Vehicle exterior surface stain detection method, device, system, storage medium and vehicle

Abstract

The application discloses a vehicle outer surface stain detection method, device, system, storage medium and vehicle. The method comprises the following steps: performing start calibration on multiple types of sensors, triggering synchronous collection after calibration, weighting and fusing three types of data of a vehicle body laser point cloud, ultrasonic distance and a visual image, dividing a unified point cloud data set after fusion into multiple subsets according to a preset area, fitting a curve for each subset, and splicing the subsets into a complete vehicle body three-dimensional contour model. Based on the spatial reference of the three-dimensional contour model, a stain area is identified based on the visual image, stain coordinates and area are marked to a corresponding position of the model, integrity verification and stain mark accuracy verification are performed on the three-dimensional contour model after marking, and a detection result is output if the verification is passed. The scheme can improve the completeness and detection accuracy of the vehicle outer surface stain detection result.

Description

Technical Field

[0001] This application relates to the field of intelligent car wash technology, and in particular to a method, device, system, storage medium, and vehicle for detecting stains on the exterior surface of a vehicle. Background Technology

[0002] Current solutions for detecting dirt on the exterior surface of vehicles in intelligent car wash scenarios have significant drawbacks: most solutions only use a single visual sensor to identify dirt, which can only output dirt information at the two-dimensional image level and cannot obtain the three-dimensional spatial coordinates corresponding to the dirt. This makes it difficult to support the high-precision trajectory planning of the car wash gun, and the point cloud data of the blind spot at the edge of the vehicle body is seriously missing, resulting in poor detection accuracy. This leads to safety hazards such as missed washing, over-washing, and even the car wash gun scratching the vehicle body. Summary of the Invention

[0003] This application provides a method, apparatus, system, storage medium, and vehicle for detecting stains on the exterior surface of a vehicle, in order to improve the integrity and accuracy of stain detection results on the exterior surface of a vehicle.

[0004] This application provides a method for detecting stains on the exterior surface of a vehicle, including: Perform startup calibration on multiple types of sensors, and trigger synchronous data acquisition after successful calibration; The three types of data collected—vehicle laser point cloud, ultrasonic distance, and visual image—are weighted and fused together. Based on the fused unified point cloud dataset, the model is divided into multiple subsets according to a preset region, and the surfaces are fitted separately and then stitched together to form a complete three-dimensional outline model of the vehicle body. With the spatial reference of the 3D contour model, the stain area is identified based on visual image, and the stain coordinates and area are marked to the corresponding position on the model; The complete 3D contour model is subjected to integrity verification and stain marking accuracy verification. If the verification passes, the detection result is output. If either verification fails, the lidar and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained from the supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

[0005] The beneficial effects of this application are as follows: by calibrating multiple types of sensors, the acquisition reference deviation is eliminated from the hardware level. Then, the three types of sensor data are weighted and fused to improve the integrity of point cloud spatial data acquisition and eliminate blind spots at the vehicle body edges. Next, the fused point cloud dataset is partitioned and fitted with curved surfaces to generate a three-dimensional contour model. The mapping of two-dimensional stain information to three-dimensional coordinates is completed under the spatial reference of the contour model. Finally, by verifying the point cloud overlap, verifying the accuracy of stain marking, and performing directional supplementary scanning, the accuracy of stain detection results is improved. This solves the problems of insufficient stain detection accuracy and many blind spots in the car wash scenario, and improves the integrity and accuracy of stain detection results on the vehicle's outer surface.

[0006] In one embodiment, performing startup calibration on multiple types of sensors includes: Calibrate the lidar ranging using a preset reference point; The ultrasonic distance is calibrated using a known distance calibration value; The camera's white balance is calibrated using a preset solid color area.

[0007] In one embodiment, the step of performing start-up calibration on multiple types of sensors and triggering synchronous data acquisition after successful calibration includes: Collect laser reflection signals from preset reference points and calculate the ranging offset of the lidar; An ultrasonic test signal is emitted toward a preset fixed calibration surface, and the distance deviation of the ultrasonic sensor is calculated. Collect sample images of a preset standard solid color area and adjust the white balance parameters of the visual camera; The ranging offset of the lidar, the distance deviation of the ultrasonic sensor, and the calibrated white balance parameters of the camera are used as calibration parameters and written into the built-in configuration register of the corresponding sensor for subsequent data acquisition.

[0008] In one embodiment, triggering synchronous data acquisition after successful calibration includes: Synchronous acquisition signals are transmitted to each sensor via the CAN bus, and sampling is triggered based on a unified timestamp; Configure the sampling frequency of various sensors so that the total time from calibration start to test result output does not exceed the preset time.

[0009] In one embodiment, after triggering synchronous acquisition, the acquisition process of vehicle body laser point cloud data includes: Laser pulses are emitted at a preset frequency to scan the vehicle body and obtain point cloud data covering the entire vehicle outline; The three-dimensional coordinate information of each point cloud point is recorded based on the point cloud data, providing the main point cloud input for weighted fusion of multi-source data.

[0010] In one embodiment, after triggering synchronous acquisition, the ultrasonic distance acquisition process includes: Ultrasonic waves are emitted toward the edge of the vehicle body at a preset frequency to collect distance data to the edge of the vehicle body. The spatial position information of the vehicle body edge is recorded based on the vehicle body edge distance data, providing a reference for point cloud edge correction.

[0011] In one embodiment, after triggering synchronous acquisition, the visual image acquisition process includes: High-definition images of the vehicle surface are captured at a preset frame rate, and pixel color and grayscale information are recorded. After each frame of image acquisition is completed, the timestamp of that frame is aligned with the timestamps of the point cloud and ultrasonic data at the current moment to ensure that the multi-source data is synchronized in time before being sent into the fusion process.

[0012] In one embodiment, the method further includes: It receives the collection signals returned by the lidar at the front of the roof rack, and the collection signals returned by the ultrasonic sensors at the four corners of the roof rack and the cameras on both sides of the roof rack; Verify the timestamps of the acquired signals. When the timestamp deviations of all signals are within the preset synchronization allowable range, send the acquired data into the data fusion process.

[0013] In one embodiment, the weighted fusion of the three types of data—the acquired vehicle body laser point cloud, ultrasonic distance, and visual image—includes: Neighborhood averaging denoising is performed on the laser point cloud as the main body to remove noise points that exceed the deviation threshold; By combining ultrasonic distance data to correct the deviation of point cloud edge contours, the image texture is mapped to the corresponding point cloud points, and a unified point cloud dataset is output.

[0014] In one embodiment, the step of dividing the fused unified point cloud dataset into multiple subsets according to a preset region, fitting surfaces to each subset, and stitching them together to form a complete three-dimensional outline model of the vehicle body includes: The point cloud is divided into subsets based on the roof area, hood area, left door area, right door area, and trunk area; The corresponding partition surface equations are generated by fitting each subset, and the complete three-dimensional outline model of the vehicle body is spliced ​​together as a spatial positioning reference.

[0015] In one embodiment, the step of identifying the stain region based on a visual image and marking the stain coordinates and area to the corresponding position on the model, under the spatial reference of a three-dimensional contour model, includes: Perform grayscale conversion and threshold segmentation on the visual image to extract grayscale anomaly regions; The percentage of pixels in abnormal areas is counted, and if it reaches a preset threshold, it is identified as a stain. After calculating the area, the coordinates are mapped to the corresponding position on the 3D model.

[0016] In one embodiment, performing integrity verification on the marked 3D contour model includes: Statistically determine the point cloud spatial overlap between the stated 3D contour model and the pre-stored standard vehicle body point cloud model; When the spatial overlap of the point cloud reaches the first preset overlap threshold, the integrity verification is deemed to be successful; otherwise, it is determined that there is an uncovered area on the vehicle body and a rescan signal is triggered. In one embodiment, the stain mark accuracy verification includes: Extract the area overlap of the stain region obtained from the visual image recognition of two adjacent frames; When the overlap reaches the second preset overlap threshold, the accuracy of the marker is verified as passed; otherwise, it is determined as a misidentification and a rescan signal is triggered.

[0017] In one embodiment, if the verification passes, the detection result is output, including: The coordinates and area of ​​the marked stain area are bound to the 3D contour model to generate detection result data for cleaning path planning; The detection results are transmitted to the subsequent trajectory planning module via the vehicle communication interface.

[0018] This application also provides a vehicle exterior surface stain detection device for implementing the vehicle exterior surface stain detection method described in any of the above embodiments, including: The data acquisition module is used to perform startup calibration on multiple types of sensors, and triggers synchronous data acquisition after the calibration is successful. The fusion module is used to perform weighted fusion of three types of data: vehicle laser point cloud, ultrasonic distance, and visual image. The fitting module is used to fit surfaces to multiple subsets of a preset region based on the fused unified point cloud dataset, and then stitch them together to form a complete 3D outline model of the vehicle body. The marking module is used to identify stain areas based on visual images under the spatial reference of the 3D contour model and mark the stain coordinates and area to the corresponding position on the model. The verification module is used to perform integrity verification and stain marking accuracy verification on the marked 3D contour model. If the verification passes, the detection result is output. If any verification fails, the LiDAR and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained from the supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

[0019] This application also provides a vehicle exterior surface stain detection system, including: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to implement the vehicle exterior surface stain detection method described in any of the above embodiments.

[0020] This application also provides a computer-readable storage medium, which, when the instructions in the storage medium are executed by a processor corresponding to a vehicle exterior surface stain detection system, enables the vehicle exterior surface stain detection system to implement the vehicle exterior surface stain detection method described in any of the above embodiments.

[0021] This application also provides a vehicle, including: The vehicle exterior surface stain detection device as described in any of the above embodiments; or Vehicle exterior surface stain detection system as described in any of the above embodiments.

[0022] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0023] The technical solution of this application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate the present application and form part of the specification. They are used together with the embodiments of the present application to explain the application and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a method for detecting stains on the exterior surface of a vehicle according to an embodiment of this application; Figure 2 This is a flowchart of a sensor calibration process in one embodiment of this application; Figure 3 This is a flowchart of the sensor calibration process in another embodiment of this application; Figure 4 This is a flowchart illustrating the sensor-triggered sampling process in one embodiment of this application; Figure 5 This is a flowchart illustrating the acquisition of vehicle body laser point cloud data in one embodiment of this application; Figure 6 This is a flowchart illustrating the acquisition of ultrasonic distance data in one embodiment of this application; Figure 7 This is a flowchart illustrating the acquisition process of visual image data in one embodiment of this application; Figure 8 This is a flowchart of time synchronization verification in one embodiment of this application; Figure 9This is a data fusion flowchart in one embodiment of this application; Figure 10 This is a flowchart of surface fitting in one embodiment of this application; Figure 11 This is a flowchart of stain identification and marking in one embodiment of this application; Figure 12 This is a flowchart illustrating the integrity verification of a three-dimensional contour model in one embodiment of this application; Figure 13 This is a flowchart illustrating the accuracy verification of stain markings in one embodiment of this application; Figure 14 This is a flowchart illustrating the output of detection results in one embodiment of this application; Figure 15 This is a schematic diagram of the structure of a vehicle exterior surface stain detection device according to an embodiment of this application; Figure 16 This is a schematic diagram of the hardware structure of a vehicle exterior surface stain detection system according to one embodiment of this application. Detailed Implementation

[0025] The preferred embodiments of this application are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit this application.

[0026] Figure 1 This is a flowchart of a method for detecting stains on the exterior surface of a vehicle according to an embodiment of this application, as follows: Figure 1 As shown, the method can be implemented as follows: S101-S105: In step S101, start-up calibration is performed on multiple types of sensors, and synchronous acquisition is triggered after the calibration is qualified; In step S102, the three types of data acquired—vehicle laser point cloud, ultrasonic distance, and visual image—are weighted and fused. In step S103, based on the fused unified point cloud dataset, the data is divided into multiple subsets according to a preset region, and the surfaces are fitted separately and then stitched together to form a complete three-dimensional outline model of the vehicle body. In step S104, under the spatial reference of the three-dimensional contour model, the stain area is identified based on the visual image, and the stain coordinates and area are marked to the corresponding position of the model. In step S105, integrity verification and stain marking accuracy verification are performed on the marked 3D contour model. If the verification passes, the detection result is output. If either verification fails, the lidar and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained from the supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

[0027] In this application, multiple types of sensors are subjected to initial calibration. Once the calibration is successful, synchronous data acquisition is triggered, thus avoiding subsequent end-to-end data deviations caused by calibration failure of a single type of sensor from the source.

[0028] Specifically, the method in this application can be applied to an in-vehicle intelligent car wash system equipped with multiple types of sensors. For example, the multiple types of sensors include a lidar fixedly installed at the front of the roof rack, four sets of ultrasonic sensors respectively set at the four corners of the roof rack, and two high-definition vision cameras symmetrically arranged on the left and right sides of the roof rack. The installation positions of all sensors are pre-calibrated, and the spatial offset parameters of each sensor relative to the vehicle's central coordinate system are recorded in advance, providing a basic reference for the spatial alignment of subsequent multi-source data. In one embodiment, when performing startup calibration on multiple types of sensors, the lidar ranging is calibrated using a preset reference point, for example, using a fixed calibration reflector pre-set at the entrance of the car wash area as a preset reference point; the ultrasonic distance is calibrated using a known distance calibration value, for example, using a preset distance calibration value pre-stored before the system leaves the factory to calibrate the ultrasonic distance detection accuracy; and the camera white balance is calibrated using a preset solid color area, for example, using a standard pure white calibration plate pre-laid on the side of the car wash area as a preset solid color area to calibrate the camera white balance. Furthermore, a lidar system can actively emit a probe beam to scan a preset reference point, collecting the laser reflection signal at that point to calculate the ranging offset; an ultrasonic sensor emits an ultrasonic test signal towards a fixed calibration reference surface, recording the round-trip flight time of the signal to calculate the current distance deviation; a vision camera continuously acquires multiple frames of images by pointing at a preset solid color area, analyzing the RGB channel grayscale distribution of each frame to adjust the white balance parameters. Then, all the offsets, deviation compensation values, and white balance parameters obtained from calibration are directly written into the built-in configuration register of the corresponding sensor. Subsequent acquisition operations will automatically call this set of calibration parameters to complete data compensation, ensuring the consistency and accuracy benchmark of all subsequent acquired data from the underlying sensor level.

[0029] After successful calibration, the vehicle-mounted central control unit sends a synchronous acquisition signal with a global timestamp to all sensors via the CAN bus. Upon receiving the instruction, all sensing devices trigger sampling based on a completely unified time reference. The system can also flexibly configure the sampling frequency of various sensors according to the detection requirements of different vehicle models. In a typical embodiment, the sampling frequency of the lidar is set to 10Hz, the sampling frequency of the ultrasonic sensor is set to 5Hz, and the sampling frame rate of the vision camera is set to 30fps. Ultimately, the total time from calibration start to detection result output can be kept within a preset time of 5 seconds, effectively improving the efficiency of the entire detection process.

[0030] The system then initiates a parallel acquisition process for three types of multi-source data. For example, when acquiring laser point cloud data, the lidar continuously emits laser pulses at a preset frequency to the outer surface of the vehicle, scanning and covering the entire outer contour of the vehicle to obtain complete three-dimensional contour point cloud data. Simultaneously, it records the X / Y / Z three-dimensional coordinate information of each point cloud point, providing the main point cloud input for subsequent multi-source data weighted fusion. When acquiring ultrasonic data, four sets of corner ultrasonic sensors emit ultrasonic waves at a preset frequency towards the edge of the vehicle, specifically acquiring distance data of the vehicle edge area where the lidar detection angle is limited. Simultaneously, it records the spatial position information of each edge detection blind spot, providing accurate reference for subsequent edge contour correction of the point cloud. When acquiring visual images, high-definition visual cameras on both sides continuously acquire high-definition images of the surface on the left and right sides of the vehicle at a preset frame rate, completely recording the RGB color information and grayscale information of each pixel. After each frame of image acquisition is completed, the timestamp of that frame is automatically aligned with the timestamp of the point cloud and ultrasonic data at the current moment to ensure the time synchronization of the three types of multi-source data before being sent to the subsequent fusion process.

[0031] In one embodiment, the vehicle central control processing unit sets up a dedicated data access verification step. It first receives the collection signal returned by the lidar at the front of the roof rack, and then sequentially receives the collection signals returned by the four sets of ultrasonic sensors at the four corners of the roof rack and the two cameras on both sides of the roof rack. It compares the deviation of the timestamp carried by each signal with the global reference timestamp. Only when the timestamp deviation of all signals is within the preset synchronization allowable range (such as 20ms) will the three types of data be sent into the subsequent data fusion process. This completely avoids the problem of spatial misalignment of multi-source data caused by asynchronous acquisition at the data access level.

[0032] The collected vehicle body laser point cloud, ultrasonic distance data, and visual image data are weighted and fused. Specifically, during the fusion process, differentiated fusion weights are set for the laser point cloud, ultrasonic distance data, and visual image data, and the spatial accuracy of the fusion result is prioritized through weighted allocation. Neighborhood averaging denoising is performed on the laser point cloud as the main component to remove noise points whose coordinate deviation from the neighboring points exceeds a preset deviation threshold. The ultrasonic distance data is used to complete the missing point cloud in the blind spot of the vehicle body edge, correcting the deviation of the vehicle body edge contour in the original point cloud. Through the pre-calibrated image-point cloud feature mapping relationship, the texture information of the visual image is mapped one by one to the corresponding point cloud points, and finally, a unified point cloud dataset with alignment is output.

[0033] Based on the fused unified point cloud dataset, the system divides the data into multiple subsets according to preset regions, fits surfaces to each subset, and then stitches them together to form a complete 3D contour model of the vehicle body. Specifically, the system automatically divides the unified point cloud dataset into five independent point cloud subsets according to the inherent structural features of the vehicle body: the roof area, the hood area, the left door area, the right door area, and the trunk area. For each subset, a corresponding curvature algorithm, such as the Bezier surface algorithm, is applied to fit and generate the surface equations for the corresponding partition. The surface equations of the five partitions are then smoothly transitioned at the junctions of the vehicle body structure and stitched together to obtain a complete 3D contour model of the vehicle body covering the entire outer surface, which serves as a unified spatial reference for subsequent stain spatial localization.

[0034] Under the spatial reference of the 3D contour model, stain areas are identified based on visual images, and the stain coordinates and area are marked on the corresponding positions of the model. Specifically, the vehicle body image captured by the visual camera is first converted to grayscale, transforming the three-channel RGB image into a single-channel grayscale image. For example, a threshold segmentation operation is performed with a grayscale deviation threshold of 30 to extract all grayscale abnormal areas where the grayscale value deviates from the basic color of the vehicle body beyond the threshold. Then, the proportion of pixels in each abnormal area to the total pixels of the corresponding local vehicle body image is counted. When the proportion reaches a preset judgment threshold (e.g., 50%), the area is judged as a valid stain. After calculating the two-dimensional pixel area of ​​the stain area, the two-dimensional pixel position of the stain is converted into three-dimensional spatial coordinates through a pre-constructed image-point cloud mapping relationship and marked on the corresponding position of the 3D contour model.

[0035] The system performs integrity verification and stain marking accuracy verification on the marked 3D contour model. If the verification passes, the detection result is output. If either verification fails, the system controls the lidar and / or visual camera to perform directional rescanning. The local point cloud and local image data obtained from the rescanning are re-weighted and fused to generate a new model and are verified again until all verifications pass or the preset number of rescanning cycles is reached. When performing integrity verification on the marked 3D contour model, the point cloud spatial overlap between the 3D contour model and the pre-stored standard vehicle body point cloud model is calculated. When the point cloud spatial overlap reaches a first preset overlap threshold, the integrity verification is considered successful; otherwise, it is determined that there is an uncovered area on the vehicle body and a rescanning signal is triggered. When performing stain marking accuracy verification, the stain areas identified from two adjacent frames of visual images are extracted, and the ratio of the overlapping pixel area of ​​the two stain areas to the total pixel area of ​​the two areas is calculated to obtain the stain area pixel overlap. When the stain area pixel overlap reaches a second preset overlap threshold, the marking accuracy verification is considered successful; otherwise, the area is determined to be a misidentified false stain and a rescanning signal is triggered. When the integrity verification and stain marking accuracy verification of the marked 3D contour model are both passed, the detection results are output. Specifically, the 3D coordinates and area parameters of all marked stain areas are bound to the vehicle body 3D contour model to generate structured detection result data that can be directly used for cleaning path planning. Then, the detection result data is transmitted in real time to the downstream car wash gun trajectory planning module through the vehicle's built-in CAN communication interface, directly providing complete and reliable input data for subsequent car wash operations.

[0036] Compared to existing technologies, this invention employs a multi-sensor pre-calibration mechanism to independently calibrate lidar ranging, ultrasonic distance detection, and camera white balance. All calibration parameters are directly written into the sensor configuration registers, eliminating system acquisition biases from different sensing devices at the hardware level and improving the consistency and reliability of the acquired data. Furthermore, it breaks through the dimensional limitations of traditional single-vision or single-laser detection schemes by using a weighted fusion of three types of data: laser point cloud as the spatial subject, ultrasonic data to fill in blind spots at the vehicle body edges, and visual images to provide texture information. This eliminates detection blind spots in traditional schemes such as those at the vehicle body edges and bottom. Moreover, it breaks away from the conventional logic of fitting the entire vehicle body model, dividing the point cloud into components such as the roof, hood, and doors. After dividing the region into subsets, surface equations are independently fitted and then stitched together to fully adapt to the curvature characteristics of different parts of the vehicle body. This avoids the problems of abrupt surface transitions and local contour distortion caused by overall fitting, ultimately generating a complete 3D contour model of the vehicle body. Using this high-precision 3D contour model as a unified spatial reference, the pixel information of the stain area obtained from visual image recognition is directly mapped to the corresponding spatial coordinates and area parameters on the 3D model, solving the industry pain point that traditional 2D stain detection solutions cannot output the 3D location of stains. Furthermore, by performing integrity verification and stain marking accuracy verification on the marked 3D contour model, and verifying the integrity of the vehicle body through point cloud spatial overlap, the pass rate of the final output detection results is improved. Through these multiple steps, the long-standing problems of insufficient accuracy, numerous blind spots, and poor reliability in stain detection in car wash scenarios are solved.

[0037] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited to the following embodiments. This embodiment is applied to an in-vehicle intelligent car wash system equipped with a complete set of sensing devices, wherein one set of lidar is fixedly installed at the front center of the roof rack, four sets of ultrasonic sensors are respectively installed at the four corners of the roof rack, and two sets of high-definition cameras are installed on the left and right sides of the roof rack. All sensing devices interact with the in-vehicle central control processing unit through a CAN bus. The specific execution flow is as follows: The first step involves the system automatically entering the sensor startup calibration phase after power-on. The central control processing unit controls the lidar to collect the strong reflected laser signal from the pre-set calibration reflector, calculates the current ranging offset of the lidar, and writes it into the configuration register. It also controls the ultrasonic sensor to emit ultrasonic test signals to measure the distance to a fixed calibration reference surface 1.5m away from the system, calculates the current distance deviation of the ultrasonic sensor, and completes parameter compensation. Finally, it controls the cameras on both sides to collect images of the pre-set pure white calibration area, automatically adjusts the white balance parameters, and writes them into the camera configuration register. After all calibration operations of the three types of sensors are completed, the system automatically determines that the calibration is qualified and enters the acquisition ready state.

[0038] The second step involves the central control unit sending synchronous acquisition commands to all sensors via the CAN bus. All sensors begin sampling based on a globally unified timestamp: the lidar emits laser pulses at a frequency of 10Hz to scan the vehicle body, acquiring point cloud data covering the entire vehicle outline and fully recording the 3D coordinate information of each point cloud point as the main input for fusion; four sets of ultrasonic sensors at the corners of the roof rack emit ultrasonic waves at a frequency of 5Hz to collect distance data in the edge areas of the vehicle body and simultaneously record the spatial position information of the edge detection blind spots; and the high-definition cameras on both sides acquire high-definition images of the vehicle body surface at a frame rate of 30fps, fully recording the color and grayscale information of all pixels. The central control unit receives the acquisition signals returned by all sensors, verifies that the timestamp deviation of all signals is within the preset 20ms synchronization allowable range, and then sends the three types of data into the weighted fusion module for processing. The total time of the entire acquisition process is controlled within 3 seconds.

[0039] The third step involves the weighted fusion module initiating multi-source data fusion processing: First, neighborhood averaging denoising is performed on the laser point cloud data as the main component, directly removing abnormal noise points with a deviation of more than 2cm from the neighboring points; then, edge distance data collected by the ultrasonic sensor is used to correct and complete the point cloud with missing edge blind areas in the original point cloud; finally, the texture information of the visual image is mapped one by one to the corresponding point cloud points, and the aligned unified point cloud dataset is output.

[0040] The fourth step is to perform partition modeling on the fused unified point cloud dataset: the complete point cloud dataset is divided into 5 independent subsets according to the roof area, hood area, left door area, right door area, and trunk area. For each subset, the Bezier surface algorithm is adapted to fit and generate the corresponding partition surface equation. The surface equations of the 5 partitions are spliced ​​and integrated to obtain a complete 3D contour model of the vehicle body covering the whole vehicle, which serves as a unified spatial positioning reference for subsequent stain recognition and marking.

[0041] The fifth step is to perform stain recognition and marking: The central control processing unit performs grayscale processing on the vehicle body image captured by the vision camera, sets the grayscale deviation threshold to 30, performs threshold segmentation, and extracts all suspected stain areas with grayscale anomalies; it calculates the pixel ratio of each suspected stain area, and when the pixel ratio reaches the preset threshold of 50% of the total pixels of the local image, the area is officially determined to be a valid stain. After calculating the two-dimensional area of ​​the stain area, the position of the stain is mapped to the corresponding spatial coordinate point of the three-dimensional contour model through the pre-established image-point cloud mapping relationship, thus completing the coordinate and area marking of the stain area.

[0042] Step 6: Initiate a dual closed-loop verification process: The first step verifies the accuracy of stain marking by extracting the stain region sets obtained from the visual image recognition of two adjacent frames and calculating the ratio of the overlapping pixel area of ​​the two regions to the total pixel area. In this embodiment, this ratio is 86%, which meets the second preset overlap threshold of 80%, thus determining that all stain markings are accurate and there are no false stains. The second step verifies the integrity by retrieving the 3D model generated in this acquisition and the pre-stored standard vehicle body point cloud model, and calculating the point cloud spatial overlap between the two sets of models. In this embodiment, the calculated point cloud spatial overlap is 87%, which meets the first preset overlap threshold of 80%, thus determining that the integrity verification is passed. Without triggering a rescan operation, the coordinates and area of ​​the marked stain regions are directly bound to the 3D contour model to generate detection result data that can be directly used for cleaning path planning. This data is then transmitted to the subsequent car wash gun trajectory planning module through the vehicle communication interface, completing the entire detection process. If, in a subsequent practical application, the pixel overlap of a detected stain area is only 72%, or the point cloud spatial overlap is only 72%, and either of these fails to meet the corresponding threshold requirements, the system will automatically trigger a supplementary scanning signal. This will control the LiDAR and vision camera to collect point cloud and image data of the local area. The local data obtained from the supplementary scanning will be sent back to the weighted fusion module to regenerate the model. The double verification will be performed again, and this process will be repeated until both verifications pass, or until the preset limit of three supplementary scanning times is reached. The final detection result will then be directly output, providing high-precision path planning for the in-vehicle intelligent car wash scenario, effectively avoiding the risks of missed cleaning and scratches during the car wash operation.

[0043] The beneficial effects of this application are as follows: by calibrating multiple types of sensors, the acquisition reference deviation is eliminated from the hardware level. Then, the three types of sensor data are weighted and fused to improve the integrity of point cloud spatial data acquisition and eliminate blind spots at the vehicle body edge. Then, the fused point cloud dataset is partitioned and fitted with curved surfaces to generate a three-dimensional contour model. The mapping of two-dimensional stain information to three-dimensional coordinates is completed under the spatial reference of the contour model. Finally, the accuracy of stain detection results is improved by point cloud overlap verification, stain marking accuracy verification, and directional supplementary scanning, which solves the problems of insufficient stain detection accuracy and many blind spots in the car wash scenario.

[0044] Figure 2 This is a flowchart of a sensor calibration process in one embodiment of this application, as follows: Figure 2 As shown, step S101 above can be implemented as steps A1-A3 as follows: In step A1, the lidar ranging is calibrated using a preset reference point; In step A2, the ultrasonic distance is calibrated using a known distance calibration value; In step A3, the camera white balance is calibrated using a preset solid color area.

[0045] Figure 3 This is a flowchart of the sensor calibration process in another embodiment of this application, as follows: Figure 3 As shown, step S101 above can be implemented as steps B1-B4 as follows: In step B1, the laser reflection signal of the preset reference point is collected, and the ranging offset of the lidar is calculated; In step B2, an ultrasonic test signal is emitted to a preset fixed calibration surface, and the distance deviation of the ultrasonic sensor is calculated. In step B3, sample images of a preset standard solid color area are acquired, and the white balance parameters of the visual camera are adjusted. In step B4, the ranging offset of the lidar, the distance deviation of the ultrasonic sensor, and the calibrated white balance parameters of the camera are used as calibration parameters and written into the built-in configuration register of the corresponding sensor for subsequent data acquisition.

[0046] Figure 4 This is a flowchart of a sensor-triggered sampling process in one embodiment of this application, as follows: Figure 4 As shown, step S101 above can be implemented as steps C1-C2 as follows: In step C1, synchronous acquisition signals are transmitted to each sensor via the CAN bus, and sampling is triggered based on a unified timestamp; In step C2, the sampling frequency of various sensors is configured so that the total time from calibration start to detection result output does not exceed the preset time.

[0047] Figure 5 This is a flowchart illustrating the acquisition process of vehicle body laser point cloud data in one embodiment of this application, as follows: Figure 5 As shown, after triggering synchronous acquisition, the acquisition process of vehicle body laser point cloud data can be implemented as follows: steps D1-D2: In step D1, laser pulses are emitted at a preset frequency to scan the vehicle body and obtain point cloud data covering the entire vehicle outline; In step D2, the three-dimensional coordinate information of each point cloud point is recorded according to the point cloud data, providing the main point cloud input for multi-source data weighted fusion.

[0048] Figure 6 This is a flowchart illustrating the acquisition of ultrasonic distance data in one embodiment of this application, as follows: Figure 6 As shown, after triggering synchronous acquisition, the ultrasonic distance acquisition process can be implemented as follows: E1-E2: In step E1, ultrasonic waves are emitted towards the edge of the vehicle body at a preset frequency to collect distance data to the edge of the vehicle body. In step E2, the spatial position information of the vehicle body edge is recorded based on the vehicle body edge distance data to provide a reference for point cloud edge correction.

[0049] Figure 7 This is a flowchart of the visual image data acquisition process in one embodiment of this application, as follows: Figure 7 As shown, after triggering synchronous acquisition, the visual image acquisition process can be implemented as follows: F1-F2: In step F1, a high-definition image of the vehicle surface is acquired at a preset frame rate, and pixel color information and grayscale information are recorded. In step F2, after each frame of image acquisition is completed, the timestamp of that frame is aligned with the timestamps of the point cloud and ultrasonic data at the current moment to ensure that the multi-source data are synchronized in time before being sent into the fusion process.

[0050] Figure 8 This is a flowchart of time synchronization verification in one embodiment of this application, as follows: Figure 8 As shown, the method can also be implemented as follows: steps G1-G2: In step G1, the system receives the acquisition signal returned by the lidar at the front of the roof rack, and the acquisition signals returned by the ultrasonic sensors at the four corners of the roof rack and the cameras on both sides of the roof rack. In step G2, the timestamps of the acquired signals are verified. When the timestamp deviations of all signals are within the preset synchronization allowable range, the acquired data is sent into the data fusion process.

[0051] Figure 9 This is a data fusion flowchart in one embodiment of this application, such as... Figure 9 As shown, step S102 above can be implemented as steps H1-H2 as follows: In step H1, neighborhood averaging is performed on the laser point cloud as the main body to remove noise points that exceed the deviation threshold. In step H2, the deviation of the point cloud edge contour is corrected by combining the ultrasonic distance data, the image texture is mapped to the corresponding point cloud points, and a unified point cloud dataset is output.

[0052] Figure 10 This is a flowchart of surface fitting in one embodiment of this application, as shown below. Figure 10 As shown, step S103 above can be implemented as steps I1-I2 as follows: In step I1, the point cloud subsets are divided according to the roof area, hood area, left door area, right door area, and trunk area; In step I2, the corresponding partition surface equations are fitted to each subset, and the complete three-dimensional outline model of the vehicle body is spliced ​​together as a spatial positioning reference.

[0053] Figure 11 This is a flowchart of stain identification and marking in one embodiment of this application, as shown below. Figure 11 As shown, step S104 above can be implemented as steps J1-J2 as follows: In step J1, the visual image is converted to grayscale and thresholded to extract grayscale abnormal regions; In step J2, the percentage of pixels in the abnormal area is counted. If it reaches a preset threshold, it is determined to be a stain. After calculating the area, the coordinates are mapped to the corresponding position mark in the 3D model.

[0054] Figure 12 This is a flowchart illustrating the integrity verification process of a three-dimensional contour model in one embodiment of this application, as follows: Figure 12 As shown, the integrity verification of the marked 3D contour model described in step S105 above can be implemented as the following steps K1-K2: In step K1, the point cloud spatial overlap between the three-dimensional contour model and the pre-stored standard vehicle body point cloud model is calculated. In step K2, when the spatial overlap of the point cloud reaches the first preset overlap threshold, the integrity verification is deemed to be successful; otherwise, it is determined that there is an uncovered area on the vehicle body and a rescan signal is triggered. Figure 13 This is a flowchart illustrating the accuracy verification of stain markings in one embodiment of this application, as follows: Figure 13 As shown, the stain marking accuracy verification described in step S105 above can be implemented as follows: steps L1-L2: In step L1, the area overlap of the stain region identified from two adjacent visual images is extracted; In step L2, when the overlap reaches the second preset overlap threshold, the accuracy verification of the marker is deemed to have passed; otherwise, it is deemed to be a misidentification and a rescan signal is triggered.

[0055] Figure 14 This is a flowchart of the output detection results in one embodiment of this application, such as... Figure 14 As shown, if the verification in step S105 above is successful, the output detection result can be implemented as follows: steps M1-M2: In step M1, the coordinates and area of ​​the marked stain area are bound to the three-dimensional contour model to generate detection result data for cleaning path planning; In step M2, the detection result data is transmitted to the subsequent trajectory planning module via the vehicle communication interface.

[0056] Figure 15 This is a schematic diagram of a vehicle exterior surface stain detection device according to one embodiment of this application, used to implement the vehicle exterior surface stain detection method described in any of the above embodiments, such as... Figure 15 As shown, the device includes: The acquisition module 151 is used to perform start-up calibration on multiple types of sensors, and trigger synchronous acquisition after the calibration is qualified; The fusion module 152 is used to perform weighted fusion of the three types of data acquired: vehicle laser point cloud, ultrasonic distance, and visual image. The fitting module 153 is used to fit surfaces to multiple subsets of a preset region based on the fused unified point cloud dataset, and then stitch them together to form a complete three-dimensional outline model of the vehicle body. The marking module 154 is used to identify the stain area based on the visual image under the spatial reference of the three-dimensional contour model and mark the stain coordinates and area to the corresponding position of the model. The verification module 155 is used to perform integrity verification and stain marking accuracy verification on the marked 3D contour model. If the verification passes, the detection result is output. If any verification fails, the LiDAR and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained by supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

[0057] Figure 16 This is a schematic diagram of the hardware structure of a vehicle exterior surface stain detection system according to an embodiment of this application, as shown below. Figure 16 As shown, the vehicle exterior surface stain detection system includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to implement the vehicle exterior surface stain detection method described in any of the above embodiments.

[0058] Reference Figure 16 The vehicle exterior surface stain detection system may include one or more of the following components: a processing component, a memory, a power supply component, an input / output (I / O) interface, a sensor component, and a communication component.

[0059] The processing component typically controls the overall operation of a vehicle exterior surface stain detection system. The processing component may include one or more processors to execute instructions to complete all or part of the steps of the methods described above. Furthermore, the processing component may include one or more modules to facilitate interaction between the processing component and other components. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0060] The memory is configured to store various types of data to support the operation of the vehicle exterior surface stain detection system. Examples of this data include instructions for any application or method operating on the vehicle exterior surface stain detection system. The memory can be an internal storage unit of the terminal device, such as a hard disk or RAM of the terminal device. The memory can also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The memory is used to store the programs and data required by this application. The memory can also be used to temporarily store data that has been output or will be output.

[0061] The power supply unit provides power to the various components of the vehicle exterior surface stain detection system. The power supply unit may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the vehicle exterior surface stain detection system.

[0062] I / O interfaces provide interfaces between processing components and peripheral interface modules, such as keyboards, click wheels, buttons, etc.

[0063] The sensor assembly includes one or more sensors for providing various aspects of the status assessment of the vehicle exterior surface stain detection system. Additionally, the sensor assembly can detect the on / off state of the vehicle exterior surface stain detection system, the relative positioning of components, and the operational status of the vehicle exterior surface stain detection system or a component of the system. In some embodiments, the sensor assembly may include lidar, ultrasonic sensors, cameras, etc.

[0064] The communication component is configured to enable the vehicle exterior surface stain detection system to communicate with other devices and cloud platforms via wired or wireless means. The vehicle exterior surface stain detection system can access wireless networks based on communication standards, such as CAN bus, automotive Ethernet, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0065] In an exemplary embodiment, the vehicle exterior surface stain detection system may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the vehicle exterior surface stain detection method described in any of the above embodiments.

[0066] This application also provides a computer-readable storage medium, which, when the instructions in the storage medium are executed by a processor corresponding to a vehicle exterior surface stain detection system, enables the vehicle exterior surface stain detection system to implement the vehicle exterior surface stain detection method described in any of the above embodiments.

[0067] This application also provides a vehicle, including: The vehicle exterior surface stain detection device as described in any of the above embodiments; or Vehicle exterior surface stain detection system as described in any of the above embodiments.

[0068] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0069] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0070] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0071] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0072] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting stains on the exterior surface of a vehicle, characterized in that, include: Perform startup calibration on multiple types of sensors, and trigger synchronous data acquisition after successful calibration; The three types of data collected—vehicle laser point cloud, ultrasonic distance, and visual image—are weighted and fused together. Based on the fused unified point cloud dataset, the model is divided into multiple subsets according to a preset region, and the surfaces are fitted separately and then stitched together to form a complete three-dimensional outline model of the vehicle body. With the spatial reference of the 3D contour model, the stain area is identified based on visual image, and the stain coordinates and area are marked to the corresponding position on the model; The complete 3D contour model is subjected to integrity verification and stain marking accuracy verification. If the verification passes, the detection result is output. If either verification fails, the lidar and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained from the supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

2. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, Perform startup calibration on multiple types of sensors, including: Calibrate the lidar ranging using a preset reference point; The ultrasonic distance is calibrated using a known distance calibration value; The camera's white balance is calibrated using a preset solid color area.

3. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, After triggering synchronous acquisition, the acquisition process for vehicle body laser point cloud data includes: Laser pulses are emitted at a preset frequency to scan the vehicle body and obtain point cloud data covering the entire vehicle outline; The three-dimensional coordinate information of each point cloud point is recorded based on the point cloud data, providing the main point cloud input for weighted fusion of multi-source data.

4. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, After triggering synchronous acquisition, the ultrasonic distance acquisition process includes: Ultrasonic waves are emitted toward the edge of the vehicle body at a preset frequency to collect distance data to the edge of the vehicle body. The spatial position information of the vehicle body edge is recorded based on the vehicle body edge distance data, providing a reference for point cloud edge correction.

5. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, After synchronous acquisition is triggered, the visual image acquisition process includes: High-definition images of the vehicle surface are captured at a preset frame rate, and pixel color and grayscale information are recorded. After each frame of image acquisition is completed, the timestamp of that frame is aligned with the timestamps of the point cloud and ultrasonic data at the current moment to ensure that the multi-source data is synchronized in time before being sent into the fusion process.

6. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, The weighted fusion of the three types of data acquired—vehicle laser point cloud, ultrasonic distance, and visual image—includes: Neighborhood averaging denoising is performed on the laser point cloud as the main body to remove noise points that exceed the deviation threshold; By combining ultrasonic distance data to correct the deviation of point cloud edge contours, the image texture is mapped to the corresponding point cloud points, and a unified point cloud dataset is output.

7. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, Based on the fused unified point cloud dataset, multiple subsets are divided into preset regions, and surfaces are fitted to each subset. These subsets are then stitched together to form a complete 3D contour model of the vehicle body, including: The point cloud is divided into subsets based on the roof area, hood area, left door area, right door area, and trunk area; The corresponding partition surface equations are generated by fitting each subset, and the complete three-dimensional outline model of the vehicle body is spliced ​​together as a spatial positioning reference.

8. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, Under the spatial reference of the 3D contour model, stain regions are identified based on visual images, and the stain coordinates and area are marked to the corresponding positions on the model, including: Perform grayscale conversion and threshold segmentation on the visual image to extract grayscale anomaly regions; The percentage of pixels in abnormal areas is counted, and if it reaches a preset threshold, it is identified as a stain. After calculating the area, the coordinates are mapped to the corresponding position on the 3D model.

9. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, The integrity verification of the marked 3D contour model includes: Statistically determine the point cloud spatial overlap between the stated 3D contour model and the pre-stored standard vehicle body point cloud model; When the spatial overlap of the point cloud reaches the first preset overlap threshold, the integrity verification is deemed successful; otherwise, it is determined that there is an uncovered area on the vehicle body and a rescan signal is triggered.

10. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, The stain marking accuracy verification includes: Extract the area overlap of the stain region obtained from the visual image recognition of two adjacent frames; When the overlap reaches the second preset overlap threshold, the accuracy of the marker is verified as passed; otherwise, it is determined as a misidentification and a rescan signal is triggered.

11. The method for detecting stains on the exterior surface of a vehicle according to claim 1, characterized in that, If the verification passes, the detection result will be output, including: The coordinates and area of ​​the marked stain area are bound to the 3D contour model to generate detection result data for cleaning path planning; The detection results are transmitted to the subsequent trajectory planning module via the vehicle communication interface.

12. A device for detecting stains on the exterior surface of a vehicle, characterized in that, include: The data acquisition module is used to perform startup calibration on multiple types of sensors, and triggers synchronous data acquisition after the calibration is successful. The fusion module is used to perform weighted fusion of three types of data: vehicle laser point cloud, ultrasonic distance, and visual image. The fitting module is used to fit surfaces to multiple subsets of a preset region based on the fused unified point cloud dataset, and then stitch them together to form a complete 3D outline model of the vehicle body. The marking module is used to identify stain areas based on visual images under the spatial reference of the 3D contour model and mark the stain coordinates and area to the corresponding position on the model. The verification module is used to perform integrity verification and stain marking accuracy verification on the marked 3D contour model. If the verification passes, the detection result is output. If any verification fails, the LiDAR and / or vision camera are controlled to perform directional supplementary scanning. The local point cloud and local image data obtained from the supplementary scanning are re-weighted and fused to generate a model and are verified again until all verifications pass or the number of cycles reaches the preset number of supplementary scanning cycles.

13. A vehicle exterior surface stain detection system, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to implement the vehicle exterior surface stain detection method as described in any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor corresponding to the vehicle exterior surface stain detection system, the vehicle exterior surface stain detection system is able to implement the vehicle exterior surface stain detection method as described in any one of claims 1-11.

15. A vehicle, characterized in that, include: The vehicle exterior surface stain detection device as described in claim 12; or The vehicle exterior surface stain detection system as described in claim 13.

Images