Method for adjusting the operating speed of sanitation vehicles based on garbage coverage rate and sanitation vehicles

Abstract

This invention discloses a method for adjusting the operating speed of a sanitation vehicle based on garbage coverage rate, and a sanitation vehicle itself. The method includes the following steps: acquiring image data within a preset area in front of the vehicle; preprocessing the image data, including dynamic ROI delineation, image enhancement and denoising, and brightness normalization; using a deep learning model based on YOLOv5s to detect garbage in the preprocessed image and outputting target boxes for the coverage area; calculating a two-layer coverage rate based on the target detection results and the Otsu threshold segmentation method, combined with color features, and outputting the current garbage coverage rate; and adjusting the operating speed of the sanitation vehicle according to the garbage coverage rate based on preset rules. This invention can improve the accuracy of road garbage coverage rate identification and significantly optimize vehicle operating energy consumption and operating costs.

Description

Technical Field

[0001] This invention belongs to the field of sanitation vehicle technology, and specifically relates to a method for adjusting the operating speed of sanitation vehicles based on garbage coverage rate, and a sanitation vehicle. Background Technology

[0002] Currently, sanitation sweeper trucks are widely used as the main tools for urban road cleaning, applied in various environments such as main roads, secondary roads, and residential areas. To adapt to different cleaning scenarios, existing sweeper trucks generally have multiple operating modes, such as a "cleaning mode" for light waste like fallen leaves and an "enhanced mode" for heavy waste like mud and sand, and control the vehicle's cleaning system by setting a constant speed. However, existing technologies generally use fixed operating speeds or rely on manual intervention for adjustment, which has the following shortcomings: poor matching between power output and waste load. In cleaning mode, the vehicle speed is usually fixed within a certain range (e.g., 1500–1550 rpm), but in actual operation, the density of road waste varies greatly. Fixing a high speed can easily lead to excessive power output and wasted energy. Manual intervention is inefficient and has a slow response. Although some sweeper trucks allow drivers to manually adjust the speed, this method requires drivers to make judgments and operate based on experience, which cannot achieve rapid and accurate response to complex road conditions (such as mixed distribution of different types of waste).

[0003] Therefore, there is an urgent need for a control method and device that can automatically adjust the operating speed based on the real-time road surface garbage coverage rate, so as to achieve precise matching between power output and operating load, thereby reducing energy consumption, improving operating efficiency, and adapting to more complex and changeable cleaning environments.

[0004] Based on this, Chinese invention patent application CN108776439A discloses a control method, control system, operating device, and sanitation vehicle for a sanitation vehicle operating device. The control method for the operating device includes acquiring road surface condition information during the operation of the sanitation vehicle; acquiring vehicle speed information during the operation of the sanitation vehicle; and controlling the speed of the auxiliary engine and / or the opening and closing of the suction nozzle air duct of the sanitation vehicle according to the road surface condition information and the vehicle speed information.

[0005] The above solution achieves automatic control of the power unit on the sanitation vehicle, but it still has the following problems: the garbage recognition is coarse, and the accuracy is low due to the use of traditional image statistics and binarization methods. It is also poorly adaptable to situations such as fallen leaves, reflections, and slippery conditions. The "coarse-grained interval + fixed mapping" approach cannot accurately match complex operating scenarios and is difficult to dynamically respond to continuously changing garbage distribution. The control mechanism is rigid, and the jump is based solely on rule judgment. The rotation speed cannot be flexibly adjusted, and it is difficult to suppress frequent switching or stable drift. Summary of the Invention

[0006] This invention provides a method for adjusting the operating speed of sanitation vehicles based on garbage coverage rate, as well as a sanitation vehicle itself. Through embedded platform adaptation and garbage identification technology optimized by a custom large-model operator, it achieves synergistic optimization of efficient end-side inference and flexible speed control. It aims to solve the problems of poor adaptability, low accuracy, coarse control granularity, and rigid control mechanisms in existing technologies.

[0007] To solve the above-mentioned technical problems, the sanitation vehicle operating speed adjustment method proposed in this invention includes the following steps: Hardware adaptation layer initialization: Image data of a preset area in front of the vehicle is collected. The image data is preprocessed, including dynamic ROI delineation, image enhancement and denoising, and brightness normalization. A deep learning model based on YOLOv5s is used to detect garbage in preprocessed images and output bounding boxes of the covered areas. Based on the target detection results and the Otsu threshold segmentation method, a two-layer coverage rate calculation is performed by combining color features, and the current garbage coverage rate is output. The operating speed of the sanitation vehicles is adjusted based on the garbage coverage rate and preset rules.

[0008] Preferably, the dynamic ROI delineation specifically involves: Based on the installation parameters of the image acquisition device, a spatial coordinate system is established to delineate a rectangular recognition area within a preset distance in front of the vehicle. The image coordinates are then converted into orthographic projection area coordinates under the road surface plane coordinate system through perspective transformation.

[0009] Preferably, the operating speed is adjusted by establishing a three-state finite state machine for the operating speed, and switching the operating speed based on the garbage coverage rate: When the detection coverage is between 1% and 5% for several consecutive frames, switch to medium speed mode; When the coverage rate of several consecutive detections is greater than or equal to 5%, switch to high speed mode; If the detection coverage is less than 1% for several consecutive frames, switch to low speed mode.

[0010] Preferably, the operating speed is set with a holding time threshold after switching, and the state rollback is not performed when the coverage does not continuously meet the switching conditions.

[0011] Preferably, the double-layer coverage rate is calculated as follows: The area of ​​a single piece of waste is calculated based on the coordinates of the target box, and the sum is used to obtain the first area. The second area of ​​all garbage regions in the preprocessed image is extracted by combining the Otsu threshold segmentation algorithm with HSV color filtering. The final garbage coverage rate is calculated by weighting the first area and the second area together with their corresponding weights.

[0012] Preferably, before outputting the current garbage coverage rate, the following processing steps are also included: Model dynamic background targets appearing in the image; if a dynamic target is detected at the same location in multiple consecutive frames, then the coverage calculation of that area is masked; and / or The coverage results calculated from multiple consecutive frames of images are subjected to time series filtering to smooth short-term fluctuations and output stable coverage parameters.

[0013] Preferably, the time series filtering performs Kalman filtering on coverage data of several consecutive frames, and the smoothing coefficient of the Kalman filter is set to 0.8.

[0014] Preferably, during the switching of the operating speed, an incremental PID algorithm is used to achieve flexible control. In the incremental PID algorithm, the proportional coefficient is 0.6, the integral coefficient is 0.1, and the derivative coefficient is 0.3.

[0015] Preferably, the switching strategy for the operating speed also includes a minimum time interval for speed adjustment.

[0016] In another aspect, the present invention also provides a sanitation vehicle, including a vehicle chassis, a power system, a sweeping and washing device, and a superstructure controller. The sanitation vehicle also includes a forward-facing camera for acquiring images, and an intelligent control system for identifying the road surface garbage coverage rate and dynamically adjusting the operating speed accordingly. The intelligent control system executes the above-mentioned operating speed adjustment method and automatically switches the speed of the engine or sweeping device based on the detected garbage coverage rate.

[0017] Compared with the prior art, the present invention has the following technical effects: 1. The operating speed adjustment method proposed in this invention acquires real-time images of the front of the vehicle using a forward-looking camera, and identifies the distribution density of leaves and garbage on the road surface using a YOLOv5s deep learning model. Based on the coverage rate, it accurately judges the operating load and dynamically adjusts the operating speed, effectively avoiding the power excess or deficiency caused by traditional fixed gear control, and achieving precise matching of power output with actual garbage working conditions.

[0018] 2. The operating speed adjustment method proposed in this invention adopts a graded state control strategy based on coverage. It automatically switches to a low-speed operating mode in areas with sparse garbage and increases the operating intensity in areas with dense garbage. It effectively controls airflow disturbance, reduces garbage splashing, and can reduce fuel consumption and optimize the overall vehicle operating efficiency.

[0019] 3. The operating speed adjustment method proposed in this invention constructs a three-state finite state machine control structure, judges state switching based on coverage continuity and hold time mechanism, and avoids instability caused by frequent jumps; at the same time, an incremental PID algorithm is introduced in the speed control process to achieve smooth transition, effectively protect the engine and improve operation smoothness.

[0020] 4. The proposed operating speed adjustment method introduces dynamic background modeling and time-series filtering mechanisms, significantly improving the system's anti-interference capability and recognition stability in actual complex operating environments. By using a Gaussian mixture model to construct a road surface background model for continuous image frames, when a dynamic target is detected to continuously appear in a specific area, the system can automatically shield the coverage calculation of that area, effectively avoiding the mistaken inclusion of non-garbage targets in the recognition area, and improving the accuracy and robustness of the overall recognition results. At the same time, the Kalman filter algorithm is used to perform time-series smoothing processing on the coverage output, making the final output coverage parameter more stable and continuous, with fluctuation amplitude controlled within 0.3%, preventing frequent state switching or control misjudgment caused by instantaneous image changes. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the operating speed adjustment method described in this invention; Figure 2 This is a schematic diagram of image acquisition as described in an embodiment of the present invention; Figure 3 This is a schematic diagram of a 1% garbage coverage rate as described in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the 4% waste coverage rate described in an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating the 6% garbage coverage rate described in an embodiment of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present application and with reference to the accompanying drawings.

[0023] Example 1 This embodiment describes a method for adjusting the operating speed of sanitation vehicles based on garbage coverage rate. Figure 1 As shown, it includes the following steps S1 to S5: S1: Hardware adaptation layer initialization, collecting image data in a preset area in front of the vehicle.

[0024] S2: Preprocess the image data, including dynamic ROI (Region of Interest) delineation, image enhancement and denoising, and brightness normalization.

[0025] S3: Utilize a deep learning model based on YOLOv5s to detect garbage in preprocessed images. The model embeds multiple custom operators for sanitation scenarios and outputs target bounding boxes for the covered areas.

[0026] S4: Based on the target detection results and Otsu threshold segmentation method (Otsu method or Otsu inter-class variance method), and combined with color features, perform two-layer coverage calculation and output the current garbage coverage rate.

[0027] S5: Adjust the operating speed of the sanitation vehicle based on the garbage coverage rate and preset rules.

[0028] In this embodiment, step S1 is completed by the forward-looking camera, mainly including an intelligent controller, a forward-looking camera, and an on-board controller. The specific hardware used in this embodiment is an embedded platform, whose main computing chip is a quad-core Cortex-A55 processor and an NXP S32K144 chip. A customized inference engine is deployed to control the image input latency to within 20ms. Image data of a preset area in front of the vehicle is collected through the forward-looking camera. The camera is a 2-megapixel industrial-grade wide-angle CMOS camera with a resolution ≥1280×720 pixels, a frame rate of 25-30 frames / second, a top-down angle of 15°-30°, supports a wide operating temperature range of -40℃ to 85℃, and has an EMC protection level of IP67, making it suitable for bumpy operation scenarios of sanitation vehicles.

[0029] Image acquisition such as Figure 2 As shown, the sanitation truck on the left is traveling to the right. A forward-facing camera captures images of the area in front of the vehicle, and an intelligent controller processes the image information to identify the extent of garbage coverage. The forward-facing camera is mounted on the front of the sanitation truck or above the driver's cab, facing directly in the direction of travel. The downward angle is set to 15°–30°, adjustable to accommodate different vehicle heights and working widths. An industrial-grade wide-angle CMOS camera or a high-resolution vehicle-mounted camera can be used, with a recommended resolution of at least 1280×720 pixels and a frame rate of 25–30 frames per second to ensure continuous and clear image acquisition. The camera is equipped with automatic exposure and white balance adjustment functions to adapt to changes in natural light at different times of day (such as morning, night, and shaded areas).

[0030] In other embodiments of the present invention, the image acquisition system can combine GPS / vehicle speed sensor information to dynamically adjust the acquisition frequency or synchronize the vehicle's motion state, thereby improving the spatiotemporal consistency of data between consecutive frames.

[0031] In this embodiment, the hardware mainly consists of an intelligent controller, a forward-facing camera, and an onboard controller. For example... Figure 2As shown, the sanitation vehicle on the left is traveling to the right. It uses a front-view camera to collect images of the area in front of the vehicle, and the intelligent controller processes the image information to identify the garbage coverage in front. The upper controller receives the speed value transmitted by the intelligent controller and adjusts the vehicle speed accordingly.

[0032] Step S2 involves preprocessing the image data collected in front of the vehicle, including two sub-steps: dynamic ROI delineation and multimodal image enhancement processing, to improve the accuracy and robustness of subsequent waste detection and coverage calculation.

[0033] In this step, the dynamic ROI delineation specifically involves: To ensure that the analysis only covers the target area within the sanitation vehicle's travel path, the system establishes a forward spatial coordinate system based on the installation parameters of the forward-facing camera (such as installation height, pitch angle, focal length, and field of view). A rectangular recognition area is delineated within a preset distance in front of the vehicle. To avoid image distortion affecting recognition accuracy, the system performs perspective transformation on the captured image after acquisition, converting the image coordinates into orthographic projection coordinates of the area in the road surface plane coordinate system. The specific method for delineating the region of interest is as follows: The recognition distance is set between 7 and 13.5 meters in front of the vehicle. The defined ROI area has a fixed length of 6.5 meters, and its width is automatically matched according to the actual working width of the vehicle-mounted sweeping disc (e.g., 2.0 to 3.5 meters). Based on the camera's viewing angle and installation position, the system uses homography transformation or inverse perspective transformation to perform perspective correction processing on the original image, converting the target area in the oblique image into a ground orthographic projection image. The perspective transformation uses calibration data of image coordinates and actual ground coordinate points to generate a transformation matrix, ensuring that the ROI area is consistent with the true geometric relationship of the cleaning path and avoiding image distortion from causing deviations in the recognition results. When the vehicle turns left or right or deviates from its path, the system can predict the vehicle's pose shift based on vehicle speed and steering wheel angle, dynamically adjusting the ROI position and width in real time to ensure that the recognition area always remains aligned with the actual cleaning path.

[0034] To ensure the stability of image recognition under different lighting conditions, this embodiment introduces various image enhancement techniques, including: Contrast Limited Adaptive Histogram Equalization (CLAHE) is employed to enhance local contrast in the acquired images, particularly in low-light environments such as shadows, dusk, and nighttime, improving the contrast between leaf outlines and the background and enhancing target visibility. Gaussian filtering is combined for noise reduction: Gaussian blur is applied to the image to filter out high-frequency noise points and false edges, improving the model's robustness to garbage target edges. This embodiment also incorporates bilateral filtering to preserve garbage edge features, with a 5×5 kernel size and a Gaussian kernel function standard deviation in the spatial domain. Standard deviation of Gaussian kernel function over pixel value range .

[0035] To address interference from strong light reflections (such as on wet roads in sunny weather or under vehicle headlights), a brightness adjustment algorithm based on the HSV color space is employed. This algorithm limits the V channel (brightness channel) of the image to a preset range [30, 220], preventing highlights or excessively dark areas from obscuring the garbage target and eliminating interference from strong light reflections and shadows for identification. This ensures the stability of garbage features under different lighting conditions. Further color balancing and saturation limiting can be applied to the normalized image to improve the distinguishability between garbage color and road surface color.

[0036] In this embodiment, step S3 involves detecting garbage targets on the image data after image preprocessing and outputting the target bounding box of the coverage area for subsequent coverage calculation and operation speed adjustment.

[0037] Specifically, it includes the following two sub-steps: first, constructing and using a lightweight deep learning detection model to identify garbage in the image; and second, calculating the coverage based on the identification results combined with image segmentation technology.

[0038] To adapt to the diversity of leaf litter and the need for real-time identification in sanitation operations, this embodiment adopts a lightweight waste detection model built on the YOLOv5s framework, which has the dual advantages of high recognition rate and high inference speed.

[0039] The specific model training and deployment process is as follows: Training samples are collected, and Mosaic data augmentation technology is introduced during the training process. By randomly stitching multiple images together, various garbage mixed and piled-up scenarios are simulated to improve the model's ability to detect small targets in complex backgrounds. The Focus structure in YOLOv5 is used to perform piecewise rearrangement processing on the input image (size 1024×512 pixels). That is, the image is divided into 4 sub-images (256×128 pixels) with a stride of 2, and then the sub-images are stitched together through channels to form a 512×256×12-dimensional feature map.

[0040] The optimized model achieves an inference speed of 65 FPS on the RTX3060 platform, adapting to the real-time recognition requirements of edge devices. The output detection results include the location information and category confidence of each target box.

[0041] Step S4 involves calculating the dual-layer fusion coverage based on the results output by the YOLOv5s target detection model and image segmentation algorithm, and combining this with anti-interference processing methods to finally output stable and reliable garbage coverage parameters, which are used to drive subsequent operation speed control.

[0042] The calculation of the double-layer coverage rate is as follows: The area of ​​a single piece of waste is calculated based on the coordinates of the target box, and the sum is used to obtain the first area. The second area of ​​all garbage regions in the preprocessed image is extracted by combining the Otsu threshold segmentation algorithm with HSV color filtering. The final garbage coverage rate is calculated by weighting the first area and the second area together with their corresponding weights.

[0043] Final coverage The calculation formula is:

[0044] In the formula, The first area corresponds to the total pixel area of ​​the object detection box in the deep learning model in step S3. The pixel area of ​​the suspected leaf region after HSV segmentation. This represents the total number of pixels in the ROI region (i.e., the coverage calculation area). These are the weighted coefficients for the detection and segmentation results, respectively. In one embodiment of the present invention, The value is 0.7. The value is 0.3.

[0045] Before outputting the current garbage coverage rate, the following processing steps are included: Model dynamic background targets appearing in the image; if a dynamic target is detected at the same location in multiple consecutive frames, then the coverage calculation of that area is masked; and / or The coverage results calculated from multiple consecutive frames of images are subjected to time series filtering to smooth short-term fluctuations and output stable coverage parameters.

[0046] Specifically, the purpose of modeling dynamic background targets is to reduce the interference of non-garbage dynamic targets (such as pedestrians, vehicles, animals, etc.) on coverage calculation and improve the system's resistance to environmental interference in real road scenarios. A dynamic background modeling mechanism is introduced, and the specific steps are as follows: A background update model is built for the image sequence using a Gaussian Mixture Model (GMM). When a dynamic target with obvious shape changes is detected at the same location in more than 3 consecutive frames, the system determines that the area is a background interference area. This area will be automatically masked in the coverage calculation of this frame and subsequent frames to avoid misjudgment caused by dynamic targets; The background model is continuously updated for each frame to adapt to changes in the background as the vehicle travels on the road.

[0047] The purpose of time series filtering is to avoid abnormal fluctuations in coverage caused by image acquisition and recognition errors or short-term garbage distribution fluctuations. Time series smoothing is added before the final output coverage parameter, as follows: The coverage results of five consecutive frames are fused using a Kalman filter. The smoothing coefficient is set to 0.8 to emphasize the dominant role of historical trends in current estimates; The coverage fluctuation of the filter output is controlled within 0.3% to ensure stable system judgment.

[0048] Step S5 involves dynamically adjusting the operating speed of the sweeper truck based on the stable garbage coverage parameters output in step S4 and the set operation control rules, so as to achieve intelligent matching between the power system and the garbage load in front, thereby improving energy efficiency and operation results.

[0049] In this embodiment, a three-state finite state machine (FSM) is established for operation speed control. Based on the garbage coverage rate, three speed states are set: medium speed mode (1350 rpm), high speed mode (1500 rpm), and low speed mode (1270 rpm).

[0050] The state switching rules are as follows: When the coverage results of several consecutive frames or more are between [1%, 5%), switch from low speed mode to medium speed mode. When the coverage result of several consecutive frames or more is ≥5%, switch directly from low speed mode or medium speed mode to high speed mode. When the coverage result of several consecutive frames is less than 1%, switch back to low speed mode from medium speed mode or high speed mode.

[0051] Examples of different waste coverage rates are as follows Figure 3-5As shown, an example of speed switching is as follows: when switching from 1270rpm to 1350rpm, the speed of 1350rpm will be maintained for 6 seconds. As long as the coverage reaches the 1350rpm requirement, another 6 seconds will be added at the current time. Only if the coverage does not meet the 1350rpm requirement for a duration of 6 seconds will it switch back to 1270rpm. At the same time, if the coverage reaches the 1500rpm level within the 6-second duration, it will be immediately increased to 1500rpm, and the 6-second strategy will be maintained simultaneously.

[0052] To avoid frequent switching of operating speed due to short-term fluctuations in coverage, which could affect engine stability and cleaning performance, a holding time threshold is set for the operating speed after switching. If the coverage does not continuously meet the switching conditions, a state rollback will not be executed. The specific mechanism is as follows: After switching to medium or high speed mode, the system starts a 6-second countdown. If the current state condition is still met during the countdown (e.g., coverage remains greater than or equal to 5%), then the timer is reset. If the coverage no longer meets the entry conditions for this state within 6 seconds and the countdown reaches zero, the system will perform a downgrade rollback. The highest priority is given to the condition that the coverage rate is greater than or equal to 5%. Once this condition is met, the system can immediately jump from any state to S3 without waiting for the countdown to end.

[0053] To ensure a smooth transition and avoid drastic fluctuations during speed adjustment, this embodiment employs an incremental PID control algorithm to flexibly adjust the target speed. The calculation model is as follows:

[0054]

[0055] In the formula, For the current moment Control the output increment. For the error at the current moment, we have This represents the difference between the current desired speed and the actual speed. and Each is the previous (time) ), the time before last (time) The error value; This is the proportional coefficient, used to control the degree of response of the system to the current error, and how fast the adjustment speed is; These are integral coefficients used to eliminate steady-state error and response error accumulation. These are differential coefficients used to predict error trends, suppress oscillations, and improve stability. The current control quantity (speed) is output. For the last (time) Control quantity (speed) output.

[0056] In this embodiment, the proportional coefficient is 0.6, the integral coefficient is 0.1, and the derivative coefficient is 0.3.

[0057] To further protect the engine and prevent frequent adjustments from impacting mechanical components, the operating speed switching strategy also includes a minimum time interval for speed adjustments. In this embodiment, the minimum time interval for speed switching is set to 2 seconds, meaning that after any speed change, the next speed switching operation must wait at least 2 seconds before it can be triggered.

[0058] Example 2 This embodiment is a sanitation vehicle. The system is used to implement the operating speed adjustment method as described in Embodiment 1, and includes a vehicle chassis, a power system, a sweeping and cleaning device, and a superstructure controller. The sanitation vehicle also includes a forward-facing camera for acquiring images, and an intelligent control system for identifying the road surface garbage coverage rate and dynamically adjusting the operating speed accordingly. The intelligent control system executes the operating speed adjustment method as described in Embodiment 1, and automatically switches the speed of the engine or sweeping device according to the detected garbage coverage rate.

[0059] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for adjusting the operating speed of sanitation vehicles based on garbage coverage rate, characterized in that, Includes the following steps: Hardware adaptation layer initialization: Image data of a preset area in front of the vehicle is collected. The image data is preprocessed, including dynamic ROI delineation, image enhancement and denoising, and brightness normalization. A deep learning model based on YOLOv5s is used to detect garbage in preprocessed images and output bounding boxes of the covered areas. Based on the target detection results and the Otsu threshold segmentation method, a two-layer coverage rate calculation is performed by combining color features, and the current garbage coverage rate is output. The operating speed of the sanitation vehicles is adjusted based on the garbage coverage rate and preset rules.

2. The method according to claim 1, characterized in that, The dynamic ROI delineation specifically refers to: Based on the installation parameters of the image acquisition device, a spatial coordinate system is established to delineate a rectangular recognition area within a preset distance in front of the vehicle. The image coordinates are then converted into orthographic projection area coordinates under the road surface plane coordinate system through perspective transformation.

3. The method according to claim 1, characterized in that, The operating speed is adjusted by establishing a three-state finite state machine based on the garbage coverage rate, and the operating speed is switched accordingly. When the detection coverage is between 1% and 5% for several consecutive frames, switch to medium speed mode; When the coverage rate of several consecutive detections is greater than or equal to 5%, switch to high speed mode; If the detection coverage is less than 1% for several consecutive frames, switch to low speed mode.

4. The method according to claim 3, characterized in that, The operating speed is set with a hold time threshold after switching. If the coverage does not continuously meet the switching conditions, the state rollback will not be performed.

5. The method according to claim 1, characterized in that, The calculation of the double-layer coverage rate is as follows: The area of ​​a single piece of waste is calculated based on the coordinates of the target box, and the sum is used to obtain the first area. The second area of ​​all garbage regions in the preprocessed image is extracted by combining the Otsu threshold segmentation algorithm with HSV color filtering. The final garbage coverage rate is calculated by weighting the first area and the second area together with their corresponding weights.

6. The method according to claim 1, characterized in that, Before outputting the current garbage coverage rate, the following processing steps are included: Model dynamic background targets appearing in the image; if a dynamic target is detected at the same location in multiple consecutive frames, then the coverage calculation of that area is masked; and / or The coverage results calculated from multiple consecutive frames of images are subjected to time series filtering to smooth short-term fluctuations and output stable coverage parameters.

7. The method according to claim 6, characterized in that, The time series filtering performs Kalman filtering on coverage data for several consecutive frames, and the smoothing coefficient of the Kalman filter is set to 0.

8.

8. The method according to claim 1, characterized in that, The switching process of the operating speed adopts an incremental PID algorithm to achieve flexible control. In the incremental PID algorithm, the proportional coefficient is 0.6, the integral coefficient is 0.1, and the derivative coefficient is 0.

3.

9. The method according to claim 1, characterized in that, The switching strategy for the operating speed also includes a minimum time interval for speed adjustment.

10. A sanitation vehicle, comprising a vehicle chassis, a power system, a sweeping and washing device, and a superstructure controller; Its features are, The sanitation vehicle also includes a forward-facing camera for acquiring images, and an intelligent control system for identifying the road surface garbage coverage rate and dynamically adjusting the operating speed accordingly. The intelligent control system executes the operating speed adjustment method as described in any one of claims 1 to 9, and automatically switches the speed of the engine or sweeping device according to the detected garbage coverage rate.

Images