Automobile transportation logistics data tracing method and system for cross-border trade

Abstract

The invention relates to the technical field of block chains and automobile transportation, and provides an automobile transportation logistics data tracing method and system for cross-border trade, and the method comprises the steps: carrying out the real-time monitoring of a target vehicle in a domestic section, an international section and a domestic section, and carrying out the tracking of the target vehicle according to a monitoring image, detecting whether the appearance of the target vehicle is damaged or not through a damage type identification model based on a neural network; when it is detected that the appearance of the target vehicle is damaged, a block is generated based on damage information; and the corresponding blocks are linked to corresponding alliance chains, including a domestic segment alliance chain, an international segment alliance chain and a foreign segment alliance chain. According to the invention, an effective automobile appearance damage condition tracing method is established, and powerful support is provided for the judgment of compensation responsibility.

Description

Technical Field

[0001] The present application relates to the fields of blockchain and automobile transportation technology, and in particular to a method and system for tracing automobile transportation logistics data for cross-border trade. Background Art

[0002] In the field of cross-border automobile trade, logistics and transportation links are usually completed by multiple different types of logistics companies. Domestic transportation is the responsibility of domestic logistics companies; international transportation is undertaken by international logistics companies; and transportation and distribution in the destination country, that is, the foreign section, is operated by local foreign logistics companies. When the foreign car buyer receives the car, he will check the appearance of the target vehicle to determine whether the appearance of the car is damaged during transportation. Once the car buyer finds that the appearance of the car is damaged, he usually asks for compensation from the logistics company responsible for the foreign section transportation. In this case, the foreign section logistics company often has to bear the compensation liability because it is difficult to provide sufficient evidence to prove that the damage is not caused by its own responsibility. But in fact, there are many possibilities for the source of damage to the appearance of the car. It may not be caused by a foreign transportation company. It may be caused by a domestic logistics company during domestic transportation, or it may be caused by an international logistics company during the international transportation link, or even it may be caused by the joint action of multiple logistics and transportation companies at the stage for which they are responsible.

[0003] Therefore, how to establish an effective method to trace the damage to the exterior of an automobile in order to provide strong support for the determination of compensation liability has become a technical issue that needs to be urgently addressed in the current field of cross-border automobile trade logistics. Summary of the invention

[0004] In response to the above technical problems, the purpose of this application is to provide a method and system for tracing automobile transportation logistics data for cross-border trade, aiming to establish an effective method for tracing the source of automobile appearance damage in order to provide strong support for the determination of compensation liability.

[0005] In a first aspect, an embodiment of the present application provides a method for tracing automobile transportation logistics data for cross-border trade, the method comprising:

[0006] The target vehicle in the domestic transportation phase is monitored in real time by the first monitoring system, and whether the appearance of the target vehicle is damaged is detected by a damage type recognition model based on a neural network according to the monitoring screen image of the first monitoring system; when it is detected that the appearance of the target vehicle is damaged, a first block is generated based on the first damage information; and the first block is linked to the domestic segment alliance chain;

[0007] The target vehicle in the international transportation stage is monitored in real time through the second monitoring system. According to the monitoring screen images of the second monitoring system, whether the appearance of the target vehicle is damaged is detected through a damage type recognition model based on a neural network; when it is detected that the appearance of the target vehicle is damaged, a second block is generated based on the second damage information; the second block is uploaded to the international segment consortium chain;

[0008] The target vehicle in the foreign transportation stage is monitored in real time through the third monitoring system. According to the monitoring screen images of the third monitoring system, whether the appearance of the target vehicle is damaged is detected through a damage type recognition model based on a neural network; when it is detected that the appearance of the target vehicle is damaged, a third block is generated based on the third damage information; the third block is uploaded to the foreign segment consortium chain; wherein, the first damage information, the second damage information and the third damage information all include time, location, target vehicle information, details of damage, monitoring screen images corresponding to the damaged time period, and information of the transportation vehicle carrying the target vehicle; information interaction is realized among the domestic segment consortium chain, the international segment consortium chain and the foreign segment consortium chain based on a cross-chain protocol; the second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

[0009] Further, the step of detecting whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network according to the monitoring screen images includes:

[0010] Obtain two adjacent frames of monitoring screen images, and use the previous frame of monitoring screen image and the next frame of monitoring screen image as the first monitoring screen image and the second monitoring screen image respectively;

[0011] Locate the target vehicle from the first monitoring screen image to obtain a first target vehicle image;

[0012] Locate the target vehicle from the second monitoring screen image to obtain a second target vehicle image;

[0013] Compare the first target vehicle image and the second target vehicle image to obtain a difference region image;

[0014] Based on the difference region image, determine whether the target vehicle is damaged and determine the damage type through a damage type recognition model based on a neural network; wherein, the damage types include: scratches, dents and component breakage;

[0015] When it is determined that the target vehicle is damaged, input the difference region image into a part recognition model based on a neural network to identify the damaged part;

[0016] If it cannot be recognized, expand the difference region until the part recognition model recognizes the damaged part.

[0017] Further, the damage type recognition model is trained by the following method:

[0018] Label the collected damaged vehicle images, marking the actual damaged areas and damage types; among them, the damage types include: scratches, dents, and component breakage;

[0019] Label the other vehicle images that may cause false alarms of damage, marking the areas and objects that cause false alarms of damage; the objects that cause false alarms of vehicle damage include: dirt, water stains, and environmental reflections;

[0020] Use a preset neural network model to perform deep learning training on the labeled images to obtain a damage type recognition model based on the neural network.

[0021] Further, the cameras of the first monitoring system, the second monitoring system, and the third monitoring system all include movable anti-shake lenses, gyroscopes, and image sensors. Before the step of obtaining adjacent two-frame monitoring screen images, the method further includes:

[0022] Obtain the angular change amplitude value compared to the previous moment through the gyroscope;

[0023] Determine the moving direction and moving distance of the anti-shake lens according to the corresponding relationship between the angular change amplitude value and the anti-shake lens compensation vector under a preset focal length;

[0024] Adjust the position of the anti-shake lens according to the moving direction and moving distance of the anti-shake lens, so that the image sensor captures an anti-shake image, obtaining an anti-shake image;

[0025] Use the anti-shake image as the blurred input image during application and input it into a pre-trained first deblurring neural network to predict the reconstructed blurred image through the first task of the first deblurring neural network;

[0026] Update the meta-training weights of the first deblurring neural network based on the anti-shake image and the reconstructed blurred image until the training is completed, obtaining a second deblurring neural network;

[0027] Input the anti-shake image into the second deblurring neural network to generate a deblurred monitoring image.

[0028] Further, after detecting that the appearance of the target vehicle is damaged and generating a block based on the damage information, the method further includes:

[0029] Obtain the vibration data collected by the vibration sensor during the damaged period of the target vehicle; among them, the vibration sensor is installed on the transportation vehicle that transports the target vehicle;

[0030] Analyze whether the vehicle is jolted based on the vibration data and vibration curve, and whether the degree of jolting is greater than a preset threshold;

[0031] If so, determine the current location of the vehicle through a positioning module installed on the vehicle;

[0032] Obtain other vehicles that will pass through the location from the logistics tracking system;

[0033] Send a prompt message indicating the location is jolted and a prompt message indicating that the appearance of the consigned target vehicle has been damaged at the location to other vehicles that will pass through the location; wherein, the prompt message indicating the location is jolted includes distance information from the location and voice prompt information about the jolting.

[0034] Further, after generating a block based on the damage information when detecting that the appearance of the target vehicle is damaged, the method further includes:

[0035] Extract the target vehicle information, the details of the damage, the surveillance video image corresponding to the damaged time period, and the vehicle information carrying the target vehicle from the damage information; wherein, the details of the damage include the type of damage and the location of damage;

[0036] Extract the damaged image of the target vehicle from the surveillance video image and prominently mark the damaged location in the damaged image of the target vehicle to obtain a damaged marked image of the target vehicle;

[0037] Send the target vehicle information, the details of the damage, the vehicle information carrying the target vehicle, and the damaged marked image of the target vehicle to the management system of the logistics company responsible for the next transportation stage;

[0038] The management system forwards the target vehicle information, the details of the damage, the vehicle information carrying the target vehicle, and the damaged marked image of the target vehicle to the corresponding docking person terminal.

[0039] In a second aspect, an automobile transportation logistics data traceability system for cross-border trade provided by an embodiment of the present application includes:

[0040] A domestic section module, configured to perform real-time monitoring on a target vehicle in the domestic transportation stage through a first monitoring system, detect whether the appearance of the target vehicle is damaged according to the surveillance video image of the first monitoring system, and generate a first block based on the first damage information when detecting that the appearance of the target vehicle is damaged; and upload the first block to the domestic section consortium chain;

[0041] The international section module is used to monitor the target vehicle in real time during the international transportation stage through the second monitoring system. According to the monitoring screen images of the second monitoring system, it detects whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network. When it detects that the appearance of the target vehicle is damaged, it generates a second block based on the second damage information and uploads the second block to the international section consortium chain.

[0042] The foreign section module is used to monitor the target vehicle in real time during the foreign transportation stage through the third monitoring system. According to the monitoring screen images of the third monitoring system, it detects whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network. When it detects that the appearance of the target vehicle is damaged, it generates a third block based on the third damage information and uploads the third block to the foreign section consortium chain. Among them, the first damage information, the second damage information, and the third damage information all include time, location, target vehicle information, details of damage, monitoring screen images corresponding to the damaged time period, and information about the transportation vehicle carrying the target vehicle. Information interaction is realized among the domestic section consortium chain, the international section consortium chain, and the foreign section consortium chain based on a cross-chain protocol. The second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

[0043] Further, detecting whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network according to the monitoring screen images includes:

[0044] Obtain two adjacent frames of monitoring screen images, and use the previous frame of monitoring screen image and the next frame of monitoring screen image as the first monitoring screen image and the second monitoring screen image respectively.

[0045] Locate the target vehicle in the first monitoring screen image to obtain the first target vehicle image.

[0046] Locate the target vehicle in the second monitoring screen image to obtain the second target vehicle image.

[0047] Compare the first target vehicle image and the second target vehicle image to obtain a difference region image.

[0048] Based on the difference region image, determine whether the target vehicle is damaged and determine the damage type through a damage type recognition model based on a neural network. Among them, the damage types include scratches, dents, and component breakage.

[0049] When it is determined that the target vehicle is damaged, input the difference region image into a part recognition model based on a neural network to identify the damaged part.

[0050] If it cannot be recognized, expand the difference area until the damaged part is recognized by the part recognition model.

[0051] Further, the damage type recognition model is trained by the following method:

[0052] Label the collected images of damaged vehicles, and mark the true damaged areas and damage types; among them, the damage types include: scratches, dents, and component breakage;

[0053] Label the other vehicle images that will cause false alarms of damage, and mark the areas and objects that cause false alarms of damage; the objects that cause false alarms of vehicle damage include; dirt, water stains, and environmental reflections;

[0054] Use a preset neural network model to perform deep learning training on the labeled images to obtain a damage type recognition model based on the neural network.

[0055] Further, the system further includes:

[0056] The first acquisition module is used to acquire the vibration data collected by the vibration sensor during the damaged period of the target vehicle; wherein, the vibration sensor is installed on the transportation vehicle transporting the target vehicle;

[0057] The analysis module is used to analyze whether the transportation vehicle is jolted and the degree of jolting is greater than a preset threshold based on the vibration data and the vibration curve;

[0058] The positioning module is used to, if so, determine the current position of the transportation vehicle through the positioning module installed on the transportation vehicle;

[0059] The second acquisition module is used to acquire other transportation vehicles that will pass by the position from the logistics tracking system;

[0060] The sending module is used to send a prompt message about the jolting of the position and a prompt message about the appearance damage of the consigned target vehicle at the position to other transportation vehicles that will pass by the position; wherein, the prompt message about the jolting of the position includes distance information from the position and a voice prompt message about the jolting.

[0061] In the embodiments of the present application, multiple monitoring systems are used to monitor the target vehicle in real time during the domestic transportation, international transportation, and foreign transportation stages, enabling the timely detection of damage to the vehicle's appearance. The damage type recognition model based on neural network ensures the accuracy of vehicle appearance damage detection and can effectively identify various types of damage. The generated damage information includes key contents such as time, location, target vehicle information, details of the damage, and information on the means of transportation, ensuring the diversity of damage data and facilitating subsequent traceability and analysis. The damage information at different stages is separately uploaded to the corresponding domestic section consortium chain, international section consortium chain, and foreign section consortium chain. Utilizing the immutable characteristic of the blockchain, the authenticity and credibility of the data are guaranteed. Information interaction is realized among the consortium chains based on cross-chain protocols, enabling the integration and sharing of vehicle damage data throughout the cross-border trade transportation process. Relevant parties, whether domestic or foreign, can conveniently obtain the damage situation of the vehicle at different transportation stages, providing strong support for liability determination, claims settlement, etc. In addition, the consortium chain can conduct strict identity authentication and permission management for participating nodes. Only authorized relevant parties such as transportation enterprises can join the consortium chain and participate in data recording and sharing, meeting the requirements of cross-border trade automotive transportation logistics business, ensuring that data only circulates among specific trusted entities, and protecting business secrets and sensitive information. Cross-border trade involves laws and regulations in different countries and regions. The consortium chain can formulate unified rules and standards within the consortium according to the legal requirements of different countries and regions to ensure that the storage, use, and sharing of data comply with local laws and regulations, reducing legal risks. The cross-border trade automotive transportation logistics data is large in volume and requires rapid processing. The number of nodes in the consortium chain is relatively small and relatively fixed, and it is relatively easy to reach a consensus mechanism. Therefore, it can complete data verification and recording in a relatively short time, improving the processing speed. Description of the Drawings

[0062] To more clearly illustrate the technical solutions of the present application, the drawings required for the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0063] Figure 1 It is a flowchart of the method for tracing automotive transportation logistics data for cross-border trade provided by the embodiments of the present application;

[0064] Figure 2 It is a schematic structural diagram of the system for tracing automotive transportation logistics data for cross-border trade provided by another embodiment of the present application. Detailed Embodiments

[0065] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further details this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely used to explain this application and are not used to limit this application.

[0066] Those skilled in the art of this technology can understand that, unless specifically stated otherwise, the singular forms "a", "an", "above-mentioned" and "the" used herein may also include the plural forms. It should be further understood that the term "including" used in the specification of this application means that there are features, integers, steps, operations, elements, modules and / or components, but does not exclude the existence or addition of one or more other features, integers, steps, operations, elements, modules, components and / or their groups. It should be understood that when we say an element is "connected" or "coupled" to another element, it can be directly connected or coupled to other elements, or there may also be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any module and all combinations of one or more related listed items.

[0067] Those skilled in the art of this technology can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those of ordinary skill in the field to which this application belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have a meaning consistent with the meaning in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as here.

[0068] Please refer to Figure 1 , an embodiment of this application provides a method for tracing automotive transportation logistics data for cross-border trade, including:

[0069] S1. Real-time monitor the target vehicle during the domestic transportation stage through the first monitoring system. According to the monitoring screen image of the first monitoring system, detect whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network; when it is detected that the appearance of the target vehicle is damaged, generate a first block based on the first damage information; upload the first block to the domestic section consortium blockchain;

[0070] S2. Real-time monitor the target vehicle during the international transportation stage through the second monitoring system. According to the monitoring screen image of the second monitoring system, detect whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network; when it is detected that the appearance of the target vehicle is damaged, generate a second block based on the second damage information; upload the second block to the international section consortium blockchain;

[0071] S3. The target vehicle during the foreign transportation stage is monitored in real time through the third monitoring system. Based on the monitoring screen images of the third monitoring system, whether the appearance of the target vehicle is damaged is detected through a damage type recognition model based on a neural network. When it is detected that the appearance of the target vehicle is damaged, a third block is generated based on the third damage information. The third block is chained to the foreign section consortium chain. Among them, the first damage information, the second damage information, and the third damage information all include time, location, target vehicle information, details of the damage, monitoring screen images corresponding to the damaged time period, and transportation vehicle information carrying the target vehicle. Information interaction is realized among the domestic section consortium chain, the international section consortium chain, and the foreign section consortium chain based on a cross-chain protocol. The second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

[0072] In the embodiment of the present application, the first monitoring system, the second monitoring system, and the third monitoring system are installed on the transportation vehicle corresponding to the target vehicle. The transportation vehicle corresponding to the first monitoring system can be a freight truck, a train, a ro-ro ship, or a passenger ro-ro ship. The transportation vehicle corresponding to the second monitoring system can be a freight truck, a train, a ro-ro ship, or a passenger ro-ro ship. The transportation vehicle corresponding to the third monitoring system can be a freight truck, a train, a ro-ro ship, or a passenger ro-ro ship. The time is the time when it is determined that the target vehicle is damaged. The location is the location information of the transportation vehicle carrying the target vehicle when the target vehicle is damaged. The target vehicle information includes one or both of the unique identification code of the target vehicle and the position information of the target vehicle fixed on the transportation vehicle. In the case where multiple cameras correspond to one target vehicle and one camera only targets one target vehicle, the position information of the target vehicle fixed on the transportation vehicle can be determined based on the relationship between the pre-set unique identification code of the camera and the position information of the target vehicle fixed on the transportation vehicle. The transportation vehicle information carrying the target vehicle includes the license plate information of the transportation vehicle. The details of the damage include the type of damage and the location of the damage. It should be understood that the first damage information, the second damage information, and the third damage information, where the first, second, and third are used to distinguish the damage information generated in different stages (domestic, international, foreign).

[0073] In the embodiments of the present application, multiple monitoring systems are used to monitor the target vehicle in real time during the domestic transportation, international transportation, and foreign transportation stages respectively, and the damage condition of the vehicle appearance can be detected in a timely manner. The damage type recognition model based on the neural network ensures the accuracy of the vehicle appearance damage detection and can effectively identify various types of damage. The generated damage information includes key contents such as time, location, target vehicle information, damage details, and carrier vehicle information, etc., ensuring the diversity of damage data and facilitating subsequent traceability and analysis. The damage information in different stages is respectively uploaded to the corresponding domestic segment consortium chain, international segment consortium chain, and foreign segment consortium chain. By using the immutable characteristic of the blockchain, the authenticity and credibility of the data are guaranteed. Information interaction is realized among the consortium chains based on the cross-chain protocol, enabling the integration and sharing of vehicle damage data during the entire cross-border trade transportation process. Relevant parties, whether domestic or foreign, can conveniently obtain the damage conditions of the vehicle in different transportation stages, providing strong support for liability determination, claims settlement, etc. In addition, the consortium chain can conduct strict identity authentication and permission management for participating nodes. Only authorized relevant parties such as transportation enterprises can join the consortium chain and participate in the recording and sharing of data, meeting the requirements of the cross-border trade automobile transportation logistics business, and ensuring that the data can only circulate among specific trusted entities, protecting business secrets and sensitive information. Cross-border trade involves laws and regulations in different countries and regions. The consortium chain can formulate unified rules and standards within the consortium according to the legal requirements of different countries and regions to ensure that the storage, use, and sharing of data comply with local laws and regulations, reducing legal risks. The cross-border trade automobile transportation logistics has a large amount of data and requires fast processing. The number of nodes in the consortium chain is relatively small and relatively fixed, and it is relatively easy to reach a consensus mechanism. Therefore, it can complete the verification and recording of data in a relatively short time, improving the processing speed.

[0074] In one embodiment, the step of detecting whether the appearance of the target vehicle is damaged according to the monitoring screen image through the damage type recognition model based on the neural network includes:

[0075] Obtain two adjacent frames of monitoring screen images, and use the previous frame of monitoring screen image and the next frame of monitoring screen image as the first monitoring screen image and the second monitoring screen image respectively;

[0076] Locate the target vehicle in the first monitoring screen image to obtain a first target vehicle image;

[0077] Locate the target vehicle in the second monitoring screen image to obtain a second target vehicle image;

[0078] Compare the first target vehicle image and the second target vehicle image to obtain a difference region image;

[0079] Based on the differential region image, determine whether the target vehicle is damaged and identify the damage type through a neural network-based damage type recognition model; wherein, the damage types include: scratches, dents, and component breakage;

[0080] When it is determined that the target vehicle is damaged, input the differential region image into a neural network-based part recognition model to identify the damaged part;

[0081] If the damaged part cannot be identified, expand the differential region until the part recognition model identifies the damaged part.

[0082] Embodiments of the present application are applicable to the first monitoring system, the second monitoring system, and the third monitoring system. That is, according to the monitoring screen image of the first monitoring system, it can be realized by using embodiments of the present application to detect whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network. The same applies to the second and third monitoring systems. The first monitoring system includes multiple cameras. By arranging and installing the cameras, comprehensive monitoring of the appearance of the target vehicle is achieved. Each image captured by each camera of the first monitoring system is considered a first monitoring screen image. Since the first monitoring screen does not include the entire vehicle but only a certain part of the vehicle, when locating the target vehicle from the first monitoring screen image, as long as a certain part of the vehicle is detected, it is considered that the target vehicle is located. The area belonging to the vehicle in the first monitoring screen image can be recognized through a target recognition model, such as the YOLO model. Specifically, the YOLOv8 model can be used. The YOLOv8 model has high frame rate processing capabilities and is suitable for monitoring video analysis. By training the YOLOv8 model with a large number of images labeled with a certain part of the vehicle, the area belonging to the vehicle in the first monitoring screen image can be recognized. Or the area belonging to the vehicle in the first monitoring screen can be recognized based on the color features, texture features, and shape features of the vehicle. In addition, by combining the YOLOv8 model with a TensorRT acceleration engine, the inference speed and efficiency can be improved. Comparing the first target vehicle image and the second target vehicle image to obtain a difference region image. Specifically, the difference image can be recognized from the second target vehicle image through the pixel difference between the first target vehicle image and the second target vehicle image, and then the difference region image is input into a damage type recognition model based on a neural network. This damage type recognition model will automatically identify and determine whether the target vehicle is damaged. If it is damaged, the damage type will be output. The damage types include scratches, dents, and component breakages. Component breakages include window breakages, headlight breakages, bumper breakages, etc. When it is determined that the target vehicle is damaged, the difference region image is input into a part recognition model based on a neural network to identify the damaged part. This model can automatically identify the damaged part, including windows, headlights, the body, the bumper, etc. If it cannot be recognized, the difference region is enlarged. By enlarging the difference region, image information related to the damaged part can be obtained, which helps the model to more accurately identify the damaged part. Enlarging the difference region here means that based on the original difference region, a larger image area is intercepted as the enlarged difference region image. The enlarged difference region is input into the part recognition model. If it still cannot be recognized, the difference region is continued to be enlarged until the damaged part is recognized.

[0083] In the embodiments of the present application, by locating the target vehicle and then processing and analyzing the located target vehicle image, the amount of calculation and the interference of the background can be reduced. By comparing the first target vehicle image and the second target vehicle image, the area where the target vehicle may be damaged can be quickly obtained. By using an artificial intelligence model to determine whether the target vehicle is damaged and identify the damaged parts, a foundation is laid for subsequent tracing of vehicle damage. In addition, during the process of identifying the damaged parts, if the initially identified difference area does not accurately contain the damaged parts, the solution can automatically expand the difference area and re-identify the parts. This flexible adjustment mechanism can reduce the amount of calculation and ensure the successful identification of the damaged parts.

[0084] In one embodiment, the damage type recognition model is trained by the following method:

[0085] Label the collected damaged vehicle images, and mark the actual damaged areas and damage types; wherein, the damage types include: scratches, dents, and component breakage;

[0086] Label the other vehicle images that may cause false alarms of damage, and mark the areas and objects that cause false alarms of damage; the objects that cause false alarms of vehicle damage include; dirt, water stains, and environmental reflections;

[0087] Use a preset neural network model to perform in-depth learning training on the labeled images to obtain a damage type recognition model based on the neural network.

[0088] In the embodiments of the present application, in order to solve the problem that the target vehicle may be stained with dirt, water stains or misidentified as damaged due to environmental reflections during transportation, the embodiments of the present application label these vehicle images that cause false alarms, mark the areas and objects that cause false alarms of damage, and add them to the training data to train the model, so that the model can distinguish the situations that are prone to false alarms such as dirt, water stains, and environmental reflections, greatly improving the accuracy of vehicle damage recognition.

[0089] In one embodiment, the part recognition model is trained based on the vehicle image and the corresponding part label. The vehicle image may only include an image of a part of the vehicle.

[0090] In one embodiment, in the cameras of the first monitoring system, the second monitoring system, and the third monitoring system all include movable anti-shake lenses, gyroscopes, and image sensors, before the step of acquiring adjacent two-frame monitoring screen images, the method further includes:

[0091] Obtain the angular change amplitude value compared to the previous moment through the gyroscope;

[0092] Determine the moving direction and moving distance of the anti-shake lens according to the corresponding relationship between the angle change amplitude value and the anti-shake lens compensation vector under the preset focal length;

[0093] Adjust the position of the anti-shake lens according to the moving direction and moving distance of the anti-shake lens, so that the image sensor captures an image after anti-shake to obtain an anti-shake image;

[0094] Use the anti-shake image as a blurred input image during application and input it into a pre-trained first deblurring neural network to predict a reconstructed blurred image through the first task of the first deblurring neural network;

[0095] Update the meta-training weights of the first deblurring neural network based on the anti-shake image and the reconstructed blurred image until the training is completed to obtain a second deblurring neural network;

[0096] Input the anti-shake image into the second deblurring neural network, and generate a deblurred monitoring image through the second task of the second deblurring neural network.

[0097] In the embodiments of the present application, in order to solve the problem of image blurring caused by the vibration of the camera and the vibration of the target vehicle during the transportation of the target vehicle, the present invention adopts optical anti-shake and electronic anti-shake technologies to improve the image quality. For the electronic anti-shake technology, in view of the problem of dynamic blurring that occurs during the transportation of the target vehicle, the present invention selects a neural network-based self-adaptive deblurring method. Specifically, the first deblurring neural network is trained through meta-training to perform the first task and the second task. The first task is to reproduce the blurred input image, and the second task is to generate a clean image based on the blurred input image. Updating the meta-training weights of the first deblurring neural network based on the anti-shake image and the reconstructed blurred image specifically means updating the training weights of the first deblurring neural network based on the loss function of the first task, the anti-shake image, and the reconstructed blurred image. The neural network-based self-adaptive deblurring method can improve the image quality of the images captured by the target vehicle in the scenario of the present invention.

[0098] In one embodiment, after detecting that the appearance of the target vehicle is damaged and generating a block based on the damage information, the method further includes:

[0099] Obtain the vibration data collected by the vibration sensor during the damaged period of the target vehicle; wherein, the vibration sensor is installed on the vehicle transporting the target vehicle;

[0100] Analyze whether the vehicle is jolted and the degree of jolting is greater than a preset threshold based on the vibration data and the vibration curve;

[0101] If so, determine the current location of the vehicle through the positioning module installed on the vehicle;

[0102] Obtain other vehicles that will pass through the location from the logistics tracking system;

[0103] Send a prompt message about the location being bumpy and a prompt message indicating that the appearance of the consigned target vehicle has been damaged at the location to other vehicles that will pass through the location; wherein, the prompt message about the location being bumpy includes distance information from the location and voice prompt information about the bump.

[0104] In the embodiment of the present application, the vibration curve is directly drawn based on the collected vibration data (time and corresponding vibration values), and it shows the change trend of the vibration value over time. Analyze whether the vehicle is bumped and the degree of bump is greater than a preset threshold based on the vibration data and the vibration curve. Specifically, the vibration data is the vibration data collected during the damaged period of the target vehicle, and the vibration data includes time and vibration values. Compare the vibration value of the damaged time period of the target vehicle with the vibration value of its adjacent time period (the vibration value of the adjacent time period of the damaged time period of the target vehicle is determined from the vibration curve), subtract the vibration value of its adjacent time period (specifically, the average vibration value) from the vibration value of the damaged time period of the target vehicle (specifically, the average vibration value), calculate the difference between the two. If the difference is greater than the preset vibration threshold, it is determined that the vehicle is bumped and the degree of bump is greater than the preset threshold. When it is determined that the degree of bump is greater than the preset threshold, determine the current location of the vehicle through the positioning module installed on the vehicle, such as a GPS module or a Beidou positioning module. The current location information includes road section information. Obtain other vehicles that will pass through the location from the logistics tracking system. Specifically, the logistics tracking system will record the target end position information of each vehicle used for transporting vehicles, and record the real-time position information of each vehicle used for transporting vehicles; analyze its forward direction based on the position information of the vehicle changing over time; determine whether the vehicle passes through the location based on the target end position information of the vehicle used for transporting vehicles, the real-time position information of the vehicle, and the forward direction of the vehicle. Through such a method, the logistics tracking system can obtain other vehicles that will pass through the location. By sending the voice prompt message about the location being bumpy and the prompt message indicating that the appearance of the consigned target vehicle has been damaged at the location to other vehicles that will pass through the location, the problem of damage to the transported vehicle due to being bumped can be reduced.

[0105] In one embodiment, after detecting that the appearance of the target vehicle is damaged and generating a block based on the damage information, the method further includes:

[0106] Extract the target vehicle information, the details of the damage, the surveillance video images corresponding to the damaged time period, and the information of the vehicle carrying the target vehicle from the damage information; wherein, the details of the damage include the type of damage and the location of the damage.

[0107] Extract the damaged image of the target vehicle from the surveillance video images and visibly mark the damaged location in the damaged image of the target vehicle to obtain the damaged marked image of the target vehicle.

[0108] Send the target vehicle information, the details of the damage, the information of the vehicle carrying the target vehicle, and the damaged marked image of the target vehicle to the management system of the logistics company responsible for the next transportation stage.

[0109] The management system forwards the target vehicle information, the details of the damage, the information of the vehicle carrying the target vehicle, and the damaged marked image of the target vehicle to the corresponding docking person's terminal.

[0110] In the embodiment of the present application, visibly marking the damaged location in the damaged image of the target vehicle is specifically achieved by drawing a border or adding a text annotation to the damaged area, so as to obtain the damaged marked image of the target vehicle. By forwarding the target vehicle information, the details of the damage, the information of the vehicle carrying the target vehicle, and the damaged marked image of the target vehicle to the corresponding docking person's terminal and sending a reminder message to the docking person's terminal, it is possible to timely remind and facilitate the docking person to check and verify the target vehicle determined to be damaged by artificial intelligence, which is beneficial to the division of responsibilities. Further, the docking person's terminal can upload the inspection and verification results, and both the damage type recognition model based on the neural network and the part recognition model based on the neural network can be optimized and trained respectively based on the inspection and verification results to improve the recognition accuracy.

[0111] As Figure 2 shown, the embodiment of the present application further provides a vehicle transportation logistics data traceability system for cross-border trade, and the system includes:

[0112] The domestic section module 1 is used to monitor the target vehicle in the domestic transportation stage in real time through the first monitoring system, and detect whether the appearance of the target vehicle is damaged according to the surveillance video images of the first monitoring system; when it is detected that the appearance of the target vehicle is damaged, generate a first block based on the first damage information; and upload the first block to the domestic section alliance chain.

[0113] The international segment module 2 is used to monitor the target vehicle in real time during the international transportation stage through the second monitoring system. According to the monitoring screen images of the second monitoring system, it detects whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network; when it detects that the appearance of the target vehicle is damaged, it generates a second block based on the second damage information; and uploads the second block to the international segment consortium chain;

[0114] The foreign segment module 3 is used to monitor the target vehicle in real time during the foreign transportation stage through the third monitoring system. According to the monitoring screen images of the third monitoring system, it detects whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network; when it detects that the appearance of the target vehicle is damaged, it generates a third block based on the third damage information; and uploads the third block to the foreign segment consortium chain; wherein, the first damage information, the second damage information, and the third damage information all include time, location, target vehicle information, details of damage, monitoring screen images corresponding to the damaged time period, and information about the means of transportation carrying the target vehicle; information interaction is realized among the domestic segment consortium chain, the international segment consortium chain, and the foreign segment consortium chain based on a cross-chain protocol; the second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

[0115] In one embodiment, detecting whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network according to the monitoring screen images includes:

[0116] Obtain two adjacent frames of monitoring screen images, and use the previous frame of monitoring screen image and the next frame of monitoring screen image as the first monitoring screen image and the second monitoring screen image respectively;

[0117] Locate the target vehicle in the first monitoring screen image to obtain a first target vehicle image;

[0118] Locate the target vehicle in the second monitoring screen image to obtain a second target vehicle image;

[0119] Compare the first target vehicle image and the second target vehicle image to obtain a difference region image;

[0120] Based on the difference region image, determine whether the target vehicle is damaged and determine the damage type through a damage type recognition model based on a neural network; wherein, the damage types include: scratches, dents, and component breakage;

[0121] When it is determined that the target vehicle is damaged, input the difference region image into a part recognition model based on a neural network to identify the damaged part;

[0122] If it cannot be recognized, expand the difference area until the damaged part is recognized by the part recognition model.

[0123] In one embodiment, the damage type recognition model is trained by the following method:

[0124] Label the collected damaged vehicle images, and mark the true damaged area and the damage type; wherein, the damage type includes: scratches, dents and component breakage;

[0125] Label the other vehicle images that may cause false alarms of damage, and mark the areas and objects that cause false alarms of damage; the objects that cause false alarms of vehicle damage include; dirt, water stains and environmental reflections;

[0126] Use a preset neural network model to perform deep learning training on the labeled images to obtain a damage type recognition model based on the neural network.

[0127] In one embodiment, the system further includes:

[0128] The first acquisition module is used to acquire the vibration data collected by the vibration sensor during the damaged period of the target vehicle; wherein, the vibration sensor is installed on the vehicle transporting the target vehicle;

[0129] The analysis module is used to analyze whether the vehicle is jolted and the degree of jolting is greater than a preset threshold based on the vibration data and the vibration curve;

[0130] The positioning module is used to, if so, determine the current position of the vehicle by the positioning module installed on the vehicle;

[0131] The second acquisition module is used to acquire other vehicles that will pass through the position from the logistics tracking system;

[0132] The sending module is used to send a prompt message of the position jolting and a prompt message of the appearance damage of the consigned target vehicle at the position to other vehicles that will pass through the position; wherein, the prompt message of the position jolting includes the distance information from the position and the voice prompt information about the jolting.

[0133] It should be understood that the above-mentioned methods for tracing automobile transportation logistics data for cross-border trade are all applicable to the system for tracing automobile transportation logistics data for cross-border trade of the present invention. Therefore, the embodiments of the present invention will not be elaborated here too much.

[0134] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided in this application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0135] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, system, article, or method including a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, system, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, system, article, or method including that element.

[0136] The above are only the preferred embodiments of this application, and do not limit the patent scope of this application accordingly. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of this application, or directly or indirectly applied in other related technical fields, shall be included in the patent protection scope of this application by the same token.

Claims

1. A method for tracing automobile transportation logistics data for cross-border trade, characterized in that: The method comprises: The target vehicle in the domestic transportation phase is monitored in real time by the first monitoring system, and whether the appearance of the target vehicle is damaged is detected by a damage type recognition model based on a neural network according to the monitoring screen image of the first monitoring system; when it is detected that the appearance of the target vehicle is damaged, a first block is generated based on the first damage information; and the first block is linked to the domestic segment alliance chain; The target vehicle in the international transport phase is monitored in real time by the second monitoring system, and whether the appearance of the target vehicle is damaged is detected by a damage type recognition model based on a neural network according to the monitoring screen image of the second monitoring system; when it is detected that the appearance of the target vehicle is damaged, a second block is generated based on the second damage information; and the second block is chained to the international segment alliance chain; The target vehicle in the foreign transportation stage is monitored in real time through the third monitoring system, and whether the appearance of the target vehicle is damaged is detected by a damage type recognition model based on a neural network according to the monitoring screen image of the third monitoring system; when it is detected that the appearance of the target vehicle is damaged, a third block is generated based on the third damage information; the third block is linked to the foreign segment alliance chain; wherein the first damage information, the second damage information and the third damage information all include time, place, target vehicle information, damage details, monitoring screen images corresponding to the damaged time period, and information of the vehicle carrying the target vehicle; information exchange is realized between the domestic segment alliance chain, the international segment alliance chain and the foreign segment alliance chain based on a cross-chain protocol; the second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

2. The automobile transportation logistics data tracing method for cross-border trade according to claim 1 is characterized in that: The step of detecting whether the exterior of the target vehicle is damaged by using a damage type recognition model based on a neural network according to the monitoring screen image includes: Acquire two adjacent monitoring screen image frames, and use the previous monitoring screen image frame and the next monitoring screen image frame as the first monitoring screen image and the second monitoring screen image, respectively; Locating the target vehicle from the first monitoring screen image to obtain a first target vehicle image; Locating the target vehicle from the second monitoring screen image to obtain a second target vehicle image; Comparing the first target vehicle image with the second target vehicle image to obtain a difference area image; Based on the difference area image, judging whether the target vehicle is damaged and determining the damage type through a damage type recognition model based on a neural network; wherein the damage type includes: scratches, dents and component damage; When it is determined that the target vehicle is damaged, the difference area image is input into a part recognition model based on a neural network to identify the damaged part; If the damaged part cannot be identified, the difference area is enlarged until the part identification model identifies the damaged part.

3. The automobile transportation logistics data tracing method for cross-border trade according to claim 2 is characterized in that: The damage type identification model is trained by the following method: Label the collected damaged vehicle images, marking the actual damaged area and damage type; wherein the damage type includes scratches, dents and component damage; Label the collected vehicle images that may cause false damage reports, and mark the areas and objects that may cause false damage reports; objects that may cause false damage reports include dirt, water stains, and environmental reflections; The preset neural network model is used to perform deep learning training on the labeled images to obtain a neural network-based damage type recognition model.

4. The automobile transportation logistics data tracing method for cross-border trade according to claim 2 is characterized in that: The cameras of the first monitoring system, the second monitoring system and the third monitoring system all include movable anti-shake lenses, gyroscopes and image sensors. Before the step of acquiring two adjacent frames of monitoring screen images, the method further includes: Obtaining the angle change amplitude value compared with the previous moment through the gyroscope; Determine the moving direction and moving distance of the anti-shake lens according to the corresponding relationship between the angle change amplitude value and the anti-shake lens compensation vector at the preset focal length; adjusting the position of the anti-shake lens according to the moving direction and moving distance of the anti-shake lens so that the image sensor captures the anti-shake image to obtain the anti-shake image; Inputting the anti-shake image as a blurred input image in application into a pre-trained first deblurring neural network, and predicting a reconstructed blurred image through a first task of the first deblurring neural network; Updating the meta-training weights of the first deblurring neural network based on the anti-shake image and the reconstructed blurred image until the training is completed to obtain a second deblurring neural network; The anti-shake image is input into the second deblurring neural network to generate a deblurred monitoring image.

5. The automobile transportation logistics data tracing method for cross-border trade according to claim 1 is characterized in that: After detecting that the exterior of the target vehicle is damaged and generating a block based on the damage information, the method further includes: Acquire vibration data collected by a vibration sensor during a period when the target vehicle is damaged; wherein the vibration sensor is installed on a vehicle that transports the target vehicle; Analyzing whether the vehicle is subjected to bumps based on the vibration data and the vibration curve and whether the bumps are greater than a preset threshold; If yes, determining the current location of the vehicle through a positioning module installed on the vehicle; Obtain other means of transport that will pass through the location from the logistics tracking system; Sending prompt information of bumpy rides at the location and prompt information of damage to the appearance of the consignment target vehicle at the location to other vehicles that will pass through the location; wherein the prompt information of bumpy rides at the location includes distance information from the location and voice prompt information about the bumps.

6. The automobile transportation logistics data tracing method for cross-border trade according to claim 1 is characterized in that: After detecting that the exterior of the target vehicle is damaged and generating a block based on the damage information, the method further includes: Extracting the target vehicle information, the damage details, the monitoring screen image corresponding to the damaged time period, and the vehicle information carrying the target vehicle from the damage information; wherein the damage details include the damage type and the damaged location; Extracting a damaged image of the target vehicle from the monitoring screen image and explicitly marking the damaged position in the damaged image of the target vehicle to obtain a damaged marked image of the target vehicle; Sending the target vehicle information, the damage details, the vehicle information carrying the target vehicle, and the target vehicle damage mark image to a management system of a logistics company responsible for the next transportation stage; The management system forwards the target vehicle information, the damage details, the vehicle information carrying the target vehicle, and the target vehicle damage mark image to the corresponding docking person terminal.

7. A data tracing system for automobile transportation logistics for cross-border trade, characterized in that: The system comprises: The domestic segment module is used to monitor the target vehicle in the domestic transportation stage in real time through the first monitoring system, and detect whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network according to the monitoring screen image of the first monitoring system; when it is detected that the appearance of the target vehicle is damaged, a first block is generated based on the first damage information; and the first block is chained to the domestic segment alliance chain; The international segment module is used to monitor the target vehicle in the international transport phase in real time through the second monitoring system, and detect whether the appearance of the target vehicle is damaged through a damage type recognition model based on a neural network according to the monitoring screen image of the second monitoring system; when it is detected that the appearance of the target vehicle is damaged, generate a second block based on the second damage information; and link the second block to the international segment alliance chain; The foreign segment module is used to monitor the target vehicle in the foreign transportation stage in real time through the third monitoring system, and detect whether the appearance of the target vehicle is damaged according to the monitoring screen image of the third monitoring system through a neural network-based damage type recognition model; when it is detected that the appearance of the target vehicle is damaged, a third block is generated based on the third damage information; the third block is linked to the foreign segment alliance chain; wherein the first damage information, the second damage information and the third damage information all include time, place, target vehicle information, damage details, monitoring screen images corresponding to the damaged time period, and information of the vehicle carrying the target vehicle; the domestic segment alliance chain, the international segment alliance chain and the foreign segment alliance chain realize information exchange based on the cross-chain protocol; the second block includes the hash value of the first block connected to it, and the third block includes the hash value of the second block connected to it.

8. The automobile transportation logistics data tracing system for cross-border trade according to claim 7 is characterized in that: According to the monitoring screen image, detecting whether the exterior of the target vehicle is damaged by using a damage type recognition model based on a neural network includes: Acquire two adjacent monitoring screen image frames, and use the previous monitoring screen image frame and the next monitoring screen image frame as the first monitoring screen image and the second monitoring screen image, respectively; Locating the target vehicle from the first monitoring screen image to obtain a first target vehicle image; Locating the target vehicle from the second monitoring screen image to obtain a second target vehicle image; Comparing the first target vehicle image with the second target vehicle image to obtain a difference area image; Based on the difference area image, judging whether the target vehicle is damaged and determining the damage type through a damage type recognition model based on a neural network; wherein the damage type includes: scratches, dents and component damage; When it is determined that the target vehicle is damaged, the difference area image is input into a part recognition model based on a neural network to identify the damaged part; If the damaged part cannot be identified, the difference area is enlarged until the part identification model identifies the damaged part.

9. The automobile transportation logistics data tracing system for cross-border trade according to claim 8 is characterized in that: The damage type identification model is trained by the following method: Label the collected damaged vehicle images, marking the actual damaged area and damage type; wherein the damage type includes scratches, dents and component damage; Label the collected vehicle images that may cause false damage reports, and mark the areas and objects that may cause false damage reports; objects that may cause false damage reports include dirt, water stains, and environmental reflections; The preset neural network model is used to perform deep learning training on the labeled images to obtain a neural network-based damage type recognition model.

10. The automobile transportation logistics data tracing system for cross-border trade according to claim 7 is characterized in that: The system further comprises: A first acquisition module is used to acquire vibration data collected by a vibration sensor during a period when the target vehicle is damaged; wherein the vibration sensor is installed on a vehicle that transports the target vehicle; An analysis module, configured to analyze whether the vehicle is subjected to bumps based on the vibration data and the vibration curve and whether the bumps are greater than a preset threshold; a positioning module, for determining the current position of the vehicle through a positioning module installed on the vehicle if yes; A second acquisition module is used to acquire other transportation vehicles that will pass through the location from a logistics tracking system; A sending module is used to send prompt information about the bumpy location and prompt information about the damage to the appearance of the consignment target vehicle at the location to other vehicles that are about to pass through the location; wherein the prompt information about the bumpy location includes the distance information from the location and voice prompt information about the bumpy location.

Images