A new energy environmental sanitation vehicle cleaning control method and system combining vision and radar, electronic equipment and storage medium

Abstract

The application provides a new energy sanitation vehicle cleaning control method and system fusing vision and radar, electronic equipment and storage medium, relates to the new energy vehicle control technical field, and through acquiring the water accumulation characteristic data of the urban secondary trunk road in the rainy day scene, the working section image and the ground distance data, generating the road surface reflection characteristic data and the ground distance correction data, identifying the water accumulation area through the multi-feature fusion identification model, and detecting the actual water accumulation parameters of the area, to generate the water accumulation area avoidance instruction and the adjusted discharge parameter; the cleaning execution component is controlled to clean according to the water accumulation area avoidance instruction and the adjusted discharge parameter, and the load data of the cleaning execution component is collected, if the preset load range is exceeded, the adjusted discharge parameter is optimized until it is within the preset load range, the autonomous cleaning of the urban secondary trunk road in the rainy day scene is completed, and the autonomous operation safety, energy efficiency utilization rate and cleaning stability of the new energy sanitation vehicle in the rainy day environment are improved.

Description

Technical Field

[0001] This application relates to the field of new energy vehicle control technology, and in particular to a new energy sanitation vehicle cleaning control method, system, electronic device and storage medium that integrates vision and radar. Background Technology

[0002] In rainy weather conditions on urban secondary roads, new energy sanitation vehicles face complex road conditions, especially the presence of waterlogged areas. This not only affects the efficiency and safety of cleaning operations but also poses potential risks to the vehicle's power system and battery operation. Due to poor lighting conditions and severe road surface reflection in rainy weather, traditional sensing systems that rely on a single sensor struggle to accurately identify slippery or waterlogged areas, leading to unreasonable cleaning path planning and even situations where vehicles wade through too much water, causing malfunctions.

[0003] Current research has proposed water accumulation identification methods that combine monocular vision with fixed threshold judgment. These methods use vehicle-mounted cameras to capture road images, compare the image's grayscale statistical features with preset reflectivity thresholds to initially delineate suspected water accumulation areas, and then use slope information provided by the vehicle's inertial navigation system to estimate potential water accumulation locations, thereby guiding the cleaning device to avoid low-lying areas. However, existing solutions rely on pre-set empirical thresholds, making it difficult to adapt to changes in reflectivity due to different environments and road surface materials, easily leading to misjudgments and missed detections. Furthermore, they cannot effectively distinguish between water accumulation areas and ordinary slippery road surfaces, resulting in overly conservative or aggressive avoidance strategies. Summary of the Invention

[0004] The purpose of this application is to provide a sweeping control method, system, electronic device and storage medium for new energy sanitation vehicles that integrates vision and radar, so as to solve the problems in the prior art that are difficult to adapt to the changes in reflective characteristics caused by different environments and road materials, which easily lead to misjudgment and missed detection; and the avoidance strategy is too conservative or aggressive.

[0005] To address the aforementioned technical problems, in a first aspect, this application provides a new energy sanitation vehicle cleaning control method integrating vision and radar, comprising:

[0006] Acquire water accumulation characteristic data, images of the work section, and ground distance data of urban secondary arterial roads in rainy weather scenarios;

[0007] An adaptive median filtering algorithm is used to filter out the interference areas in the image of the work section to obtain road surface reflectivity feature data;

[0008] The ground distance data is subjected to deviation compensation processing to obtain ground distance correction data. Combined with the road surface reflective feature data, the water accumulation area is identified through a multi-feature fusion recognition model.

[0009] The depth detection component mounted on the new energy sanitation vehicle is controlled to detect the actual water accumulation parameters of the water accumulation area, and the actual water accumulation parameters are compared with the water accumulation feature data to generate a water accumulation area avoidance command to bypass the water accumulation area.

[0010] Based on the actual water accumulation parameters, adjust the discharge parameters of the vehicle battery to obtain the adjusted discharge parameters;

[0011] The cleaning execution component is controlled to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and the load data of the cleaning execution component is collected during the cleaning process. If the load data exceeds the preset load range, the adjusted discharge parameters are optimized based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

[0012] Optionally, an adaptive median filtering algorithm is used to filter out interference areas in the image of the work section to obtain road surface reflectivity feature data, including:

[0013] Based on preset division rules, the operation section image is divided into multiple operation section sub-images;

[0014] Calculate the grayscale value of all pixels in each sub-image of the work section, and adjust the filtering window of the adaptive median filtering algorithm according to the degree of difference of the grayscale values ​​to obtain the adjusted filtering window;

[0015] Based on the adjusted filtering window, the sub-image of the work section is filtered to obtain a filtered sub-image, and all filtered sub-images are stitched together in reverse according to the preset division rules to obtain a de-interference image.

[0016] Based on the grayscale characteristics of the reflective area of ​​the water accumulation on the urban secondary arterial road in a rainy weather scenario, a grayscale value range for the reflective area of ​​the water accumulation is set, and reflective pixels with grayscale values ​​within the grayscale value range are selected.

[0017] The coordinate positions of the reflective pixels in the de-interference image are statistically analyzed, and the area of ​​the image region covered by all reflective pixels is calculated. The coordinate positions and the area of ​​the image region are combined to form road surface reflective feature data.

[0018] Secondly, this application provides a sweeping control method, system, electronic device, and storage medium for new energy sanitation vehicles that integrates vision and radar, including:

[0019] The acquisition module is used to acquire water accumulation characteristic data, images of the work section, and ground distance data of urban secondary arterial roads in rainy weather scenarios.

[0020] The filtering module is used to filter out interference areas in the image of the work section using an adaptive median filtering algorithm to obtain road surface reflective feature data.

[0021] The compensation module is used to perform deviation compensation processing on the ground distance data to obtain ground distance correction data. Combined with the road surface reflective feature data, the water accumulation area is identified through a multi-feature fusion recognition model.

[0022] The generation module is used to control the depth detection component mounted on the new energy sanitation vehicle to detect the actual water accumulation parameters of the water accumulation area, compare the actual water accumulation parameters with the water accumulation feature data, and generate a water accumulation area avoidance command to bypass the water accumulation area.

[0023] The adjustment module is used to adjust the discharge parameters of the vehicle battery according to the actual water accumulation parameters, so as to obtain the adjusted discharge parameters.

[0024] The cleaning module is used to control the cleaning execution component to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and to collect the load data of the cleaning execution component during the cleaning process. If the load data exceeds the preset load range, the adjusted discharge parameters are optimized based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

[0025] Thirdly, this application provides an electronic device, comprising:

[0026] Memory, used to store computer programs;

[0027] A processor is configured to execute the computer program to implement the steps of a new energy sanitation vehicle cleaning control method integrating vision and radar as described in the first aspect above.

[0028] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the new energy sanitation vehicle cleaning control method integrating vision and radar as described in the first aspect above.

[0029] This application provides a new energy sanitation vehicle cleaning control method that integrates vision and radar. By acquiring water accumulation feature data, working section images, and ground distance data in rainy weather scenarios, and employing an adaptive median filtering algorithm to filter image interference areas, it effectively suppresses strong glare and noise interference in rainy weather, improves the accuracy of road surface reflectivity feature extraction, and overcomes the misjudgment problem caused by environmental changes in fixed threshold methods. By performing deviation compensation processing on ground distance data and combining it with road surface reflectivity feature data, a multi-feature fusion recognition model is used to identify water accumulation areas. This achieves collaborative perception of visual and radar information, enhances the ability to identify real water accumulation areas, and avoids missed detections caused by relying solely on image grayscale or slope estimation. To address excessive avoidance, the system utilizes depth detection components to detect actual parameters of identified waterlogged areas and compares these parameters with initial identification results to generate avoidance commands. This enhances the reliability and safety of path planning, preventing vehicles from accidentally entering deep water. The system dynamically adjusts battery discharge strategies based on actual water parameters, ensuring power output better matches the actual workload and improving energy efficiency. Real-time load data collection and feedback optimization of discharge parameters during cleaning operations create a closed-loop control mechanism of perception-decision-execution-feedback. This prevents motor overload or energy waste, ensuring the stability and endurance of cleaning operations and enabling more precise, intelligent, and reliable autonomous cleaning control for new energy sanitation vehicles in complex rainy conditions. Furthermore, by acquiring radar detection parameters and rainy environmental parameters, a radar detection area coordinate system is established, and real-time air refractive index is calculated. This allows for deviation compensation of the original distance values ​​at each radar detection point, generating accurate ground distance correction data. This correction data is then correlated with the coordinates of reflective areas extracted from images. By fusing reflective coverage features with corrected ground height variation features, and using a multi-feature fusion model based on a joint criterion of area and distance differences, waterlogged areas are identified. It solves the problem that traditional monocular vision methods have difficulty distinguishing between water accumulation and ordinary slippery road surfaces due to the lack of real ground elevation information. At the same time, it overcomes the impact of environmental refractive index changes on radar ranging accuracy, improves the spatial alignment accuracy and fusion reliability of multi-sensor data, so that water accumulation area identification no longer relies on empirical thresholds, has stronger environmental adaptability and recognition robustness, and improves path planning errors and operational risks caused by inaccurate perception. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1A flowchart illustrating a new energy sanitation vehicle cleaning control method integrating vision and radar, provided as an embodiment of this application;

[0032] Figure 2 This is a schematic diagram of a new energy sanitation vehicle cleaning control system that integrates vision and radar, provided as an embodiment of this application. Detailed Implementation

[0033] To address the issues of inaccurate water accumulation identification during urban secondary road operations in rainy weather due to changes in lighting and road surface reflection, and the shortcomings of single-vision solutions relying on fixed thresholds that are prone to misjudgment and missed detection, this application integrates visual and radar data, utilizes adaptive filtering technology to suppress image interference, extracts reliable road surface reflection features, compensates for environmental parameters in radar ranging data to improve ground height measurement accuracy, and identifies water accumulation areas through a multi-feature fusion recognition model based on both, achieving precise positioning of water accumulation areas. Furthermore, a depth detection component is introduced to perform secondary verification of the identification results, generating safe and reliable avoidance path instructions, and dynamically adjusting the battery discharge strategy according to the actual water accumulation situation to match the power output with the actual operating load. Load data of the cleaning execution components during the cleaning process is collected, and a load feedback mechanism is established to optimize the discharge parameters in a closed loop, avoiding motor overload or energy waste. This improves the operational safety, energy efficiency, and cleaning stability of new energy sanitation vehicles in complex rainy environments.

[0034] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0035] The core of this application is to provide a sweeping control method for new energy sanitation vehicles that integrates vision and radar. A flowchart of one specific implementation method is shown below. Figure 1 As shown, the method includes:

[0036] Step 101: Obtain water accumulation characteristic data, operation section images, and ground distance data of urban secondary arterial roads in rainy weather scenarios.

[0037] In this step, the rainy weather scenario refers to rainfall exceeding 0.1 mm / h, resulting in significant water accumulation, glare, or slipperiness on urban secondary roads, reflecting the characteristics of water accumulation interfering with sanitation operations. Water accumulation characteristic data refers to reference data related to water accumulation obtained through statistical analysis of historical rainy weather operation data for urban secondary roads, including the average range of water depth, water accumulation time thresholds, and baseline values ​​for the proportion of water-affected areas. The operation section image refers to real-time images of the current urban secondary road surface captured by high-definition vehicle-mounted cameras on new energy sanitation vehicles, reflecting the visual characteristics of the road surface (such as areas of water accumulation and glare, areas of debris interference, and ordinary slippery areas). Ground distance data refers to real-time distance data between various points on the ground and radar detection points on the current operation section, collected by radar detection devices on new energy sanitation vehicles, reflecting the ground's height and spatial distribution characteristics.

[0038] In this embodiment, the target of the new energy sanitation vehicle is first determined to be a secondary arterial road in the city. Considering that this road is prone to water accumulation in rainy weather, the required basic data is obtained through multi-source data acquisition: water accumulation feature data is obtained by retrieving the historical rainy water accumulation database of the secondary arterial road in the city; the image of the working section is collected in real time by the vehicle-mounted high-definition camera; and the ground distance data is collected in real time by the vehicle-mounted radar detection device to collect the ground height and distance information of the current working section, providing radar dimension data support for subsequent water accumulation identification.

[0039] Step 102: Use an adaptive median filtering algorithm to filter out the interference areas in the image of the work section to obtain road surface reflective feature data.

[0040] In this step, the interference region refers to the area in the image of the work section that affects the extraction of water accumulation reflection features. This includes non-water accumulation reflections on the road surface during rainy weather (such as street light reflections), shadows of road debris, and vehicle reflections. The grayscale features of these areas are easily confused with water accumulation reflections and need to be filtered using an adaptive median filtering algorithm to avoid misjudgment. Road surface reflection feature data refers to the feature data related to water accumulation reflections extracted from the image after interference removal. This includes the coordinate positions of reflective pixels in the image and the area of ​​the image region covered by all reflective pixels, reflecting the spatial location and scale of the road surface reflection area, and serving as the visual input features for the multi-feature fusion recognition model.

[0041] Step 103: Perform deviation compensation processing on the ground distance data to obtain ground distance correction data. Combine the road surface reflective feature data and identify the water accumulation area through a multi-feature fusion recognition model.

[0042] In this step, ground distance correction data refers to the accurate distance data obtained after bias compensation processing of the ground distance data. It consists of the corrected distance values ​​of all radar detection points and their coordinates in the radar detection area coordinate system, reflecting the true distance characteristics of each point on the ground and serving as the radar input feature for the multi-feature fusion recognition model. Waterlogged areas refer to areas on urban secondary roads where water accumulates during rainy weather. This is identified through collaborative analysis of road surface reflectivity data and ground distance correction data using the multi-feature fusion recognition model.

[0043] Step 104: Control the depth detection component on the new energy sanitation vehicle to detect the actual water accumulation parameters of the water accumulation area, compare the actual water accumulation parameters with the water accumulation feature data, and generate a water accumulation area avoidance command to bypass the water accumulation area.

[0044] In this step, the actual water accumulation parameters refer to the specific parameters obtained in real time through the depth detection component, including the actual water depth, actual water accumulation time, actual water accumulation area percentage, and water accumulation distribution density, reflecting the true state of the current water accumulation area. The water accumulation area avoidance instruction is an operational command generated based on the comparison results of the actual water accumulation parameters and water accumulation characteristic data, used to guide new energy sanitation vehicles to avoid water accumulation areas, ensuring that vehicles avoid water accumulation areas.

[0045] Step 105: Adjust the discharge parameters of the vehicle battery according to the actual water accumulation parameters to obtain the adjusted discharge parameters.

[0046] In this step, the discharge parameters of the vehicle battery refer to the core parameters when the vehicle power battery outputs electrical energy to the sweeping execution components and drive system. These values ​​need to be adjusted according to the actual water accumulation parameters to adapt to changes in sweeping load during rainy weather. The adjusted discharge parameters refer to the vehicle battery discharge parameters obtained after adjustment based on the actual water accumulation parameters and the basic adjustment strategy. These parameters are then verified against the rated input parameters of the vehicle motor to ensure compatibility with the load requirements of the sweeping execution components.

[0047] Step 106: Control the cleaning execution component to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and collect the load data of the cleaning execution component during the cleaning process. If the load data exceeds the preset load range, optimize the adjusted discharge parameters based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

[0048] In this step, load data refers to load-related data generated by the cleaning actuator during operation, including the torque of the cleaning brush and the operating current of the drive motor. This reflects the current workload of the cleaning actuator and is used to determine whether the discharge parameters are suitable for actual operational needs. The preset load range refers to the safe load range set based on the rated performance of the cleaning actuator (such as the rated torque and rated current of the motor). This range serves as a benchmark for judging whether the load data is normal; if it exceeds the range, the discharge parameters need to be optimized and adjusted.

[0049] In this embodiment, the sweeping execution components (such as sweeping brushes and vacuuming devices) of the new energy sanitation vehicle are controlled to perform sweeping operations according to the water accumulation area avoidance command (driving along the avoidance path) and the adjusted discharge parameters (outputting according to the appropriate power). During the sweeping process, load data (such as the torque of the sweeping brush and the current load of the motor) are collected in real time by the load sensor mounted on the sweeping execution component. The collected load data is compared with the preset load range (set according to the rated load of the sweeping execution component, such as torque 50-80 N·m). If the load data exceeds the range (such as torque exceeding 80 N·m), the discharge parameters are optimized and adjusted based on the actual water accumulation parameters (such as the discharge parameters are further adjusted to reduce the output power if the water depth increases, resulting in a higher load). This optimization process is repeated until the load data is within the preset load range, and finally the autonomous sweeping of the urban secondary arterial road in rainy weather is completed.

[0050] The embodiments of this application can accurately distinguish between waterlogged areas and ordinary slippery road surfaces, and generate waterlogged area avoidance instructions, avoiding vehicle malfunctions caused by wading through too deep water, optimizing the safety of cleaning path planning; ensuring cleaning quality, improving energy utilization efficiency, and enhancing the autonomous operation capability and range stability of new energy sanitation vehicles in rainy weather scenarios on urban secondary roads.

[0051] This application provides a specific embodiment. Step 102 involves using an adaptive median filtering algorithm to filter out interference areas in the image of the work section to obtain road surface reflectivity feature data. This specifically includes the following steps:

[0052] Step 201: Based on the preset division rules, divide the operation section image into multiple operation section sub-images.

[0053] In this step, the preset segmentation rules refer to the rules pre-defined for segmenting the work section image. These rules are determined based on the resolution of the work section image (e.g., 1920×1080), the efficiency of local interference processing, and the correlation of road surface areas. Fixed pixel blocks (e.g., 200×200 pixels) or road surface functional partitioning (e.g., by lane width) can be used to divide the entire work section image into multiple sub-images that can be processed independently, adapting the filtering strategy to the interference levels of different areas. The work section sub-image refers to a small local image obtained after segmenting the work section image according to the preset segmentation rules. It reflects the visual characteristics of a specific local area of ​​the work section (e.g., local road surface reflections, object shadows), and is used to calculate grayscale values, adjust the filtering window, and perform filtering processing independently, avoiding the inaccurate interference filtering problem caused by uniform processing of the entire image.

[0054] In this embodiment of the application, a preset division rule for segmenting the operation road segment image is determined, such as a fixed size of 200×200 pixels or a division according to road surface functional zoning (such as lane line spacing). Based on the preset division rule, the operation road segment image is divided into multiple independent operation road segment sub-images. Each operation road segment sub-image corresponds to a local area of ​​the operation road segment, which facilitates subsequent adjustment of the filtering strategy for the interference level of different areas.

[0055] Step 202: Calculate the grayscale value of all pixels in the sub-image of each work section, and adjust the filtering window of the adaptive median filtering algorithm according to the degree of difference of the grayscale values ​​to obtain the adjusted filtering window.

[0056] In this step, the degree of difference in grayscale values ​​refers to the dispersion of the grayscale values ​​of all pixels within the sub-image of the work section. It is usually obtained by calculating the standard deviation of the grayscale values, reflecting the amount of interference areas within the sub-image. A large degree of difference indicates the presence of more interference such as reflections and shadows, while a small degree of difference indicates more uniform visual features in the area. The filtering window refers to the rectangular area used to cover and process image pixels in the adaptive median filtering algorithm. Its size is set in pixels, and the grayscale values ​​of the pixels within the window are used to calculate the median to replace the interference pixels in the center of the window. It is the core processing unit for filtering interference areas. The adjusted filtering window refers to the filtering window whose size is dynamically adjusted according to the degree of difference in grayscale values ​​of the sub-image of the work section. When the degree of difference in grayscale values ​​is large, the window size is increased to cover more interference pixels; when the degree of difference is small, the window size is decreased to retain details of water accumulation and reflections. This is used to adapt to the interference situation of different sub-images to achieve accurate filtering.

[0057] In this embodiment, for each sub-image of the work section, the grayscale value of all pixels in each sub-image of the work section is calculated using an image grayscale extraction method (such as reading the values ​​of the red, green, and blue channels pixel by pixel, and calculating the grayscale value as 0.299 × red channel value + 0.587 × green channel value + 0.114 × blue channel value, or directly calling the image grayscale interface). Then, by calculating the degree of difference in grayscale values, such as calculating the standard deviation of the grayscale values ​​of all pixels in each sub-image of the work section, the larger the standard deviation, the more dispersed the pixel grayscale distribution and the more interference areas in the sub-image. The filtering window of the adaptive median filtering algorithm is dynamically adjusted. If the degree of difference in grayscale values ​​is large (such as the standard deviation being greater than 50), the filtering window is enlarged (such as from 3×3 to 5×5) to cover more interference pixels; if the degree of difference in grayscale values ​​is small (such as the standard deviation being less than 20), the filtering window is reduced (such as keeping it at 3×3) to avoid over-filtering and losing the details of water accumulation and reflection. Finally, an adjusted filtering window adapted to each sub-image of the work section is obtained.

[0058] Step 203: Based on the adjusted filtering window, filter the sub-image of the work section to obtain the filtered sub-image, and then stitch all the filtered sub-images together in reverse according to the preset division rules to obtain the interference-free image.

[0059] In this step, the filtered sub-image refers to the sub-image obtained after applying adaptive median filtering to the sub-image of the work section using an adjusted filtering window. Median replacement removes interference pixels such as reflective interference and shadows from debris, while preserving the true visual characteristics of the road surface (such as reflective pixels from accumulated water). It is the basic unit for stitching together the de-interference image. The de-interference image refers to the complete image obtained by reverse stitching all the filtered sub-images according to a preset division rule. It removes most of the interference areas in the work section image (such as non-accumulated water reflection and shadows), and the pixel grayscale values ​​within the image better reflect the true state of the road surface (such as reflected water and ordinary slippery road surface), which is used for subsequent screening of reflected water pixels.

[0060] In this embodiment, the adjusted filtering window is used as the processing unit to perform adaptive median filtering on the corresponding work section sub-image: the adjusted filtering window traverses each pixel in each work section sub-image, sorts the gray values ​​of all pixels in the adjusted filtering window from smallest to largest, and takes the median value after sorting to replace the pixel value in the center of the window (if the center pixel is an interference pixel, the median value can effectively replace it to remove interference), thus completing the filtering process of a single work section sub-image and obtaining the filtered sub-image; after all work section sub-images have been filtered, all filtered sub-images are reverse-stitched according to a preset division rule (such as first stitching by row, then by column, to ensure that the stitching position is consistent with the original image) to form a complete, interference-free image that has removed reflection interference and shadow interference.

[0061] Step 204: Based on the grayscale characteristics of the water accumulation and reflective area of ​​the urban secondary arterial road in a rainy weather scenario, set the grayscale value range of the water accumulation and reflective area, and filter out reflective pixels whose grayscale values ​​are within the grayscale value range.

[0062] In this step, the reflective area of ​​accumulated water refers to the area on a secondary urban road where water has accumulated on the surface during rainy weather, and this area is formed by light reflection. Its visual characteristic is that its grayscale value is significantly higher than that of ordinary slippery road surfaces. Grayscale features refer to the inherent properties of the reflective area of ​​accumulated water in the grayscale dimension, obtained through statistical analysis of historical images of accumulated water on secondary urban roads during rainy weather. These features are manifested in the grayscale value range and the uniformity of grayscale distribution (usually more uniform than that of ordinary slippery road surfaces). The grayscale value range refers to the grayscale interval set based on the grayscale characteristics of the reflective area of ​​accumulated water, used to filter reflective pixels. This range can be fine-tuned according to the real-time light intensity of the current working scene (lowering the interval when the light is weak and raising it when the light is strong) to ensure that only pixels with grayscale values ​​within this range are marked as reflective pixels, reducing misclassification of non-accumulated water pixels. Reflective pixels refer to pixels in the image after interference removal whose gray values ​​fall within the gray value range of the water reflection area. These pixels initially correspond to the water reflection area of ​​the road surface (excluding non-water reflection, shadow and other interfering pixels), providing a basis for forming road surface reflection feature data.

[0063] In this embodiment, based on the grayscale characteristics of the reflective areas of water accumulation on urban secondary roads in rainy weather, a grayscale value range for the reflective areas of water accumulation adapted to the current scene is set. For example, it is finely adjusted according to the real-time light intensity, with the range lowered when the light is weak and kept constant when the light is strong. Then, all pixels in the image after interference removal are traversed, and each pixel's grayscale value is determined to be within the set grayscale value range. Pixels with grayscale values ​​within this range are marked as reflective pixels. These reflective pixels initially correspond to areas on the road surface where water accumulation may occur, providing core pixel objects for subsequent extraction of road surface reflective feature data.

[0064] Step 205: Calculate the coordinate positions of the reflective pixels in the de-interference image, and calculate the area of ​​the image region covered by all reflective pixels. Combine the coordinate positions and the area of ​​the image region to form road surface reflective feature data.

[0065] In this step, the image area refers to the area of ​​the road surface covered by all reflective pixels in the image, reflecting the size of the water accumulation and reflective area in the image, and is an important component of the road surface reflective feature data.

[0066] In this embodiment, for all reflective pixels, an image coordinate positioning method is used, such as taking the upper left corner of the image after interference removal as the origin, the horizontal axis as the x-axis and the vertical axis as the y-axis, and recording the (x, y) coordinates of each reflective pixel to statistically determine its coordinate position in the image after interference removal; at the same time, the area of ​​the image region covered by all reflective pixels is calculated: first, the boundary coordinates of all reflective pixels (i.e., maximum x value, minimum x value, maximum y value, minimum y value) are determined, and the area of ​​the rectangular region is calculated using the formula (maximum x value - minimum x value) × (maximum y value - minimum y value). This area is the area of ​​the image region covered by all reflective pixels; finally, the statistically obtained reflective pixel coordinate positions and the calculated image region area are integrated to form road surface reflective feature data, which serves as the visual dimension input of the multi-feature fusion recognition model.

[0067] This application's embodiments prevent the loss of water accumulation details due to over-filtering in areas with less interference, thus improving the clarity of the image after interference removal; it can accurately filter out reflective pixels and eliminate the influence of non-water accumulation reflective pixels, shadows, and other interfering pixels; the road surface reflective feature data provides high-quality visual dimension input for the multi-feature fusion recognition model, and combined with radar data, it can further distinguish water accumulation areas from ordinary slippery roads, improving the accuracy of water accumulation recognition in rainy weather.

[0068] This application provides a specific embodiment. Step 103 involves performing deviation compensation processing on the ground distance data to obtain corrected ground distance data. Combined with the road surface reflectivity data, a multi-feature fusion recognition model is used to identify the water accumulation area. Specifically, this includes the following steps:

[0069] Step 301: Obtain the detection parameters of the radar detection device mounted on the new energy sanitation vehicle and the real-time rainy weather environmental parameters of the urban secondary arterial road in a rainy weather scenario, so as to determine the radar detection area coordinate system and the real-time air refractive index of the working area of ​​the new energy sanitation vehicle. The detection parameters include the radar detection angle range, the radar detection distance range, and the coordinate definition rules of the radar detection point.

[0070] In this step, the detection parameters refer to the inherent performance parameters and measurement rules of the radar detection device mounted on the new energy sanitation vehicle, including the radar detection angle range, radar detection distance range, and the coordinate definition rules of the radar detection point. These parameters reflect the effective detection range and data recording specifications of the radar and are used to determine the coordinate system of the radar detection area and the benchmark for subsequent data processing. Real-time rainy weather environmental parameters refer to meteorological data collected in real-time on urban secondary roads during rainy weather, including ambient temperature, atmospheric pressure, water vapor pressure, and rainfall intensity, reflecting the real-time atmospheric conditions of the work area. The radar detection area coordinate system refers to the spatial coordinate system established based on the installation location and detection parameters of the radar detection device, using the coordinate definition rules of the radar detection point as a benchmark, and is used to calibrate the spatial position of the radar detection point. Real-time air refractive index refers to the refractive characteristic parameter of the atmosphere in the work area for electromagnetic wave propagation, used to correct errors in radar detection distance caused by atmospheric refraction. The coordinate definition rules of the radar detection point refer to the rules specifying the way the radar detection point is represented in the coordinate system, including the origin setting, axis direction, and coordinate value calculation methods, used to unify the spatial positioning standards of the radar detection point.

[0071] In this embodiment, the detection parameters of the radar detection device are read through the local area network bus interface of the vehicle controller of the new energy sanitation vehicle. The radar detection angle range is the maximum horizontal angle detectable by the radar, and the radar detection distance range is the maximum forward and backward distance detectable by the radar. The coordinate definition rule for the radar detection point is as follows: the installation point of the radar on the chassis of the sanitation vehicle is the origin of the coordinate system, the horizontal rightward direction (to the right of the sanitation vehicle's forward direction) is the positive x-axis, and the vertical forward direction (towards the sanitation vehicle's forward direction) is the positive y-axis. The coordinate unit is meters. Simultaneously, real-time environmental parameters of the work area in rainy weather are collected through the vehicle-mounted environmental sensor, specifically including ambient temperature, atmospheric pressure, and water vapor pressure. Then, the coordinates of the radar detection point are... Based on the defined rules, the coordinates of each detection point are determined by combining the radar detection angle range and distance range. For example, if the detection angle of a certain detection point is 30 degrees and the detection distance is 10 meters, its x and y coordinates are calculated using trigonometric functions. The coordinate positions of all detection points are delineated in this way, and then a radar detection area coordinate system covering the operation area is established. Then, the real-time rainy weather environmental parameters are substituted into the Hopfield model to calculate the real-time air refractive index of the operation area. In this model, the dry refractive index is calculated from atmospheric pressure and temperature, and the wet refractive index is calculated from water vapor pressure and temperature. The total real-time air refractive index is the sum of the dry refractive index and the wet refractive index. Finally, the real-time air refractive index of the operation area is calculated.

[0072] Step 302: Calculate the distance deviation value of each radar detection point based on the real-time air refractive index, extract the original detection distance value of each radar detection point from the ground distance data, and calculate the corrected distance value of each radar detection point by combining the corresponding distance deviation value.

[0073] In this step, the distance deviation value refers to the error in radar detection range caused by atmospheric refraction, used to correct the original detection range to obtain the true distance. The original detection range value refers to the initial distance data from each radar detection point to the ground target directly measured by the radar detection device. It reflects the distance information without correction for atmospheric refraction errors and is the basis for calculating the corrected distance value. The corrected distance value refers to the accurate distance data after correcting for atmospheric refraction errors in the original detection range value, used to accurately reflect the true distance from the radar detection point to the ground.

[0074] In this embodiment, the distance deviation value of each radar detection point is calculated based on the difference between the real-time air refractive index and the standard air refractive index (1.00030 under standard conditions), combined with the radar detection principle. The specific calculation formula is: Distance deviation value = Original detection distance value × (Standard air refractive index - Real-time air refractive index) / Standard air refractive index. For example, if the original detection distance value of a radar detection point is 10.5 meters, substituting this into the formula, the distance deviation value is: 10.5 meters × (1.00030 - 1.00025) / 1.00030 ≈ 0.000 525 meters; then, from the ground distance data output by the radar detection device through the data interface, according to the correspondence between the x-coordinate and y-coordinate of the detection point, the original detection distance value of each radar detection point is extracted one by one (e.g., the original value of the detection point at coordinates (5, 8.66) is 10.5 meters); finally, each original detection distance value is subtracted from the corresponding distance deviation value to obtain the corrected distance value of each radar detection point (e.g., 10.5 meters - 0.000525 meters ≈ 10.499475 meters), thereby eliminating the distance measurement error caused by atmospheric refraction in rainy weather.

[0075] Step 303: Integrate the corrected distance values ​​of all radar detection points and the corresponding coordinates of the detection points in the radar detection area coordinate system into ground distance correction data.

[0076] In this step, the detection point coordinates refer to the spatial location data of the radar detection point in the radar detection area coordinate system, which is used to establish a correlation with the coordinates of the reflective pixels.

[0077] In this embodiment, the corrected distance values ​​of all radar detection points are associated one by one with their corresponding x-coordinates and y-coordinates in the radar detection area coordinate system. The data are arranged in ascending order of x-coordinates and ascending order of y-coordinates under the same x-coordinate, forming structured ground distance correction data. The specific form is a table with columns named x-coordinate of detection point (meters), y-coordinate of detection point (meters), and corrected distance value (meters).

[0078] Step 304: Extract the coordinates of reflective pixels and the coverage area of ​​reflective pixels from the road surface reflective feature data.

[0079] In this step, the reflective pixel coordinates refer to the positional data of pixels in the image whose grayscale values ​​fall within the grayscale range of the water accumulation reflective area, extracted from the road surface reflective feature data. This reflects the pixel-level spatial location of potential water accumulation on the road surface and is used for matching with radar detection point coordinates. The reflective pixel coverage area refers to the image region composed of multiple consecutive reflective pixels, reflecting the image range of the suspected water accumulation area on the road surface and serving as the visual feature object for water accumulation identification.

[0080] In this embodiment of the application, the coordinates of reflective pixels and the coverage area of ​​reflective pixels are separated from the road surface reflective feature data, thereby clarifying the suspected water accumulation range in the visual dimension.

[0081] Step 305: Associate the coordinates of the reflective pixels with the coordinates of the detection points in the ground distance correction data to determine the set of radar detection points corresponding to the coverage area of ​​each reflective pixel, and calculate the average corrected distance within each set of radar detection points.

[0082] In this step, the radar detection point set refers to the set of all radar detection points spatially corresponding to the coverage area of ​​a certain reflective pixel. It is obtained by transforming the coordinates of the reflective pixels to the radar detection area coordinate system and matching them with the coordinates of the detection points, and is used to extract the radar range features corresponding to that area. The corrected mean distance refers to the average of the corrected distance values ​​of all radar detection points within the radar detection point set, reflecting the average ground distance of the area corresponding to that set.

[0083] In this embodiment, the coordinates of the reflective pixels are first converted to the radar detection area coordinate system according to the mapping relationship between the image and the radar coordinate system. The mapping relationship is calculated based on the intrinsic parameters of the camera (such as a focal length of 8 mm and a pixel size of 3 micrometers) and the extrinsic parameters (i.e., the relative position between the camera and the radar, such as an offset of 0.5 meters in the x-direction, 1 meter in the y-direction, and 0.3 meters in the vertical direction). For example, the process of converting the image coordinates (200, 300) to radar coordinates is as follows: first, the image coordinates are converted to three-dimensional coordinates in the camera coordinate system through the intrinsic parameters, and then converted to coordinates in the radar detection area coordinate system through the extrinsic parameters, finally obtaining the radar coordinates (5, 10). Then, the set of radar detection points corresponding to each reflective pixel coverage area is matched: that is, all radar detection points in the radar detection area coordinate system whose coordinates are within the converted range of the reflective pixel coverage area are selected to form the radar detection point set of the area. Then, the corrected distance values ​​in each set are summed and divided by the number of detection points to obtain the average corrected distance corresponding to each reflective pixel coverage area.

[0084] Step 306: Based on the reflective pixel coverage area and the corresponding corrected average distance value, combined with the area threshold and the distance difference threshold, the water accumulation area is identified through a multi-feature fusion recognition model.

[0085] In this step, the area threshold refers to the critical value of the reflective pixel coverage area used to filter candidate waterlogged areas. The threshold is lowered when the light is weak and raised when the light is strong to filter out interference areas that are too small. The distance difference threshold refers to the critical value of the radar distance difference used to distinguish between waterlogged areas and dry road surfaces. Different materials correspond to different thresholds to adapt to the differences in road surface reflectivity.

[0086] Optionally, step 306 involves identifying the water accumulation area based on the reflective pixel coverage area and the corresponding corrected average distance value, combined with an area threshold and a distance difference threshold, using a multi-feature fusion recognition model. This specifically includes the following steps:

[0087] Step 311: Obtain historical water accumulation identification data of the urban secondary arterial road in rainy weather and real-time light intensity of the work area. Based on the comparison results between the real-time light intensity and the historical average light value in the historical water accumulation identification data, determine the area threshold.

[0088] In this step, historical waterlogging identification data refers to the collection of data on the location, area, radar distance, and corresponding environmental parameters (such as light intensity and rainfall) of waterlogged areas recorded on urban secondary roads in past rainy weather scenarios. This data reflects the characteristics and identification results of waterlogged areas under different rainy weather conditions and is used to determine the identification threshold and compare morphological features.

[0089] Step 312: Based on the distance difference range between historical dry road surfaces and historical waterlogged areas in the historical waterlogging identification data, select a distance difference threshold that is suitable for the road surface material of the work area from the distance difference range.

[0090] In this step, "historically dry road surface" refers to road areas recorded in the historical water accumulation identification data that did not experience water accumulation. The corresponding radar distance data reflects the radar reflection characteristics of the road surface in a dry state and is used to compare with the distance data of historically water-accumulated areas to obtain the distance difference range. "Historically water-accumulated area" refers to road areas recorded in the historical water accumulation identification data that were confirmed to have water accumulation. The corresponding radar distance data reflects the radar reflection characteristics of the road surface in a water-accumulated state and is compared with the distance data of historically dry road surfaces to obtain the distance difference range. The distance difference range refers to the interval between the radar distance values ​​of historically dry road surfaces and historically water-accumulated areas, reflecting the typical range of radar distance changes when water accumulation is present. It is used to select a distance difference threshold suitable for the current road surface material. The road surface material of the work area refers to the paving material properties of the work section, such as asphalt or cement. Different materials have different dielectric constants and reflection characteristics, affecting the radar distance difference. It is used to select a suitable distance difference threshold from the distance difference range.

[0091] Step 313: Using a multi-feature fusion recognition model, reflective pixel coverage areas with a pixel coverage area greater than or equal to an area threshold, or edge pixel grayscale change rate greater than a preset grayscale change threshold, are marked as candidate water accumulation areas.

[0092] In this step, the pixel coverage area refers to the image area corresponding to the total number of pixels contained in the reflective pixel coverage area, reflecting the size of the reflective area and used to filter candidate water accumulation areas. The grayscale change rate of edge pixels refers to the rate of change of the grayscale value of pixels at the edge of the reflective pixel coverage area, reflecting the clarity of the area boundary and used to assist in filtering candidate water accumulation areas. The preset grayscale change threshold refers to a pre-set critical value used to judge the edge features of the reflective pixel coverage area. It is obtained based on the statistical analysis of the edge grayscale features of historical reflective areas and real water accumulation areas and is used to distinguish real water accumulation areas from interfering reflective areas. Candidate water accumulation areas refer to the areas suspected of water accumulation initially screened by the multi-feature fusion recognition model, serving as intermediate objects for further identification of water accumulation areas.

[0093] Step 314: Obtain multiple dry road surface detection distance values ​​of the radar detection device in the work area, and determine the dry road surface distance value by combining the ground structure parameters of the work area.

[0094] In this step, the dry road surface detection distance value refers to the distance data obtained by the radar detection device from detecting dry road surfaces within the work area that are confirmed to be free of water accumulation. This is obtained by selecting the radar detection point distance values ​​corresponding to areas within the work area that do not exhibit reflective characteristics, and serves as the baseline data for determining the dry road surface distance value. The ground structure parameters of the work area refer to the physical structural properties of the road surface in the work section, such as road surface smoothness, the presence and magnitude of slope, etc., reflecting the spatial morphological characteristics of the ground. These parameters are used to correct and integrate multiple dry road surface detection distance values ​​to determine the accurate dry road surface distance value. The dry road surface distance value refers to the standard distance data from the road surface to the radar detection device in a dry state within the work area, used to calculate the distance difference to determine the presence of water accumulation.

[0095] Step 315: Calculate the difference between the mean corrected distance value and the actual distance value of the dry road surface for each candidate waterlogged area, and the average difference between the corresponding corrected distance value and the distance value of the dry road surface.

[0096] In this step, the actual distance difference refers to the difference between the mean corrected distance value corresponding to each candidate waterlogged area and the distance value to the dry road surface, reflecting the degree of distance difference between the area and the dry road surface. The average difference refers to the average of the differences between all corrected distance values ​​and the distance values ​​to the dry road surface corresponding to each candidate waterlogged area, reflecting the overall level of distance difference in that area.

[0097] Step 316: Mark the candidate water accumulation areas whose actual distance difference is greater than the distance difference threshold and whose average difference is greater than or equal to the preset difference threshold as suspected water accumulation areas.

[0098] In this step, the preset difference threshold refers to a pre-set critical value used to determine whether the distance difference between candidate waterlogged areas is significant, and is used to help mark suspected waterlogged areas. Suspected waterlogged areas refer to areas that are highly suspected of having waterlogging, as identified through distance difference filtering, and are the objects of area merging and final confirmation.

[0099] Step 317: Merge any two suspected waterlogged areas that meet the region merging rules of the multi-feature fusion recognition model into a waterlogged area to be determined, and confirm the waterlogged area by combining the historical morphological characteristics of historical waterlogged areas.

[0100] In this step, the region merging rules of the multi-feature fusion recognition model refer to the conditional rules for determining whether two suspected waterlogged areas need to be merged. These rules include center-to-center distance conditions and spatial positional relationship conditions, used to merge adjacent or overlapping interfering small areas into a complete area, improving recognition accuracy. The waterlogged area to be determined refers to the complete area obtained after performing the merging operation on the suspected waterlogged areas. By repeatedly applying the region merging rules until no area meets the conditions, the fragmentation problem of suspected areas is resolved, providing objects for final confirmation. The historical morphological characteristics of historical waterlogged areas refer to the inherent spatial morphological attributes of waterlogged areas extracted from historical data, including the ratio of the major and minor axes of the outline, the actual coverage area, and the ratio of the area of ​​the circumscribed rectangle, reflecting the typical shape and scale characteristics of the waterlogged area, used for the final confirmation of the waterlogged area.

[0101] Optionally, step 317 involves merging any two suspected waterlogged areas that satisfy the region merging rules of the multi-feature fusion recognition model into a waterlogged area to be determined, and confirming the waterlogged area by combining the historical morphological characteristics of historical waterlogged areas. This specifically includes the following steps:

[0102] Step 321: Based on the first coordinate range and center coordinates of each suspected water accumulation area in the radar detection area coordinate system, calculate the center distance between each suspected water accumulation area and any other suspected water accumulation area.

[0103] In this step, the first coordinate range refers to the spatial extent occupied by the suspected waterlogged area in the radar detection area coordinate system. It is determined by the maximum and minimum values ​​of the coordinates of all detection points within the area, reflecting the spatial boundary of the suspected area. The center coordinates of each suspected waterlogged area refer to the position data of the geometric center of the suspected waterlogged area in the radar detection area coordinate system, used to calculate the center-to-center distance between areas. The center-to-center distance refers to the straight-line distance between the center coordinates of two suspected waterlogged areas, used to determine whether the areas meet the distance conditions for merging.

[0104] Step 322: According to the region merging rules of the multi-feature fusion recognition model, merge all suspected water accumulation areas until there are no suspected water accumulation areas that meet the region merging rules, and obtain the water accumulation area to be determined. The region merging rule is that the center distance between each suspected water accumulation area and any other suspected water accumulation area is less than a preset merging distance threshold, and the first coordinate range overlaps or the edge distance is less than the preset edge distance.

[0105] In this step, the preset merging distance threshold refers to a pre-set critical distance value used to determine whether suspected waterlogged areas are adjacent. It is obtained based on statistical analysis of the typical distribution density of historical waterlogged areas and serves as the standard for judging the center-to-center distance between areas for merging. Overlap refers to the partial spatial overlap of the first coordinate ranges of two suspected waterlogged areas within the radar detection area coordinate system. This is determined by comparing whether the coordinate ranges of the two areas intersect and is one of the spatial location conditions for area merging. Edge spacing refers to the shortest distance between the boundaries of two non-overlapping suspected waterlogged areas, reflecting the degree of proximity between the areas. The preset edge spacing refers to a pre-set critical boundary distance value used to determine whether suspected waterlogged areas are sufficiently close, and is one of the spatial location conditions for area merging.

[0106] Step 323: Extract historical morphological features of historical waterlogged areas that match the work area from the historical waterlogged area morphology database. The historical morphological features include the ratio of the major and minor axes of the outline, the actual coverage area, and the ratio of the area of ​​the circumscribed rectangle.

[0107] In this step, the historical waterlogged area morphology database refers to a database that stores the morphological characteristics of historical waterlogged areas under different operating areas and different rainy weather conditions. It is established based on long-term accumulated waterlogging identification data and morphological parameter extraction, and is used to provide the historical morphological features required for comparison. The profile major-minor axis ratio refers to the ratio of the longest axis length to the shortest axis length of the waterlogged area's profile, reflecting the degree of shape deviation of the waterlogged area. The actual coverage area refers to the physical area occupied by the waterlogged area on the actual road surface, reflecting the actual scale of the waterlogged area. The circumscribed rectangle area ratio refers to the ratio of the actual coverage area of ​​the waterlogged area to the area of ​​the smallest rectangle enclosing the area, reflecting how closely the shape of the waterlogged area resembles a rectangle.

[0108] Step 324: Based on the second coordinate range of each water accumulation area to be determined in the radar detection area coordinate system, calculate the actual morphological characteristics of each water accumulation area to be determined, compare the actual morphological characteristics with the historical morphological characteristics, and mark all water accumulation areas to be determined that conform to the historical morphological characteristics as water accumulation areas.

[0109] In this step, the second coordinate range refers to the spatial extent of the water accumulation area to be determined within the radar detection area coordinate system. It is determined by the maximum and minimum values ​​of the coordinates of all detection points within the area, reflecting the spatial boundary of the area to be determined. Actual morphological characteristics refer to the true spatial attributes of the water accumulation area to be determined, including the ratio of its major and minor axes, the actual coverage area, and the ratio of the area of ​​its circumscribed rectangle. These are used to compare with historical morphological characteristics to confirm the water accumulation area.

[0110] In this embodiment, multi-feature fusion recognition is performed through step 306 and its sub-steps: First, step 311 is executed: historical water accumulation recognition data of the current working area is retrieved from the historical water accumulation recognition database stored on the vehicle. This data includes information such as the light intensity and area of ​​the water accumulation area during each water accumulation. At the same time, the real-time light intensity of the working area is obtained through the vehicle's light sensor. The historical average light value of the area under rainy weather conditions is extracted from the historical water accumulation recognition data. The real-time light intensity is compared with the historical average light value, and the area threshold is adjusted according to the comparison result. If the real-time light intensity is lower than the historical average light value, it indicates that the current reflective area may be too small, so the area threshold is appropriately lowered. If the real-time light intensity is higher than the historical average light value, the area threshold is adjusted accordingly. If the illumination value is too high, the area threshold is adjusted accordingly to finally determine the area threshold suitable for the current illumination. Step 312: Extract the radar distance values ​​of historical dry road surfaces and historical waterlogged areas from the historical waterlogging identification data, and obtain the distance difference range by subtracting the two. Confirm the road surface material of the current work area using the vehicle-mounted road surface material sensor (or pre-stored road segment information). Based on the influence of different road surface materials on the radar distance difference (e.g., the radar distance difference is smaller for some road surface materials and larger for others when waterlogged), select the distance difference threshold suitable for the current road surface material from the distance difference range. Step 313: Input all reflective pixel coverage areas into the multi-feature fusion recognition model. The model first calculates the image of each area... The pixel coverage area (i.e., the total number of reflective pixels in the statistical area) is calculated, and then the grayscale change rate of each edge pixel is calculated: by selecting pixels at the edge of the area (such as the outermost ring of pixels), the difference in grayscale value between each edge pixel and its adjacent non-edge pixels is calculated, and then divided by the pixel spacing; if the pixel coverage area of ​​the area is greater than or equal to the area threshold, or the grayscale change rate of the edge pixels is greater than the preset grayscale change threshold, then the area is marked as a candidate water accumulation area; Step 314: From the ground distance correction data, select the radar detection points corresponding to areas without reflective features in the working area (such as dry lanes without water accumulation), and extract the corrected distance values ​​of these detection points as the dry road surface detection distance values; through the vehicle-mounted slope The sensor acquires ground structure parameters (such as slope) of the work area, and corrects the dry road surface detection distance value according to the ground structure parameters (e.g., slope will cause deviation in radar distance value, which needs to be corrected according to slope-related rules); the average value of the corrected dry road surface detection distance value is taken to calculate the dry road surface distance value; step 315 is executed: for each candidate waterlogged area, the average corrected distance value is subtracted from the dry road surface distance value to obtain the actual distance difference (if the average corrected distance value is less than the dry road surface distance value, the absolute value of the difference is taken); at the same time, the difference between the corrected distance value and the dry road surface distance value of all radar detection points in the candidate area is calculated, and these differences are summed and divided by the number of detection points to obtain the average difference value.Step 316: Determine whether the actual distance difference of each candidate waterlogged area is greater than the distance difference threshold, and whether the average difference is greater than or equal to the preset difference threshold. If both conditions are met, the candidate waterlogged area is marked as a suspected waterlogged area; if either condition is not met, it is not marked as a suspected waterlogged area. Step 317: Further optimize the identification results: First, calculate the center coordinates based on the first coordinate range of each suspected waterlogged area (i.e., the maximum x-coordinate, minimum x-coordinate, maximum y-coordinate, and minimum y-coordinate of all radar detection points within the area) using step 321. (The x-coordinate is the average of the x-values ​​in the first coordinate range, and the y-coordinate is the average of the x-values ​​in the first coordinate range.) Take the average value of the y-values ​​in the first coordinate range); calculate the center-to-center distance between each suspected waterlogged area and any other suspected waterlogged area using the distance formula between two points; execute step 322: merge the suspected waterlogged areas according to the region merging rules of the multi-feature fusion recognition model: preset merging distance threshold and preset edge spacing. If the center-to-center distance between two suspected areas is less than the preset merging distance threshold, and the first coordinate ranges of the two areas overlap (i.e., the x-coordinate range of one area intersects with the x-coordinate range of another area, and the y-coordinate ranges also intersect), or the two areas do not overlap but the edge spacing (the shortest distance between the boundaries of the first coordinate ranges of the two areas) is less than the preset edge spacing. If the distance between the edges is not specified, the two areas are merged into one region; this merging operation is repeated until there are no more suspected waterlogged areas that meet the region merging rules, and finally the waterlogged area to be determined is obtained; step 323 retrieves the historical morphological features of historical waterlogged areas that match the current work area (such as road surface material and road type) from the historical waterlogged area morphology database. The historical morphological features include the ratio of the major and minor axes of the outline, the actual coverage area, and the ratio of the area of ​​the circumscribed rectangle; step 324 is executed: the actual morphology of the waterlogged area to be determined is calculated based on the second coordinate range (i.e., the maximum x-coordinate, minimum x-coordinate, maximum y-coordinate, and minimum y-coordinate of the merged area). Features: The contour of the waterlogged area to be determined is fitted by the second coordinate range. The lengths of the longest axis (major axis) and the shortest axis (minor axis) of the contour are measured, and the ratio of the major and minor axes is calculated. The area of ​​the circumscribed rectangle of the waterlogged area to be determined is calculated based on the second coordinate range (the boundary of the circumscribed rectangle is the boundary of the second coordinate range). The road surface area corresponding to the actual pixels within the waterlogged area is counted, and the ratio of the circumscribed rectangle area is calculated. The actual morphological features of the waterlogged area to be determined are compared with the retrieved historical morphological features. If the deviation between the actual morphological features and the historical morphological features is within a preset allowable range, the waterlogged area to be determined is marked as a waterlogged area.

[0111] Furthermore, the construction, training, and evaluation process of the multi-feature fusion recognition model is as follows:

[0112] The construction process of the multi-feature fusion recognition model includes: First, clarifying that the model input consists of two core features: road surface reflectivity features from the visual modality and ground distance correction features from the radar modality. The road surface reflectivity features from the visual modality include the coordinate range of the reflective pixel coverage area (including left boundary coordinates, right boundary coordinates, front boundary coordinates, and back boundary coordinates), the coverage area, and the edge grayscale change rate. The ground distance correction features from the radar modality include the corrected distance mean, distance standard deviation, and the difference between the corrected distance and the actual distance to the dry road surface. Then, a three-order architecture of independent feature processing, cross-modal fusion, and classification decision is adopted. The visual features are encoded and dimensionality reduced through a two-layer fully connected network. The first layer has an input dimension of 5 (the left boundary coordinates, right boundary coordinates, and other coordinates of the reflective pixel coverage area). The first layer has an input dimension of 64 and an output dimension of 32, using a linear rectified function. The second layer has an input dimension of 64 and an output dimension of 32, also using a linear rectified function. Radar features are encoded through a fully connected network with an input dimension of 3 (corrected mean distance, standard deviation of distance, and difference between the actual distance and the dry road surface) and an output dimension of 32, using a linear rectified function with a negative slope (0.01). These two layers are concatenated and then weighted by a 1×1 convolutional layer for channel attention, resulting in a 64-dimensional weighted fusion feature. Finally, a two-layer fully connected network outputs the classification result for either waterlogged or non-waterlogged areas. The first layer has an input dimension of 64 and an output dimension of 32, using a linear rectified function. 4. The output dimension is 16, and the activation function is a linear rectified function; the second layer has an input dimension of 16 and an output dimension of 2, and the activation function is a soft maximization function. The model architecture is adapted to in-vehicle embedded devices. The above construction logic can be derived from the feature definition and recognition process. The training process constructs a training dataset based on historical water accumulation recognition data, real-time light intensity, road surface material, etc. The dataset covers three rainfall intensities: light rain (0.1-10mm / 24h), moderate rain (10-25mm / 24h), and heavy rain (25-50mm / 24h). It covers two types of urban secondary road surface materials: asphalt and cement. It covers four types of water accumulation: sheet-like water accumulation, strip-like water accumulation, point-like water accumulation, and mixed water accumulation. Each type... No fewer than 500 samples were collected. Sample labeling employed a dual-verification method combining manual labeling and depth detection component measurements. Initial labeling was based on distance difference thresholds (distance difference from dry road surface greater than 5mm) and area thresholds (reflective area greater than 0.01㎡), followed by confirmation through depth detection component measurements. Data augmentation was achieved by randomly perturbing the grayscale change rate of visual features by ±10%, randomly scaling the coordinate range by ±5%, adding ±0.02m Gaussian noise to the corrected distance mean of radar features, and randomly adjusting the actual distance difference by ±0.01m. During training, a weighted cross-entropy loss was used to balance the class distribution (30% for water accumulation samples and 70% for non-water accumulation samples, with water accumulation samples assigned a 2.05% threshold).The model uses an adaptive momentum estimation optimizer with an initial learning rate of 0.001 and a weight decay coefficient of 0.00001. Cosine annealing scheduling is used for the learning rate, with a period of 10 rounds and a minimum learning rate of 0.00001. The training run consists of 30 rounds with a batch size of 32. A graphics processing unit (GPU) is used for training, employing mixed precision training for acceleration. After each training round, accuracy is calculated on the validation set. If the validation set accuracy does not improve for three consecutive rounds, an early stopping mechanism is triggered, and the current optimal model weights are saved. After training, the model is pruned (removing connections with weights less than 0.0004) to reduce weight density. The model parameters are limited to no more than 500,000 and are compatible with vehicle-mounted hardware. The evaluation process is based on feature judgment standards (area threshold, distance difference threshold) and business requirements. The specific content and quantitative parameters of the business requirements are as follows: First, to avoid damage to the sweeping mechanism due to missed water accumulation. The requirement is to ensure that new energy sanitation vehicles do not miss any water accumulation areas on the road during rainy weather, and to prevent the sweeping execution components from malfunctioning due to excessive load caused by excessive contact with water accumulation. The quantitative parameters include a water accumulation depth threshold of ≥5mm, a sweeping mechanism load threshold such as torque of 50-80N・m, and a current of ≤15A (i.e., the upper limit of the load for safe operation of the sweeping mechanism). Secondly, to avoid misjudging slippery road surfaces leading to ineffective deceleration or detours, the requirement is to ensure that new energy sanitation vehicles do not misjudge ordinary slippery road surfaces as waterlogged areas, thus avoiding unnecessary deceleration or detours that reduce cleaning efficiency, prolong working time, and increase additional energy consumption. Quantitative parameters include an ineffective deceleration time threshold of ≤30s / 100m (set based on waterlogged area avoidance instructions and speed adjustments; this threshold represents the maximum permissible ineffective deceleration time per 100m of working section; misjudgment will cause the vehicle to drop from normal working speed to avoidance speed), and an ineffective detour length threshold of ≤5m / suspected area (set based on the detour length of the avoidance path; if misjudged...). The judgment will result in additional detours; this threshold is the maximum invalid detour length allowed for a single suspected area. Thirdly, it ensures a balance between rainy weather cleaning coverage and battery life. The requirements are to ensure that, while ensuring accurate avoidance of waterlogged areas and safe operation of the cleaning mechanism, the rainy weather cleaning coverage of urban secondary roads is not lower than that of dry roads, while controlling vehicle battery energy consumption to avoid a significant decrease in battery range. Quantitative parameters include a cleaning coverage threshold of ≥98% (set based on the completion of autonomous rainy weather cleaning of urban secondary roads and the completion of cleaning of all cleanable target areas; this threshold is the minimum cleaning coverage rate required for rainy weather operations) and a battery energy consumption threshold of ≤0.6kWh / km (based on adjusting the vehicle battery discharge parameters and optimizing the discharge parameters based on load data, this threshold is the maximum energy consumption per unit mile allowed for operation in rainy weather). This business requirement is related to the evaluation indicators of the multi-feature fusion recognition model: recall rate ≥95% and precision rate ≥90%. By meeting the indicator standards, we ensure that the quantitative parameters meet the requirements, so as to ultimately achieve the goal of improving the safety, energy efficiency and cleaning stability of autonomous operation. The core evaluation indicators employ a judgment logic of partial achievement plus comprehensive balance, including: Classification accuracy, which reflects the overall prediction correctness of the model, is calculated as: Classification accuracy = (Number of true positives + Number of true negatives) ÷ (Number of true positives + Number of true negatives + Number of false positives + Number of false negatives), where the number of true positives represents the number of samples that are actually waterlogged areas but the model predicts them as waterlogged areas; the number of true negatives represents the number of samples that are actually non-waterlogged areas but the model predicts them as non-waterlogged areas; the number of false positives represents the number of samples that are actually non-waterlogged areas but the model predicts them as waterlogged areas; and the number of false negatives represents the number of samples that are actually waterlogged areas but the model predicts them as non-waterlogged areas. Waterlogged area recall focuses on missed detection control, corresponding to the need to avoid damage to the cleaning mechanism due to missed water, and is calculated as: Waterlogged area recall = Number of true positives ÷ (Number of true positives + Number of false negatives). Waterlogged area precision focuses on misjudgment control, corresponding to the need to avoid ineffective deceleration due to misjudging slippery surfaces, and is calculated as: Waterlogged area... Precision = Number of true positives ÷ (Number of true positives + Number of false positives); F1 score, as the harmonic mean of precision and recall, is used to balance the contradiction between missed detections and false positives. The calculation formula is F1 score = (2 × Precision × Recall) ÷ (Precision + Recall); During evaluation, first check whether each indicator meets the preset thresholds (precision not less than 90%, recall not less than 95%, precision not less than 90%, F1 score not less than 92%), and then verify the value of fusion through ablation experiments: compare the indicator differences between pure visual models and pure radar models to prove the advantages of cross-modal fusion in improving recall and precision. All evaluation data are generated based on an independent test set (3000 samples, including 900 samples from waterlogged areas and 2100 samples from non-waterlogged areas) containing new scenarios (residual water on the road after rain, water covered with fallen leaves). In the above training and evaluation process, the feature judgment criteria directly determine the sample labeling logic, and the indicator thresholds directly correspond to the safety and efficiency requirements of sanitation vehicle operations.

[0113] This application's embodiments eliminate the impact of atmospheric refraction on distance detection during rainy weather by correcting the refractive index of radar data, thereby improving the accuracy of radar distance data; avoid the defects of single visual data being affected by reflective interference or single radar data being unable to locate a specific range; reduce misjudgment of interference areas; and combine historical morphological features to finally confirm the merged area, eliminating interference from non-waterlogged areas and improving the reliability of the recognition results; the overall process achieves accurate identification of waterlogged areas in rainy weather scenarios, providing reliable support for the safe operation of new energy sanitation vehicles and urban flooding early warning.

[0114] This application provides a specific embodiment. Step 104 involves comparing the actual water accumulation parameters with the water accumulation characteristic data to generate a water accumulation area avoidance instruction, which specifically includes the following steps:

[0115] Step 401: Compare the actual water depth, actual water time, and actual water area percentage in the actual water accumulation parameters with the average range of water depth, water time threshold, and water area percentage benchmark value in the water accumulation feature data to determine the avoidance priority of the water accumulation area.

[0116] In this step, the actual waterlogged area ratio refers to the ratio of the actual area of ​​the waterlogged area to the total working area of ​​the current work section. It reflects the degree to which the waterlogged area occupies the work path and is used to determine whether the waterlogging affects normal operations and the extent of that impact. The average waterlogging depth range refers to the fluctuation range of the statistical average waterlogging depth when waterlogging occurred on similar road sections (such as asphalt secondary roads) during the same period in history. It is calculated based on a large amount of waterlogging depth data in the historical waterlogging identification database and is used to compare the abnormality of the current actual waterlogging depth to help determine the waterlogging hazard level. The waterlogging time threshold refers to the preset critical duration of waterlogging. It is set based on empirical data showing that prolonged waterlogging can easily cause road damage, hidden manhole cover defects, and other hidden dangers. It is used to determine the continued danger of the waterlogged area and provides a time dimension basis for prioritizing avoidance. The waterlogged area ratio benchmark value refers to the critical value for determining whether waterlogging obstructs operations. It is calculated based on the operating width of new energy sanitation vehicles and the effective traffic width of the road. It is used to determine the degree of obstruction of the work path by the waterlogged area and is one of the core indicators for determining the necessity of avoidance. The avoidance priority refers to the emergency level of avoidance based on the degree of danger of water accumulation and its impact on operations. It is used to guide new energy sanitation vehicles to prioritize the avoidance of high-risk water accumulation areas, so as to ensure operational safety and efficiency.

[0117] In this embodiment, actual water accumulation parameters (including actual water accumulation depth, actual water accumulation time, and actual water accumulation area percentage) are retrieved from the water accumulation monitoring module, and water accumulation characteristic data (including average water accumulation depth range, water accumulation time threshold, and water accumulation area percentage benchmark value) are retrieved from the historical water accumulation characteristic database. Then, three comparisons are performed to determine whether the actual water accumulation depth exceeds the average water accumulation depth range, whether the actual water accumulation time exceeds the water accumulation time threshold, and whether the actual water accumulation area percentage is higher than the water accumulation area percentage benchmark value. The comparison results are then integrated according to preset rules (e.g., all three exceeding the standard is the highest priority, two exceeding the standard is the medium priority, and one exceeding the standard is the low priority) to obtain the avoidance priority for the water accumulation area.

[0118] Step 402: Obtain the operating position, operating width, and minimum turning radius of the new energy sanitation vehicle, and determine the minimum offset distance of the avoidance path by combining the third coordinate range of the water accumulation area.

[0119] In this step, the operating position of the new energy sanitation vehicle refers to the real-time geographical coordinates of the sanitation vehicle currently performing cleaning operations. This coordinate is collected based on the vehicle's onboard GPS device and used to determine the starting reference point for the avoidance path, ensuring that the path planning matches the vehicle's current state. The vehicle's operating width refers to the maximum lateral working distance of the sanitation vehicle's operating devices (such as sweeping brushes and water spray nozzles) after deployment. This is obtained based on the vehicle's factory parameters or current operating mode settings and is used to calculate the lateral offset space required for avoidance, preventing the operating devices from contacting accumulated water. The minimum turning radius refers to the minimum distance from the turning center to the outer tire track of the vehicle when turning at the minimum safe driving speed. This is obtained based on the vehicle's chassis structure and steering system parameters and is used to ensure that the steering action of the avoidance path conforms to the vehicle's mechanical performance, avoiding steering failure. The third coordinate range refers to the spatial boundary parameters of the water accumulation area in the radar detection area coordinate system, including the maximum and minimum x-coordinates, the maximum and minimum y-coordinates. This is obtained based on the radar detection point coordinates of the water accumulation area and is used to accurately locate the spatial position and range of the water accumulation area. The minimum offset distance of the avoidance path refers to the minimum lateral distance that a sanitation vehicle needs to deviate from its original working path to avoid a flooded area. This is used to ensure that the vehicle and its working equipment do not come into contact with the flooded area during the operation.

[0120] In this embodiment, the operating position of the new energy sanitation vehicle is obtained through the vehicle-mounted positioning device. The vehicle's operating width and minimum turning radius are retrieved from the vehicle parameter database stored on the vehicle. Then, taking the outer boundary of the third coordinate range of the water accumulation area as a reference, combined with the vehicle's operating width (to ensure that the operating device does not come into contact with the water accumulation), and referring to the minimum turning radius (to avoid the vehicle from entering the water accumulation area when turning), the minimum lateral distance that the sanitation vehicle needs to deviate from the original operating path is determined through intuitive distance measurement. This distance is the minimum offset distance of the avoidance path.

[0121] Step 403: Calculate the detour length of the avoidance path based on the minimum offset distance and the length of the water accumulation area.

[0122] In this step, the length of the waterlogged area refers to the maximum linear distance of the waterlogged area along the operating direction of the new energy sanitation vehicle. It is calculated based on the coordinate difference of the operating direction within the third coordinate range and is used to measure the extension range of the waterlogged area on the operating path, providing a basic parameter for calculating the detour length. The detour length of the avoidance path refers to the additional path length that the sanitation vehicle takes compared to the original operating path. It is calculated based on the minimum offset distance and the length of the waterlogged area and is used to assess the increased driving cost of the avoidance operation, assisting in optimizing operating efficiency.

[0123] In this embodiment, the length of the waterlogged area (i.e., the maximum linear distance of the waterlogged area along the sanitation vehicle's operating direction) is extracted based on the third coordinate range. Then, the detour length is calculated based on the geometric relationship. The avoidance path is a broken line shape of offset-straight-return. The detour length is equal to the total length of the broken line minus the length of the original operating path (a straight line through the waterlogged area). Specifically, the square root of the sum of the square of the minimum offset distance and the square of half the length of the waterlogged area is multiplied by two and then the length of the waterlogged area is subtracted. The resulting value is the detour length of the avoidance path.

[0124] Step 404: Based on the operating location, third coordinate range, minimum offset distance, and detour length of the new energy sanitation vehicle, determine the starting point, turning point, and ending point of the avoidance path to form a set of avoidance path points.

[0125] In this step, the avoidance path point set refers to the ordered set of coordinates of the start point, turning point, and end point of the avoidance path, which is used to provide sanitation vehicles with specific and executable avoidance driving trajectory guidance.

[0126] In this embodiment, the starting point of the new energy sanitation vehicle is taken as the starting reference. The starting and ending boundaries of the waterlogged area are determined by combining the third coordinate range. Then, the lateral offset is determined according to the minimum offset distance. The total extension range of the path is controlled by referring to the detour length. The starting point (the coordinate point in front of the working position that is about to reach the waterlogged area), the turning point (the three key coordinate points of starting offset, maintaining offset, and starting return), and the ending point (the coordinate point that returns to the original working path and is located behind the waterlogged area) of the avoidance path are marked in sequence. These coordinate points are arranged in the driving order to form an avoidance path point set.

[0127] Step 405: Generate a water accumulation area avoidance command by taking the avoidance priority, avoidance path point set, minimum offset distance, and detour length.

[0128] In this embodiment, the avoidance priority obtained in the first step, the set of avoidance path points obtained in the fourth step, the minimum offset distance of the avoidance path obtained in the second step, and the detour length of the avoidance path obtained in the third step are combined and spliced ​​according to a preset instruction format to ensure that each parameter is clearly corresponding and conforms to the recognition specifications of the vehicle control system, and finally generate an avoidance instruction for sanitation vehicles to avoid waterlogged areas.

[0129] This application's embodiments solve the problems of traditional obstacle avoidance relying on manual judgment, low accuracy, and poor safety, achieving precision and intelligence in water accumulation avoidance; at the same time, by clarifying path parameters and trajectory points, it ensures that the avoidance action conforms to vehicle performance and does not affect the integrity of the operation, reducing the accident risk of operating in waterlogged sections and improving the safety and efficiency of sanitation operations.

[0130] This application provides a specific embodiment. Step 105 involves adjusting the discharge parameters of the vehicle battery based on the actual water accumulation parameters to obtain the adjusted discharge parameters. This specifically includes the following steps:

[0131] Step 501: Set the first adjustment coefficient based on the correlation between the actual water depth and the road surface material of the waterlogged area in the actual water accumulation parameters.

[0132] In this step, the correlation refers to a pre-established table based on historical rainy weather operation data, showing the correspondence between actual water depth, road surface material in the waterlogged area, and adjustment coefficients. This table reflects the degree of power compensation required by the sweeping actuator under different water depths (greater resistance with deeper water) and road surface materials (e.g., cement roads have a lower coefficient of friction and greater water resistance than asphalt roads). It is used to quickly and accurately set the first adjustment coefficient to ensure the discharge current adapts to changes in water resistance. The first adjustment coefficient is a value used to adjust the standard discharge current of the vehicle battery. It is determined based on the correlation between actual water depth and road surface material in the waterlogged area, reflecting the current compensation requirement for sweeping power under waterlogged conditions. Its value range is typically 0.8-1.5 (greater than 1 for high water resistance, less than 1 for low resistance), ensuring the current output adapts to the water resistance.

[0133] In this embodiment, the actual water depth is obtained from the actual water accumulation parameters from the depth detection component, the road material (such as asphalt or cement) of the water accumulation area is extracted from the vehicle-mounted road material database (or pre-stored road information of the work area), and then the association table of actual water accumulation depth, road material of water accumulation area, and adjustment coefficient is retrieved based on historical rainy day operation data. According to the currently obtained actual water accumulation depth and road material, the corresponding coefficient value is matched in the association table and the coefficient value is set as the first adjustment coefficient. The logic for establishing the association table is: the greater the water accumulation depth and the smaller the friction coefficient of the road material (such as cement road), the larger the first adjustment coefficient is, so as to adapt to the greater sweeping resistance.

[0134] Step 502: Based on the matching relationship between the water distribution density in the actual water accumulation parameters and the preset load level of the cleaning execution component, set a second adjustment coefficient.

[0135] In this step, the preset load level refers to the level (usually divided into low, medium, and high) based on the load of the cleaning actuator under different operating conditions. It is set based on historical cleaning load data and motor output capacity, reflecting the power intensity required by the cleaning actuator (higher load levels correspond to greater motor output demands). This is used to match the water distribution density with the second adjustment coefficient, ensuring that the voltage output adapts to load changes. The matching relationship refers to a pre-established rule table of correspondence between water distribution density and preset load level, reflecting the correlation between water distribution density (higher density means higher frequency of water contact with the cleaning actuator, resulting in a greater load) and the cleaning load. This is used to quickly determine the corresponding preset load level based on the current water distribution density, providing a basis for setting the second adjustment coefficient. The second adjustment coefficient is a coefficient value used to adjust the standard discharge voltage of the vehicle battery. It is determined based on the matching relationship between water distribution density and the preset load level of the cleaning actuator, reflecting the voltage compensation requirement of the load level for cleaning power. The value range is usually 0.9-1.4 (the coefficient for high load levels is greater than 1, and the coefficient for low load levels is less than 1), ensuring that the voltage output adapts to the load intensity.

[0136] In this embodiment, the water distribution density in the actual water accumulation parameters is calculated (calculated by dividing the area of ​​the water accumulation area by the total area of ​​the working section). The preset load level (divided into three levels—low, medium, and high—based on the magnitude of the cleaning resistance, corresponding to different motor output requirements) is retrieved from the control parameter library of the sweeping execution component. Then, the pre-established matching relationship table of water distribution density and preset load level is retrieved. The corresponding preset load level is determined based on the current water distribution density. Next, the matching coefficient value is matched in the load level-adjustment coefficient correspondence table based on the preset load level. This coefficient value is set as the second adjustment coefficient. The core logic of the matching relationship is: the higher the water distribution density, the higher the matched preset load level, and the larger the corresponding second adjustment coefficient, so as to meet the voltage requirements under high load.

[0137] Step 503: Determine the basic adjustment strategy based on the remaining charge and battery temperature of the vehicle battery. The basic adjustment strategy includes a normal adjustment mode, a balance adjustment mode, and an energy-saving adjustment mode.

[0138] In this step, the basic adjustment strategy refers to the overall rules for adjusting discharge parameters based on the remaining battery power and battery temperature. This includes three modes: regular adjustment mode, balanced adjustment mode, and energy-saving adjustment mode. These reflect the constraints imposed by the battery state on the discharge strategy (e.g., prioritizing energy saving when the battery power is low). They are used to set constraints for the initial calculation of current, voltage, and output power, ensuring that the discharge meets operational requirements while protecting the battery. The regular adjustment mode is one of the basic adjustment strategies, suitable for scenarios where the remaining battery power is high (typically >60%) and the battery temperature is normal (15-35℃). Its core principle is to prioritize the power output of the cleaning execution components, with a more lenient range of discharge parameter adjustments (e.g., higher current and power limits), used to efficiently complete cleaning operations when the battery is in good condition. The Balanced Adjustment Mode is one of the basic adjustment strategies, suitable for scenarios where the vehicle battery has a moderate remaining charge (30%-60%) and a slightly high battery temperature (35-45℃). Its core principle is to balance cleaning power with battery protection. The discharge parameter adjustment range is between that of the normal and energy-saving modes (e.g., a moderate power upper limit), used to balance operational efficiency and battery life when the battery condition is average. The Energy-Saving Adjustment Mode is another basic adjustment strategy, suitable for scenarios where the vehicle battery has a low remaining charge (<30%) or an excessively high battery temperature (>45℃). Its core principle is to prioritize reducing energy consumption and protecting the battery. The discharge parameter adjustment range is strictly constrained (e.g., lower current and power upper limits), used to extend range and prevent battery damage when the battery condition is poor.

[0139] In this embodiment, the remaining battery power (e.g., presented as a percentage) and battery temperature (e.g., presented in degrees Celsius) of the vehicle battery are obtained through the vehicle battery management system, and then filtered according to preset judgment rules: if the remaining battery power is higher than 60% and the battery temperature is between 15-35℃ (normal temperature range), the basic adjustment strategy is determined to be the conventional adjustment mode (prioritizing cleaning power); if the remaining battery power is between 30%-60% and the battery temperature is between 35-45℃ (slightly higher temperature), the balance adjustment mode is determined (balancing power and battery protection); if the remaining battery power is lower than 30% or the battery temperature is higher than 45℃ (high temperature), the energy-saving adjustment mode is determined (prioritizing energy consumption reduction and battery protection). The three modes correspond to different discharge parameter adjustment constraint ranges (e.g., the upper limit of discharge power in energy-saving mode is lower than that in conventional mode).

[0140] Step 504: Under the constraint of the basic adjustment strategy, multiply the standard discharge current of the vehicle battery and the first adjustment coefficient to obtain the preliminary adjustment current, and multiply the standard discharge voltage of the vehicle battery and the second adjustment coefficient to obtain the preliminary adjustment voltage.

[0141] In this step, the constraint premise refers to the limiting conditions of the basic adjustment strategy on the calculation of discharge parameters. It reflects the safety boundary of the battery state for power output and is used to ensure that the calculation of the preliminary adjustment current, preliminary adjustment voltage, and output power does not exceed the safe operating range of the battery, avoiding over-discharge or high-temperature overload. The standard discharge current refers to the default discharge current of the vehicle battery when operating on dry urban secondary roads. Based on the battery's rated capacity and the conventional load setting of the sweeping actuator, it reflects the normal current output level when there is no water accumulation interference, ensuring that current adjustment has a clear reference basis. The preliminary adjustment current refers to the current value obtained by multiplying the standard discharge current of the vehicle battery by the first adjustment coefficient, reflecting the initial current requirement adapted to the current water depth and road surface material. The standard discharge voltage of the vehicle battery refers to the default discharge voltage of the vehicle battery when operating on dry urban secondary roads. Based on the battery's rated voltage and the conventional voltage requirement setting of the sweeping actuator, it reflects the normal voltage output level when there is no water accumulation interference. The initial adjustment voltage refers to the voltage value obtained by multiplying the standard discharge voltage of the vehicle battery by the second adjustment coefficient. It reflects the initial voltage requirement to adapt to the current water accumulation density and preset load level. It must comply with the constraints of the basic adjustment strategy and is the core voltage parameter that constitutes the discharge parameters and is verified.

[0142] In this embodiment, the standard discharge current and standard discharge voltage of the vehicle battery are retrieved from the battery parameter manual (or stored battery rating parameters). The standard discharge current is then multiplied by the first adjustment coefficient obtained in the first step to obtain the preliminary adjustment current (if the basic adjustment strategy is energy-saving mode, it must be ensured that the preliminary adjustment current does not exceed the current limit of the energy-saving mode). At the same time, the standard discharge voltage is multiplied by the second adjustment coefficient obtained in the second step to obtain the preliminary adjustment voltage (similarly, it must comply with the voltage constraints of the basic adjustment strategy), ensuring that the preliminary current and voltage are adapted to the current water accumulation condition and battery protection requirements.

[0143] Step 505: Based on the ratio of the area of ​​the water accumulation zone to the working width of the vehicle, and in conjunction with the preset load level of the sweeping execution component, set the power correction value.

[0144] In this step, the area of ​​the waterlogged area refers to the actual coverage area of ​​the waterlogged area on the ground, determined by radar detection and visual recognition. It is calculated based on the coordinate range and image scale of the waterlogged area, reflecting its size. The power correction value is a value used to compensate for the power loss of the sweeping actuator under waterlogged conditions. It is determined based on the ratio of the waterlogged area to the vehicle's working width and the preset load level of the sweeping actuator, reflecting the additional power demand under the continuous influence of waterlogging (the larger the ratio and the higher the load level, the larger the correction value), ensuring that the power output meets the actual operating resistance.

[0145] In this embodiment, the vehicle operating width is retrieved from the vehicle parameter library, and the ratio of the area of ​​the water accumulation area to the vehicle operating width is calculated (this ratio reflects the extent of the water accumulation area along the operating direction; the larger the ratio, the longer the sweeping component is subjected to water resistance). Then, combined with the preset load level of the sweeping execution component determined in the second step, a pre-established correspondence table of area-operating width ratio-preset load level-power correction value is retrieved. Based on the current ratio and load level, the corresponding power correction value is matched in the table (the larger the ratio and the higher the load level, the larger the power correction value, used to compensate for power loss under long-term high resistance).

[0146] Step 506: Based on the initial adjustment current, the initial adjustment voltage, and the power correction value, calculate the initial output power, and adjust the initial output power according to the battery temperature to obtain the adjusted output power.

[0147] In this step, the initial output power refers to the power value obtained by multiplying the initial adjustment current and the initial adjustment voltage, and adding the power correction value. It reflects the initial power requirement for adapting to water accumulation conditions and does not consider the impact of battery temperature on power. It is the basic power parameter for subsequent temperature adjustment. The adjusted output power refers to the power value obtained after adjusting the initial output power in conjunction with battery temperature. It reflects the final power requirement that balances water accumulation conditions and battery temperature protection, ensuring that cleaning power is met while avoiding excessive battery wear due to high or low temperatures. It is the core power parameter that makes up the adjusted discharge parameters.

[0148] In this embodiment, the initial power value = initial adjustment current × initial adjustment voltage, and the initial output power = initial power value + power correction value; the battery temperature of the vehicle battery is obtained, and the initial output power is adjusted according to the preset battery temperature-power adjustment ratio rule: if the battery temperature is higher than 35°C, the initial output power is reduced by 3% for every 5°C increase (to avoid excessive power at high temperatures that aggravates battery loss); if the battery temperature is lower than 15°C, the initial output power is increased by 2% for every 5°C decrease (to compensate for the impact of battery capacity reduction at low temperatures), and the adjusted output power is obtained after adjustment.

[0149] Step 507: Combine the adjusted output power, the preliminary adjusted current, and the preliminary adjusted voltage to form the discharge parameters of the vehicle battery. Perform an adaptability check on the discharge parameters. If the discharge parameters are within the rated input parameter range of the vehicle motor, then the discharge parameters are used as the adjusted discharge parameters. If the discharge parameters exceed the rated input parameter range, then the discharge parameters are adjusted to the rated input parameter range to obtain the adjusted discharge parameters.

[0150] In this step, the rated input parameter range of the vehicle motor refers to the range of input current, voltage, and power that the vehicle motor is allowed to operate safely. Based on the motor's design parameters and safety standards, it reflects the motor's safe operating boundaries and is used to verify the compatibility of the vehicle battery's discharge parameters to avoid motor overload damage.

[0151] In this embodiment, the adjusted output power, preliminary adjusted current, and preliminary adjusted voltage are integrated to form the discharge parameters of the vehicle battery. Then, the rated input parameter range of the vehicle motor (including the rated input current range, rated input voltage range, and rated input power range) is retrieved from the vehicle motor's parameter manual, and the combined discharge parameters are verified: if the preliminary adjusted current is within the rated input current range, the preliminary adjusted voltage is within the rated input voltage range, and the adjusted output power is within the rated input power range, then the discharge parameter is directly used as the adjusted discharge parameter; if any parameter exceeds the rated range (e.g., the preliminary adjusted current is too high), then the parameter exceeding the rated range is adjusted according to the upper or lower limit of the rated range (e.g., the current is reduced to the upper limit of the rated current), and the parameters are recombined after adjustment until all parameters are within the rated input parameter range, finally obtaining the adjusted discharge parameters.

[0152] The embodiments of this application enable the discharge current and voltage to adapt to changes in water resistance, avoiding insufficient power or overload caused by fixed parameters; ensuring precise power supply under different water conditions and battery statuses; and eliminating the risk of motor failure caused by parameters exceeding the range by verifying the rated parameters of the vehicle motor.

[0153] Figure 2 This is a schematic diagram illustrating a specific implementation of a new energy sanitation vehicle cleaning control system integrating vision and radar, provided in an embodiment of this application. (Refer to...) Figure 2 The system may include:

[0154] The acquisition module 21 is used to acquire water accumulation characteristic data, operation section images, and ground distance data of urban secondary arterial roads in rainy weather scenarios;

[0155] Filtering module 22 is used to filter the interference areas in the image of the work section using an adaptive median filtering algorithm to obtain road surface reflective feature data;

[0156] The compensation module 23 is used to perform deviation compensation processing on the ground distance data to obtain ground distance correction data. Combined with the road surface reflective feature data, the water accumulation area is identified through a multi-feature fusion recognition model.

[0157] The generation module 24 is used to control the depth detection component mounted on the new energy sanitation vehicle to detect the actual water accumulation parameters of the water accumulation area, compare the actual water accumulation parameters with the water accumulation feature data, and generate a water accumulation area avoidance command to bypass the water accumulation area.

[0158] The adjustment module 25 is used to adjust the discharge parameters of the vehicle battery according to the actual water accumulation parameters, so as to obtain the adjusted discharge parameters.

[0159] The cleaning module 26 is used to control the cleaning execution component to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and to collect the load data of the cleaning execution component during the cleaning process. If the load data exceeds the preset load range, the adjusted discharge parameters are optimized based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

[0160] This application provides an embodiment of a new energy sanitation vehicle cleaning control system that integrates vision and radar to implement the aforementioned new energy sanitation vehicle cleaning control method that integrates vision and radar. Therefore, the specific implementation of the new energy sanitation vehicle cleaning control system that integrates vision and radar can be found in the embodiment section of the new energy sanitation vehicle cleaning control method that integrates vision and radar mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0161] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the above-described new energy sanitation vehicle cleaning control method that integrates vision and radar.

[0162] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for controlling the cleaning of a new energy sanitation vehicle that integrates vision and radar.

[0163] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0164] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the new energy sanitation vehicle cleaning control method integrating vision and radar.

[0165] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0166] The foregoing has provided a detailed description of a new energy sanitation vehicle cleaning control method, system, electronic device, and storage medium integrating vision and radar, as provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A cleaning control method for new energy sanitation vehicles integrating vision and radar, characterized in that, include: Acquire water accumulation characteristic data, images of the work section, and ground distance data of urban secondary arterial roads in rainy weather scenarios; An adaptive median filtering algorithm is used to filter out the interference areas in the image of the work section to obtain road surface reflectivity feature data; The ground distance data is subjected to deviation compensation processing to obtain ground distance correction data. Combined with the road surface reflective feature data, the water accumulation area is identified through a multi-feature fusion recognition model. The depth detection component mounted on the new energy sanitation vehicle is controlled to detect the actual water accumulation parameters of the water accumulation area, and the actual water accumulation parameters are compared with the water accumulation feature data to generate a water accumulation area avoidance command to bypass the water accumulation area. Based on the actual water accumulation parameters, adjust the discharge parameters of the vehicle battery to obtain the adjusted discharge parameters; The cleaning execution component is controlled to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and the load data of the cleaning execution component is collected during the cleaning process. If the load data exceeds the preset load range, the adjusted discharge parameters are optimized based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

2. The method according to claim 1, characterized in that, An adaptive median filtering algorithm is used to filter out interference areas in the image of the work section to obtain road surface reflectivity feature data, including: Based on preset division rules, the operation section image is divided into multiple operation section sub-images; Calculate the grayscale value of all pixels in each sub-image of the work section, and adjust the filtering window of the adaptive median filtering algorithm according to the degree of difference of the grayscale values ​​to obtain the adjusted filtering window; Based on the adjusted filtering window, the sub-image of the work section is filtered to obtain a filtered sub-image, and all filtered sub-images are stitched together in reverse according to the preset division rules to obtain a de-interference image. Based on the grayscale characteristics of the reflective area of ​​the water accumulation on the urban secondary arterial road in a rainy weather scenario, a grayscale value range for the reflective area of ​​the water accumulation is set, and reflective pixels with grayscale values ​​within the grayscale value range are selected. The coordinate positions of the reflective pixels in the de-interference image are statistically analyzed, and the area of ​​the image region covered by all reflective pixels is calculated. The coordinate positions and the area of ​​the image region are combined to form road surface reflective feature data.

3. The method according to claim 1, characterized in that, The ground distance data is subjected to deviation compensation processing to obtain ground distance correction data. Combined with the road surface reflectivity data, a multi-feature fusion recognition model is used to identify water accumulation areas, including: The detection parameters of the radar detection device mounted on the new energy sanitation vehicle and the real-time rainy weather environmental parameters of the urban secondary trunk road in rainy weather are obtained to determine the radar detection area coordinate system and the real-time air refractive index of the working area of ​​the new energy sanitation vehicle. The detection parameters include the radar detection angle range, the radar detection distance range and the coordinate definition rules of the radar detection point. Based on the real-time air refractive index, the distance deviation value of each radar detection point is calculated. The original detection distance value of each radar detection point is extracted from the ground distance data. Combined with the corresponding distance deviation value, the corrected distance value of each radar detection point is calculated. The corrected distance values ​​of all radar detection points and the corresponding coordinates of the detection points in the radar detection area coordinate system are integrated into ground distance correction data. The coordinates of reflective pixels and the coverage area of ​​reflective pixels are extracted from the road surface reflective feature data; The coordinates of the reflective pixels are associated with the coordinates of the detection points in the ground distance correction data to determine the set of radar detection points corresponding to the coverage area of ​​each reflective pixel, and the average corrected distance within each set of radar detection points is calculated. Based on the reflective pixel coverage area and the corresponding corrected average distance value, combined with the area threshold and distance difference threshold, the water accumulation area is identified through a multi-feature fusion recognition model.

4. The method according to claim 3, characterized in that, Based on the reflective pixel coverage area and the corresponding corrected average distance value, combined with area threshold and distance difference threshold, a multi-feature fusion recognition model identifies the water accumulation area, including: Acquire historical water accumulation identification data of the urban secondary arterial road in rainy weather scenarios and real-time light intensity of the work area. Based on the comparison results between the real-time light intensity and the historical average light value in the historical water accumulation identification data, determine the area threshold. Based on the distance difference range between historical dry road surfaces and historical waterlogged areas in the historical waterlogging identification data, a distance difference threshold that is suitable for the road surface material of the work area is selected from the distance difference range; By using a multi-feature fusion recognition model, reflective pixel coverage areas with a pixel coverage area greater than or equal to an area threshold, or edge pixel grayscale change rate greater than a preset grayscale change threshold, are marked as candidate water accumulation areas. The radar detection device acquires multiple dry road surface detection distance values ​​for the work area, and determines the dry road surface distance value by combining the ground structure parameters of the work area. Calculate the difference between the mean corrected distance value and the actual distance value of the dry road surface for each candidate waterlogged area, and the average difference between the corresponding corrected distance value and the distance value of the dry road surface; Candidate water accumulation areas whose actual distance difference is greater than a distance difference threshold and whose average difference is greater than or equal to a preset difference threshold are marked as suspected water accumulation areas. Any two suspected waterlogged areas that meet the region merging rules of the multi-feature fusion identification model are merged into a waterlogged area to be determined. The waterlogged area is then confirmed by combining the historical morphological characteristics of historical waterlogged areas.

5. The method according to claim 4, characterized in that, Any two suspected waterlogged areas that meet the region merging rules of the multi-feature fusion identification model are merged into a waterlogged area to be determined. Combined with the historical morphological characteristics of historical waterlogged areas, the waterlogged areas are confirmed, including: Based on the first coordinate range and center coordinates of each suspected water accumulation area in the radar detection area coordinate system, the center distance between each suspected water accumulation area and any other suspected water accumulation area is calculated. According to the region merging rules of the multi-feature fusion recognition model, all suspected water accumulation areas are merged until there are no suspected water accumulation areas that meet the region merging rules, and the water accumulation area to be determined is obtained. The region merging rule is that the center distance between each suspected water accumulation area and any other suspected water accumulation area is less than a preset merging distance threshold, and the first coordinate range overlaps or the edge distance is less than a preset edge distance. Historical morphological features of historical waterlogged areas matching the work area are extracted from the historical waterlogged area morphology database. These historical morphological features include the ratio of the major and minor axes of the outline, the actual coverage area, and the ratio of the area of ​​the circumscribed rectangle. Based on the second coordinate range of each water accumulation area to be determined in the radar detection area coordinate system, the actual morphological characteristics of each water accumulation area to be determined are calculated, and the actual morphological characteristics are compared with the historical morphological characteristics to mark all water accumulation areas to be determined that conform to the historical morphological characteristics as water accumulation areas.

6. The method according to claim 1, characterized in that, The actual water accumulation parameters are compared with the water accumulation characteristic data to generate a water accumulation area avoidance instruction, including: The actual water depth, actual water time, and actual water area percentage in the actual water accumulation parameters are compared with the average range of water depth, water time threshold, and water area percentage benchmark value in the water accumulation feature data to determine the avoidance priority of the water accumulation area. The operating position, operating width, and minimum turning radius of the new energy sanitation vehicle are obtained, and the minimum offset distance of the avoidance path is determined by combining the third coordinate range of the water accumulation area. The detour length of the avoidance path is calculated based on the minimum offset distance and the length of the water accumulation area; Based on the operating location, third coordinate range, minimum offset distance, and detour length of the new energy sanitation vehicle, the starting point, turning point, and ending point of the avoidance path are determined, forming a set of avoidance path points. The avoidance priority, avoidance path point set, minimum offset distance, and detour length are used to generate a water accumulation area avoidance command to bypass the water accumulation area.

7. The method according to claim 1, characterized in that, Based on the actual water accumulation parameters, the discharge parameters of the vehicle battery are adjusted to obtain the adjusted discharge parameters, including: The first adjustment coefficient is set based on the correlation between the actual water depth and the road surface material of the waterlogged area in the actual water accumulation parameters. A second adjustment coefficient is set based on the matching relationship between the water distribution density in the actual water accumulation parameters and the preset load level of the cleaning execution component; Based on the remaining charge and temperature of the vehicle battery, a basic adjustment strategy is determined, which includes a normal adjustment mode, a balance adjustment mode, and an energy-saving adjustment mode. Under the constraint of the aforementioned basic adjustment strategy, the standard discharge current of the vehicle battery and the first adjustment coefficient are multiplied to obtain the preliminary adjustment current, and the standard discharge voltage of the vehicle battery and the second adjustment coefficient are multiplied to obtain the preliminary adjustment voltage. Based on the ratio of the area of ​​the waterlogged area to the working width of the vehicle, and in conjunction with the preset load level of the sweeping execution components, a power correction value is set. Based on the initial adjustment current, the initial adjustment voltage, and the power correction value, the initial output power is calculated. According to the battery temperature, the initial output power is adjusted to obtain the adjusted output power. The adjusted output power, the initial adjusted current, and the initial adjusted voltage are combined to form the discharge parameters of the vehicle battery. The compatibility of the discharge parameters is verified. If the discharge parameters are within the rated input parameter range of the vehicle motor, then the discharge parameters are used as the adjusted discharge parameters. If the discharge parameters exceed the rated input parameter range, then the discharge parameters are adjusted to the rated input parameter range to obtain the adjusted discharge parameters.

8. A new energy sanitation vehicle cleaning control system integrating vision and radar, characterized in that, include: The acquisition module is used to acquire water accumulation characteristic data, images of the work section, and ground distance data of urban secondary arterial roads in rainy weather scenarios. The filtering module is used to filter out interference areas in the image of the work section using an adaptive median filtering algorithm to obtain road surface reflective feature data. The compensation module is used to perform deviation compensation processing on the ground distance data to obtain ground distance correction data. Combined with the road surface reflective feature data, the water accumulation area is identified through a multi-feature fusion recognition model. The generation module is used to control the depth detection component mounted on the new energy sanitation vehicle to detect the actual water accumulation parameters of the water accumulation area, compare the actual water accumulation parameters with the water accumulation feature data, and generate a water accumulation area avoidance command to bypass the water accumulation area. The adjustment module is used to adjust the discharge parameters of the vehicle battery according to the actual water accumulation parameters, so as to obtain the adjusted discharge parameters. The cleaning module is used to control the cleaning execution component to perform cleaning operations according to the water accumulation area avoidance command and the adjusted discharge parameters, and to collect the load data of the cleaning execution component during the cleaning process. If the load data exceeds the preset load range, the adjusted discharge parameters are optimized based on the actual water accumulation parameters until the load data is within the preset load range, so as to complete the autonomous cleaning of the urban secondary road in rainy weather.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the new energy sanitation vehicle cleaning control method integrating vision and radar as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a new energy sanitation vehicle cleaning control method that integrates vision and radar as described in any one of claims 1 to 7.

Images