the communication are improved.
3. The invention generates a simulated image through a depth generation countermeasure network modelCalculate the analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
in calculating the structural similarity index In the process, the proportion of the accident scene image damage area to the vehicle image is included in the calculation process, so that the accuracy of judging the vehicle damage degree is greatly enhanced, and an opinion suggestion of accuracy can be given to a vehicle insurance solution.
4. The application inputs the preprocessed video frame image of the accident scene of the vehicle into the trained depth generation countermeasure network model, and the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained depth generation countermeasure network model is trained by adopting a sample data set, and a simulation image is generated by the depth generation countermeasure network model, namely, an image before an accident is generated and an image after the accident is compared, so that the accuracy of the damage degree is greatly improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of steps of an insurance remote video exploration method based on adaptive communication selection in accordance with the present invention;
FIG. 2 is a schematic diagram of an insurance remote video survey system based on adaptive communication selection in accordance with the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1:
in view of the above-mentioned problems mentioned in the prior art, it is shown in fig. 1: the application provides an insurance remote video exploration method based on self-adaptive communication selection, which comprises the following steps:
s1: vehicle-mounted communication terminal real-time detection satellite communication received signal strength Transmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delay/>Bandwidth utilization/>；
In one embodiment, the vehicle is equipped with an advanced on-board communication terminal having the capability of receiving satellite communication and base station communication signals simultaneously, the terminal having built-in sensors and modules for real-time monitoring of the following parameters: satellite communication received signal strength, satellite communication transmission delay, satellite antenna azimuth and pitch angle, base station communication received signal strength, base station communication signal-to-noise ratio, network delay and bandwidth utilization; and 2, monitoring communication conditions in real time, wherein an accident of a vehicle occurs on a remote mountain area, and the vehicle-mounted communication terminal starts to monitor parameters of satellite and base station communication in real time. For example, the base station communication may have a received signal strength of only-95 dBm, and a very low signal to noise ratio, a network delay as high as 500 ms, and a bandwidth utilization approaching 90%, indicating poor quality of communication through the base station, which is unsuitable for transmitting large amounts of data. Meanwhile, the received signal strength of satellite communication is-75 dBm, the transmission time delay is 600 ms, but the satellite communication is not affected by terrain, and stable communication connection can be maintained after the azimuth angle and the pitch angle of the antenna are adjusted to the optimal positions; step 3, intelligent selection of a communication mode, wherein an intelligent algorithm built in the vehicle-mounted communication terminal calculates a communication switching value parameter, the algorithm comprehensively considers all monitored communication condition parameters and compares the communication switching value parameter with a preset threshold value, and the communication switching value parameter calculated by the algorithm determines that satellite communication is used as a more preferable choice; step 4, transmitting the video frame image of the accident scene to a server of an insurance company by the vehicle-mounted communication terminal through satellite communication, wherein the transmission of the image data can be kept stable even in a remote mountain area due to the selection of the satellite communication; and 5, evaluating the damage degree of the vehicle, and inputting the image into a pre-trained depth to generate an countermeasure network model after the image is received by the insurance company server. The model generates a simulated image similar to the vehicle state prior to the occurrence of the accident and calculates a Structural Similarity Index (SSIM) between the simulated image and the actual accident scene image; based on the SSIM value, the intelligent algorithm evaluates the severity of the vehicle damage and feeds back the evaluation to the insurance company for quick processing of claims. The result shows that the damage degree of the vehicle is displayed on the display screen of the insurance company, so that the claims settlement staff can quickly know the accident situation and provide efficient claims settlement service for clients.
S2: calculating communication handover value parametersIf the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
In one embodiment, the vehicle-mounted communication terminal at the vehicle accident scene monitors the following communication parameters in real time: base station communication received signal strength: -85dBm, signal to noise ratio: 10 dB, network delay: 300ms, bandwidth utilization: 75%, satellite communication received signal strength: -70 dBm, transmission delay: 500 ms, the optimal azimuth angle and pitch angle of the satellite antenna are adjusted to the position capable of receiving the strongest signal; step 2, defining weight factors, namely defining the following weight factors for calculating communication switching value parameters: base station communication received signal strength weighting factors: 0.25, signal to noise ratio weighting factor: 0.15, network delay weight factor: -0.1, bandwidth utilization weighting factor: -0.1, satellite communication received signal strength weighting factor: 0.3, transmission delay weight factor: -0.05; step 3, calculating the communication switching value parameter If the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used, and stable communication connection can be maintained after the azimuth angle and the elevation angle of the antenna are adjusted to the optimal positions.
S3: the vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module, and preprocesses the video frame images of the vehicle accident scene; the transmission communication mode adopts satellite communication or base station communication determined in the step S2;
In one embodiment, an accident occurs on a highway, a vehicle-mounted video frame image acquisition terminal is used for acquiring video frame image data of an accident scene after the accident occurs, the video data is stored in a memory of a vehicle-mounted communication terminal in a frame form, the video frame image is transmitted, and the vehicle-mounted communication terminal selects a process (S2) through a previous communication mode to transmit the data by using satellite communication due to the accident occurring in an area with weak communication signals of a base station. The accident video frame images are safely and stably uploaded to a server of an insurance company through satellite communication, the video frame images are preprocessed, and the received accident scene video frame images are firstly preprocessed by a series of preprocessing steps at the server end of the insurance company so as to improve the image quality and prepare for subsequent analysis: denoising: and denoising each frame of image by adopting a Gaussian filter algorithm. Since various illumination conditions may exist at the accident site, the gaussian filtering helps to reduce image noise and improve image quality. Brightness and contrast adjustment: the brightness and contrast of the video frame are automatically adjusted, so that details of the accident scene are ensured to be more clearly visible. Size standardization: all video frames are adjusted to a uniform resolution, for example 1080p or higher resolution per frame image, to ensure consistency of image analysis, taking into account the need for subsequent processing. Frame selection: key frames, i.e., the moment of occurrence of the accident and key pictures before and after the accident, are selected from the uploaded video to reduce the amount of data processed while ensuring the accuracy of damage assessment. Through the preprocessing steps, the quality of the video frame image of the accident scene is obviously improved, and clear and reliable input data is provided for further using depth generation countermeasure network (GAN) to evaluate the damage degree of the vehicle. Through the optimized video frames, the insurance company can analyze the accident situation more accurately and provide quick and accurate claim settlement service for clients.
S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; Calculating analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The damaged area is the proportion of the accident scene image to the vehicle image.
In one embodiment, a depth generation countermeasure network model is trained, specifically for processing vehicle accident images, which model includes two main components: generator (Generator): the arbiter (Discriminator) is responsible for distinguishing whether the image is an actual incident scene image or a simulated image produced by the generator, in charge of generating a simulated image, i.e., an image that is as close as possible to the state of the vehicle before the incident occurred. The model uses a number of vehicle images labeled "pre-accident" and "post-accident" during the training phase. In this way, the generator learns how to generate a "pre-accident" simulated image corresponding thereto based on the "post-accident" image. After an automobile collides with another automobile on an urban road, the front end of the automobile is seriously damaged, and after an accident occurs, a series of video frame images of accident sites are automatically recorded by an on-board communication terminal on the automobile, and the video frame images are optimized through the preprocessing steps. The key frame images selected by the image input are transmitted to a server of an insurance company through a network and then are input into a trained GAN model, and after the generator part of the simulated image generation GAN model receives the image of the accident scene, the simulated image generation GAN model starts to work so as to try to generate a simulated image showing the state of the vehicle before the accident occurs. For example, if the input image shows that the front of the automobile is severely damaged, the generator attempts to reconstruct a pattern of the front of the automobile before it is damaged. Evaluation of results: the generated simulated images are then provided to a arbiter, and to an incident assessment team for comparative analysis with the actual incident scene images. By comparing the generated simulated image with the actual accident scene image, the insurance company can accurately evaluate the degree of damage to the vehicle. This approach not only improves the accuracy of the assessment, but also greatly speeds up the processing of claim requests, and if the differences between the generated simulated image and the scene image are concentrated primarily at the front of the vehicle, then damage assessment emphasis will be placed on this area as well.
In one embodiment, the deep Generation Antagonism Network (GAN) is composed of two main parts: a Generator (Generator) and a discriminator (Discriminator). The two parts compete with each other in a antagonistic learning process, thereby improving the performance of the whole model. The GAN model structure includes a Generator (Generator) function: the task of the generator is to generate as realistic data as possible from random noise or specific inputs to fool a discriminant, typically comprising a multi-layer neural network, such as a Convolutional Neural Network (CNN) or a fully connected layer (DENSE LAYERS), to convert the input random noise into output data (e.g., images) with a specific morphology by layer-by-layer processing. In vehicle damage assessment applications, the generator input may be random noise in combination with post-accident vehicle state information, or directly post-accident image features. The generator outputs an image of the vehicle before the simulated accident. The discriminator (Discriminator) functions: the task of the arbiter is to distinguish whether the input image is real or a false image generated by the generator. The structure is as follows: also, a multi-layer neural network structure such as a Convolutional Neural Network (CNN) is employed, focusing on extracting image features to judge images, generated images from a generator, and vehicle images before an accident occurs. And (3) outputting: a probability value representing the likelihood that the input image is a true image, the principle of generating a simulated image, during which the generator and the arbiter learn countermeasures by: training a discriminator: first, parameters of the generator are fixed, and the capacity of the discriminator is improved. In this step, the arbiter receives two part inputs: one part is the image generated by the generator and the other part is the image of the vehicle before the actual accident occurs, and the objective of the discriminator is to correctly classify whether these input images are actual or generated. Training generator: subsequently, the parameters of the discriminators are fixed, the capability of the generator is improved, the generator tries to generate images which are enough to deceive the discriminators, namely, the discriminators misjudge the generated images as real images, and the aim of the generator is to maximize classification errors of the discriminators, and iterative optimization is carried out: the two steps are repeated, the parameters of the generator and the discriminator are continuously and iteratively optimized, and as training is carried out, the generator becomes more and more good at generating images, and the discriminator becomes more and more good at distinguishing images.
In some embodiments, in the application of vehicle damage assessment, the generator, after appropriate training, can generate corresponding simulated images of the vehicle state before the occurrence of the accident from the vehicle images after the accident. The comparison between these simulated images and the actual scene of an accident images can be used to assess the extent of damage to the vehicle. The simulated image generation uses the GAN technique to generate a simulated image of the vehicle before the accident according to the state of the vehicle after the accident. This image shows the intact state of the vehicle before it is damaged. Image contrast evaluation: the generated simulated image is compared with an image of the actual accident scene, in particular the area of concern for vehicle damage. Step 2: evaluation of the severity of damage by comparing the two images (simulated undamaged image and post-accident damaged image), the evaluator can intuitively identify the location and extent of the damage to the vehicle. The evaluation is based mainly on the following aspects: the visibility of the damaged area is improved, if the front end of the vehicle is obviously damaged by the image after the accident, and the front end of the simulation image is intact, the evaluators can immediately identify the damaged area; degree of damage: the evaluators judge the severity of the lesions by observing the size, depth and extent of the effect of the lesions. For example, if there are only a few scratches and depressions, this may be rated as a "slight" damage; if the front end impact causes significant deformation, it may be rated as a "significant" damage. The scene image shows that the front bumper of the vehicle is severely damaged, while from the simulated image generated by the GAN, the front bumper of the vehicle is intact. By comparing the two figures, the evaluators can clearly see the area and extent of the damage. If damage results in damage to the overall structure of the front bumper, this may be directly judged as moderate to significant damage. The insurance company determines the amount of claims accordingly and feeds back the extent of damage to the vehicle owner.
S5: the damage degree of the vehicle is displayed on an insurance company display screen.
In one embodiment, a vehicle is involved in a collision accident in urban areas, and the front end of the vehicle is significantly damaged. After the accident occurs, the vehicle owner submits an accident report to the insurance company through the mobile phone application, and uploads a video of the accident scene. The vehicle-mounted communication terminal also automatically records the accident occurrence and transmits the data back to the insurance company. And 1, collecting and processing data, and automatically triggering an accident handling flow after the server of the insurance company receives the accident data. The method comprises the steps of preprocessing an accident scene video frame image and generating a simulated vehicle state image before accident occurrence through a depth generation countermeasure network (GAN) model. And 2, evaluating the damage degree, wherein the system compares the preprocessed actual accident scene image with the generated simulation image, and evaluates the damage degree of the vehicle by using algorithms such as a Structural Similarity Index (SSIM) and the like. Based on comparative analysis, the system classifies the extent of damage into three categories, "mild", "moderate" and "significant". And 3, displaying the result, and displaying the evaluation result and related information on a display screen in the insurance company by the system after the damage degree is evaluated, and simultaneously providing information for a vehicle owner through a client service platform of the insurance company. And (3) displaying an internal display screen: in the claims department of the insurance company, pictures of accident vehicles (including actual pictures of accident sites and generated simulation pictures), vehicle information, accident profiles and damage degree assessment results are displayed on a display screen. In addition, the proposed claims amount is presented and the advice is processed for the claims personnel to review and make decisions. Customer service platform demonstrates: the vehicle owner may view the damage assessment results of his vehicle, including the extent of damage and possible repair advice, through the insurance company's cell phone application or website. This increases the transparency of the claim settlement process, enabling the vehicle owner to track the claim settlement progress in real time. By the method, the damage degree of the vehicle is displayed, and an insurance company can accelerate the claim settlement process and improve the working efficiency. Meanwhile, the vehicle owners have better understanding and satisfaction degree on the claim settlement process, so that the service quality of clients is improved.
In one embodiment, the step S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
In one embodiment, one vehicle is impacted by another vehicle in a parking lot, resulting in damage to the rear bumper and tail lights. Owners submit videos and pictures of incidents to insurance companies, and depth Generation Antagonism Network (GAN) techniques are used to evaluate the extent of damage in order to quickly and accurately process claims. Step 1: data collection and preprocessing, the insurance company pre-processes the video and pictures collected from the vehicle owners, including image denoising, brightness adjustment and size standardization, to ensure that the images input to the GAN model are clear and consistent. Step 2: using GAN to generate a simulated image, a generator task attempts to reconstruct a simulated image of the vehicle state prior to the occurrence of the accident from the damaged vehicle image of the accident scene. This requires that the generator be able to understand the structure of the vehicle and infer an undamaged raw state based on the damaged portion. The task of the discriminator: the arbiter receives two classes of images, one class being the simulated image generated by the generator and the other class being the actual vehicle image (obtained from a database of similar vehicle models) before the accident. The task of the arbiter is to distinguish between these two classes of images while improving the authenticity of the generated images. During the training phase, the generator and arbiter boost performance through constant iterative antagonism. The generator tries to generate more and more realistic simulated images, while the arbiter strives to improve the ability to distinguish between genuine and fake images. Step 3: damage assessment once training is completed, the generator generates a simulated pre-accident vehicle image from the submitted damaged vehicle image for the particular claim case. By comparing the damaged image of the accident scene with the generated simulated image, particularly focusing on the area of the rear bumper and the tail light, the evaluator can intuitively recognize the location and extent of the damage. The accident scene image shows that the rear bumper has obvious pits and scratches and the tail lamp is broken. With GAN technology, the generator successfully reconstructs a vehicle simulation image of an undamaged rear bumper and an intact tail light. The arbiter evaluates that this simulated image is very close to the image of the vehicle before the actual accident, demonstrating the high quality of the generated image. Evaluation of the damage degree: by comparing the scene of accident image with the generated simulated image, the apparent visible damage allows the evaluator to quickly and accurately assess the extent of the damage as "medium", and calculate a reasonable claim amount therefrom. And (3) processing the claims: the evaluation result and the claim amount are recorded through an internal system of the insurance company and are notified to the vehicle owner through an email and a mobile phone application, so that the claim settlement process is quickened, and the customer satisfaction is improved.
In one embodiment, the preprocessing of the video frame image of the vehicle accident scene comprises denoising by Gaussian filtering. There is more noise in the video frame image that may interfere with subsequent image analysis and evaluation of the extent of damage. Therefore, after receiving the accident video, the insurance company decides to pre-process the video frame image first to improve the image quality, including denoising with gaussian filtering. Gaussian filter denoising processing principle: gaussian filtering is an image smoothing technique that can effectively remove high frequency noise from an image. It is achieved by replacing the value of each pixel point in the image with a weighted average of the pixel values around that point, the weights being determined by a gaussian function, with the pixels closer to the center having higher weights. The implementation steps are as follows: selecting an appropriate core size: depending on the resolution and noise level of the accident video, an appropriate kernel size (e.g., 5x5 or 3x 3) is selected for gaussian filtering. Calculating Gaussian weights: and calculating a weight matrix of the filter according to the Gaussian distribution formula and the kernel size. A filter is applied: and (3) applying a Gaussian filter to each pixel point for each frame of image in the video, and calculating weighted average so as to achieve the purpose of denoising. The processed image is effectively removed while the edges and details are maintained, so that the details of the accident are more clearly visible, and a high-quality image is provided for subsequent damage evaluation. In the image before processing, the damaged part of the front end of the car appears blurred due to night shooting, and a plurality of noise points interfere with the visual effect. After Gaussian filtering denoising treatment is applied, the image becomes smoother, noise points are reduced, the damage degree and detail of damaged parts become clearer, and assessment staff of insurance companies can judge the responsibility and damage degree of accidents more accurately.
In one embodiment, the step S5: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone. Damage assessment: the professional assessment team determines the specific location and extent of the vehicle damage by analyzing videos and pictures of the accident scene. Meanwhile, vehicle simulation images generated by using the GAN technology before accidents occur help an assessment team to judge the damage degree more accurately. Degree of damage classification: the degree of damage is classified into three categories, namely "mild", "moderate" and "significant", depending on the severity of the damage. In this example, the damage level was rated "medium" because the right side door and front bumper of the vehicle need to be replaced. An internal display system: the damage evaluation report of the car is displayed on an internal information system of the insurance company through an LCD computer screen, and comprises images of accident scene, damage degree evaluation, suggested maintenance measures and estimated claim settlement. This information helps the claims department to quickly determine the details of the claims and the process flow. Customer service platform: and (3) mobile phone application display: through the mobile phone application of the insurance company, the vehicle owner can directly check the damage evaluation result of the vehicle, including a concise damage report, photos and videos of damage parts, damage degree ("medium") and estimated claim amount. In addition, the application provides a feedback option for the owner to query or request more detailed assessment information. The insurance company also sends a detailed version of the damage-assessment report to the vehicle owner via email, including details of the analysis of the incident and the subsequent claims.
In one embodiment, the index of similarity according to structureJudging the severity of damage to the vehicle whenThe severity of the vehicle damage is slight when/>The severity of the damage to the vehicle is significant whenWhen the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
In one embodiment, one vehicle is inadvertently bumped by another while parked, causing damage to the vehicle body. To evaluate the extent of damage, insurance companies decide to use a Structural Similarity Index (SSIM) to compare the actual image of the vehicle after the accident with a simulated image of the vehicle before the accident generated by a deep generation countermeasure network (GAN). The evaluation process, the insurance company sets two thresholds of SSIM to determine the extent of damage: first threshold (higher): for distinguishing "mild" lesions from more severe lesions; second threshold (lower): for distinguishing "medium" from "significant" lesions. Slight damage: the SSIM score for a vehicle was 0.95, above the first threshold (assuming 0.9), which means that the post-accident and pre-accident images were very similar, with lesions rated as "mild". May be only a scratch or a slight depression of the surface, without affecting the function of the vehicle. Moderate damage: if the SSIM score of another vehicle is 0.75, between the first threshold (0.9) and the second threshold (0.7), this indicates that the damage to the vehicle is more severe than the mild damage, but not so severe that the damage is rated "medium". This situation may require more complex maintenance, such as replacement of parts. Significant damage: if the SSIM score for the third vehicle is 0.65, below the second threshold (assuming 0.7), this means that the difference from the simulated pre-accident image is large, and the damage is rated as "significant". This may involve damage to the vehicle structure, requiring extensive repair or even scrapping. The insurance company displays the evaluation results to claim processing personnel through an internal system and informs the vehicle owners through a client service platform. For example: for cases rated as "mild" damage, insurance companies may suggest simple appearance repairs and complete claim settlement quickly; for "moderate" damage situations, it may be necessary to schedule the vehicle for more detailed inspection and repair; for "major" damage, insurance companies will conduct more extensive surveys, possibly involving higher claims and long-term maintenance procedures.
The application also provides an insurance remote video exploration system based on self-adaptive communication selection, as shown in fig. 2, comprising: real-time detection module of vehicle-mounted communication terminal for detecting satellite communication received signal strength in real timeTransmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delayBandwidth utilization/>；
In some embodiments, the hardware includes a vehicle-mounted communication terminal that functions to capture video and image data of an accident scene, perform preliminary data processing, such as compression and format conversion, and includes a high-definition camera, a data processing unit (CPU or GPU), a built-in storage, and a satellite communication interface. The satellite communication module is used for providing communication capability between the vehicle-mounted terminal and the satellite and supporting uploading and downloading of data. The satellite communication module functional components include a satellite antenna, a modem, and a signal converter. Accident video and image data from the vehicle is received and stored, and further data processing and damage-assessment algorithms are performed. The high performance server is equipped with sufficient storage space and computing power to support big data processing and deep learning models. There are also vehicle damage assessment results presented, providing claims progress tracking and customer communication services, consisting of Web servers, databases and user interfaces (including Web interfaces and mobile applications). And after the accident of the vehicle occurs, the vehicle-mounted communication terminal automatically or manually starts the camera to record video and image data of the accident scene. The vehicle-mounted terminal compresses and converts the format of the captured data to prepare for transmission. And the vehicle-mounted communication terminal uploads the preprocessed data to the insurance company server through a satellite through the satellite communication module. The satellite communication module establishes connection with the satellite by using an antenna, and completes data transmission and reception by a modem and a signal converter. After receiving the data, the insurance company server further analyzes the video and the image, generates a simulation image by using a deep learning algorithm, calculates a Structural Similarity Index (SSIM), and evaluates the damage degree of the vehicle. The damage evaluation result is displayed to the vehicle owner through the client service platform, and the vehicle owner can check the damage degree, the estimated repair cost and the claim settlement progress through a Web interface or a mobile application.
Communication handover value parameterA calculation module for calculating the communication switching value parameter/>If the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
The vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module to preprocess the video frame images of the vehicle accident scene;
Vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image ; Calculating analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
/>
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The proportion of the image damage area of the accident scene to the vehicle image;
And a display module: the damage degree of the vehicle is displayed on an insurance company display screen.
In one embodiment, the vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
In one embodiment, the preprocessing of the video frame image of the vehicle accident scene comprises denoising by Gaussian filtering.
In one embodiment, the display module: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone.
In one embodiment, the index of similarity according to structureJudging the severity of damage to the vehicle whenThe severity of the vehicle damage is slight when/>The severity of the damage to the vehicle is significant whenWhen the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
The invention provides an insurance remote video exploration method and system based on self-adaptive communication selection, which can realize the following beneficial technical effects:
1. the invention intelligently selects the optimal communication mode (satellite communication or ground base station communication) to transmit the video frame image of the vehicle accident scene, and automatically generates a simulation image similar to the vehicle state before the accident by utilizing the trained depth generation countermeasure network model. By calculating the Structural Similarity Index (SSIM) of the simulation image and the accident scene image, the method can accurately judge the damage severity of the vehicle, thereby providing scientific and objective claim settlement basis for insurance companies, greatly improving the accuracy and efficiency of judging the damage severity of the vehicle and greatly increasing the user experience.
2. The invention adopts the calculation of the communication switching value parameterIf the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used: by considering the intensity weight of the received signal of the base station communication, the signal-to-noise ratio weight, the network delay weight, the bandwidth utilization ratio weight, the satellite communication received signal intensity weight, the transmission delay weight and the satellite antenna weight, the advantages and disadvantages of the satellite communication and the base station communication can be fully combined for selective communication, the communication cost is greatly reduced, and the stability and the accuracy of the communication are improved.
3. The invention generates a simulated image through a depth generation countermeasure network modelCalculate the analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
in calculating the structural similarity index In the process, the proportion of the accident scene image damage area to the vehicle image is included in the calculation process, so that the accuracy of judging the vehicle damage degree is greatly enhanced, and an opinion suggestion of accuracy can be given to a vehicle insurance solution.
4. The application inputs the preprocessed video frame image of the accident scene of the vehicle into the trained depth generation countermeasure network model, and the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained depth generation countermeasure network model is trained by adopting a sample data set, and a simulation image is generated by the depth generation countermeasure network model, namely, an image before an accident is generated and an image after the accident is compared, so that the accuracy of the damage degree is greatly improved.
The foregoing has described in detail a method and system for insurance remote video exploration based on adaptive communication selection, and specific examples have been applied herein to illustrate the principles and embodiments of the present invention, the above examples being for the purpose of aiding in understanding the core concept of the present invention; also, as will be apparent to those skilled in the art in light of the present teachings, the present disclosure should not be limited to the specific embodiments and applications described herein.
Claims (10)
1. An insurance remote video exploration method based on self-adaptive communication selection is characterized by comprising the following steps:
s1: vehicle-mounted communication terminal real-time detection satellite communication received signal strength Transmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delay/>Bandwidth utilization/>；
S2: calculating communication handover value parametersIf the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
S3: the vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module, and preprocesses the video frame images of the vehicle accident scene;
S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image ; Calculating analog image/>With accident scene image/>Is of the structural similarity index of (2); According to structural similarity index/>Judging the severity of damage to the vehicle;
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The proportion of the image damage area of the accident scene to the vehicle image;
s5: the damage degree of the vehicle is displayed on an insurance company display screen.
2. The method for remote video exploration of insurance based on adaptive communication selection according to claim 1, wherein said S4: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
3. A method of insurance remote video investigation based on adaptive communication selection according to claim 1, wherein said pre-processing of vehicle accident scene video frame images comprises denoising with gaussian filtering.
4. The method for remote video exploration of insurance based on adaptive communication selection according to claim 1, wherein said S5: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone.
5. The method for remote video exploration of insurance based on adaptive communication selection according to claim 1, wherein said index of similarity based on structureJudging the severity of damage to the vehicle, when/>The severity of the vehicle damage is slight when/>The severity of the vehicle damage is significant when/>When the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
6. An insurance remote video survey system based on adaptive communication selection, comprising:
Real-time detection module of vehicle-mounted communication terminal for detecting satellite communication received signal strength in real time Transmission delay/>Satellite antenna azimuth/>And pitch/>; Simultaneously detecting the communication received signal strength/>, of a base station in real timeSignal to noise ratio/>Network delayBandwidth utilization/>；
Communication handover value parameterA calculation module for calculating the communication switching value parameter/>If the communication switching value parameter/>Greater than a set threshold/>Then base station communication is used, otherwise satellite communication is used:
Wherein, The method comprises the steps of respectively obtaining a base station communication received signal strength weight factor, a signal-to-noise ratio weight factor, a network delay weight factor, a bandwidth utilization weight factor, a satellite communication received signal strength weight factor, a transmission delay weight factor and a satellite antenna weight factor;
The vehicle-mounted communication terminal transmits the collected video frame images of the vehicle accident scene to the insurance company processing module to preprocess the video frame images of the vehicle accident scene;
Vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image ; Calculating analog image/>With accident scene image/>Structural similarity index/>; According to structural similarity index/>Judging the severity of damage to the vehicle;
Wherein, Is an analog image/>Average brightness of (2); /(I)Is accident scene image/>Average brightness of (2); /(I)、/>Analog images/>, respectivelyVariance, accident scene image/>Variance; /(I)For simulating images/>With accident scene image/>Is a covariance of (2); /(I)For the first set value,/>Is a second set value; /(I)The proportion of the image damage area of the accident scene to the vehicle image;
And a display module: the damage degree of the vehicle is displayed on an insurance company display screen.
7. The adaptive communication selection-based insurance remote video exploration system of claim 6, wherein said vehicle damage severity calculation module: inputting the preprocessed video frame image of the accident scene of the vehicle into a trained depth generation countermeasure network model, wherein the depth generation countermeasure network model generates a simulation image; The trained depth generation countermeasure network model comprises a generator and a discriminator; the generator is responsible for generating a simulation image of the vehicle before the accident, which corresponds to the input vehicle accident scene image; the discriminator is responsible for calculating the similarity between the generated analog image and the accident scene image; the trained deep-drawn countermeasure network model is trained by employing a sample dataset.
8. An adaptive communication selection based insurance remote video survey system according to claim 6, wherein said preprocessing of vehicle scene video frame images includes denoising with gaussian filtering.
9. The adaptive communication selection-based insurance remote video survey system of claim 6, wherein said display module: the damage degree of the vehicle is displayed on a display screen of an insurance company, the damage degree of the vehicle comprises slight, medium and great, and the display comprises the display on an LCD computer screen or the display on a mobile phone.
10. The adaptive communication selection-based insurance remote video survey system of claim 6, wherein said index of similarity according to structureJudging the severity of damage to the vehicle, when/>The severity of the vehicle damage is slight when/>The severity of the vehicle damage is significant when/>When the severity of the damage to the vehicle is moderate,/>For the first threshold of the damage degree,/>A second threshold for the degree of damage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410347596.6A CN117956120B (en) | 2024-03-26 | 2024-03-26 | Insurance remote video exploration method and system based on self-adaptive communication selection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410347596.6A CN117956120B (en) | 2024-03-26 | 2024-03-26 | Insurance remote video exploration method and system based on self-adaptive communication selection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117956120A CN117956120A (en) | 2024-04-30 |
| CN117956120B true CN117956120B (en) | 2024-06-04 |
Family
ID=90801824
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410347596.6A Active CN117956120B (en) | 2024-03-26 | 2024-03-26 | Insurance remote video exploration method and system based on self-adaptive communication selection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117956120B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886436A (en) * | 2017-11-08 | 2018-04-06 | 广东翼卡车联网服务有限公司 | Method, storage medium and the car-mounted terminal that a kind of car insurance is settled a claim automatically |
| CN114245421A (en) * | 2021-12-23 | 2022-03-25 | 军事科学院***工程研究院网络信息研究所 | Satellite communication network flow balancing method based on tilt factor |
| CN116980028A (en) * | 2023-08-29 | 2023-10-31 | 重庆邮电大学 | User terminal network switching access method based on star-ground fusion network |
-
2024
- 2024-03-26 CN CN202410347596.6A patent/CN117956120B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886436A (en) * | 2017-11-08 | 2018-04-06 | 广东翼卡车联网服务有限公司 | Method, storage medium and the car-mounted terminal that a kind of car insurance is settled a claim automatically |
| CN114245421A (en) * | 2021-12-23 | 2022-03-25 | 军事科学院***工程研究院网络信息研究所 | Satellite communication network flow balancing method based on tilt factor |
| CN116980028A (en) * | 2023-08-29 | 2023-10-31 | 重庆邮电大学 | User terminal network switching access method based on star-ground fusion network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117956120A (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108109385B (en) | System and method for identifying and judging dangerous behaviors of power transmission line anti-external damage vehicle | |
| CN110728236A (en) | Vehicle loss assessment method and special equipment thereof | |
| CN112749654A (en) | Deep neural network model construction method, system and device for video fog monitoring | |
| CN111046797A (en) | Oil pipeline warning method based on personnel and vehicle behavior analysis | |
| CN111667655A (en) | Infrared image-based high-speed railway safety area intrusion alarm device and method | |
| CN106339657A (en) | Straw incineration monitoring method and device based on monitoring video | |
| US11748664B1 (en) | Systems for creating training data for determining vehicle following distance | |
| CN112084928A (en) | Road traffic accident detection method based on visual attention mechanism and ConvLSTM network | |
| CN112287861A (en) | Road information enhancement and driving early warning method based on night environment perception | |
| CN112637550A (en) | PTZ moving target tracking method for multi-path 4K quasi-real-time spliced video | |
| CN117237597A (en) | Data processing terminal based on Beidou satellite data and AI graph fusion | |
| CN117956120B (en) | Insurance remote video exploration method and system based on self-adaptive communication selection | |
| WO2022104798A1 (en) | 5g-based unmanned electronic traffic police duty system | |
| CN112229586B (en) | Bridge support health state detection method and terminal equipment | |
| CN105631425A (en) | License plate recognition method based on video streams and system thereof and intelligent digital video camera | |
| CN112258402A (en) | Dense residual generation countermeasure network capable of rapidly removing rain | |
| CN115760839A (en) | Mobile NetV3 internet of vehicles detection system with roadside unit caching function | |
| CN207115438U (en) | Image processing apparatus for vehicle-mounted fisheye camera | |
| CN115527118A (en) | Remote sensing image target detection method fused with attention mechanism | |
| CN117935127B (en) | Intelligent damage assessment method and system for panoramic video exploration | |
| CN112836619A (en) | Embedded vehicle-mounted far infrared pedestrian detection method, system, equipment and storage medium | |
| Wu et al. | Deep learning based nighttime target enhancement detection algorithm for intelligent vehicles | |
| CN115861825B (en) | 2C detection method based on image recognition | |
| CN118135338B (en) | Vehicle risk assessment method and system based on automatic detection of external damage of vehicle | |
| CN212873683U (en) | High-speed railway safety zone intrusion alarm device based on infrared image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |