Sweeper cleaning brush control method and system based on backbone image recognition and medium

Abstract

The invention relates to the technical field of image processing, in particular to a control method and system for a sweeping brush of a sweeping vehicle based on backbone image recognition and a medium, and the method comprises the steps: obtaining a road surface image, and extracting bounding boxes of various types of garbage in the road image; determining the garbage type probability distribution of pixels in each bounding box, and calculating the cross degree of each type of garbage in combination with the garbage type probability distribution and the pixel position of each type of garbage; establishing pollution weights of the bounding boxes based on the garbage type of each pixel, and determining garbage concentration of each bounding box based on the cross degree and the pollution weight of each type of garbage in each bounding box; performing clustering analysis on each bounding box based on the garbage concentration and position distribution of each bounding box to obtain a cleaning area; determining the cleaning difficulty of the cleaning area based on the garbage concentration of each boundary frame in the cleaning area, and controlling the angle and rotating speed of the cleaning brush in each cleaning area according to the cleaning difficulty; efficient and accurate sweeping can be achieved, and the sweeping efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method, system and medium for controlling a cleaning brush of a cleaning vehicle based on basic image recognition. Background Art

[0002] When cleaning road debris, street sweepers often struggle to accurately identify the distribution of debris, resulting in low cleaning efficiency. Traditional methods rely on manual experience, lack intelligent control, and are difficult to adapt to complex and changing cleaning environments.

[0003] In related technologies, image recognition technology has been preliminarily applied in garbage identification and cleaning path planning, but intelligent adjustment of the cleaning brush angle and speed has not yet been achieved, resulting in unstable cleaning effects. Summary of the Invention

[0004] To solve the above problems, the present invention provides a method, system and medium for controlling the cleaning brush of a sweeper based on image recognition. The method aims to automatically analyze the distribution and concentration of garbage through image recognition technology, intelligently adjust the angle and speed of the brush, achieve efficient and accurate cleaning, and improve cleaning efficiency.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] In one aspect, an embodiment of the present invention provides a method for controlling a sweeping brush of a sweeper based on image recognition, the method comprising the following steps:

[0007] Obtain road images and use the object detection model to extract the bounding boxes of various types of garbage in the road images;

[0008] Determine the probability distribution of garbage types for pixels within each bounding box, and calculate the intersection degree of each garbage type within the bounding box based on the garbage type probability distribution and the pixel position of each garbage type;

[0009] The pollution weight of the bounding box is established based on the garbage type of each pixel in the bounding box, and the garbage concentration of each bounding box is determined based on the intersection degree and pollution weight of each type of garbage in each bounding box;

[0010] Perform cluster analysis on each bounding box based on its garbage concentration and location distribution to obtain the cleaning area;

[0011] The cleaning difficulty of the cleaning area is determined based on the garbage concentration of each bounding box in the cleaning area, and the angle and speed of the cleaning brush in each cleaning area are controlled according to the cleaning difficulty.

[0012] Optionally, determining the probability distribution of garbage types of pixels within each bounding box and calculating the intersection degree of garbage within the bounding box in combination with the probability distribution of garbage types and the pixel positions of each type of garbage may include:

[0013] The probability of the target type of garbage in the bounding box is obtained by summing the ratio of the frequency of each pixel belonging to the target type of garbage to the total number of pixels in the bounding box;

[0014] Calculate the pixel position difference of the target type garbage in the two bounding boxes to obtain the pixel difference of the target type garbage in the two bounding boxes;

[0015] The probability of each target type of garbage in the two bounding boxes is multiplied by the pixel difference and then added together to obtain the intersection degree of the target type of garbage in the two bounding boxes. The intersection degree of each target type of garbage in the bounding box is accumulated to obtain the intersection degree of each type of garbage in the bounding box.

[0016] Optionally, establishing a pollution weight for the bounding box based on the garbage type of each pixel within the bounding box, and determining the garbage concentration of each bounding box based on the intersection degree and pollution weight of each type of garbage within each bounding box, includes:

[0017] Set a basic weight coefficient based on the type of garbage, and multiply the garbage concentration of each type of garbage within the boundary box by the corresponding basic weight coefficient to obtain the pollution weight of each type of garbage;

[0018] The pollution weights of each type of garbage in the bounding box and their intersection are weighted and summed to obtain the comprehensive pollution index of the bounding box;

[0019] Based on the comprehensive pollution index of the bounding box and the area of ​​the bounding box, the garbage concentration value per unit area is calculated as the garbage concentration of the bounding box.

[0020] Optionally, performing cluster analysis on each bounding box based on the garbage concentration and location distribution of each bounding box to obtain a cleaning area includes:

[0021] The sweeping radius is set based on the average value of the brush coverage area and the garbage concentration of each bounding box. The minimum number of points in the road image is set based on the minimum amount of garbage to be considered as a covered area. The cluster center is determined based on the garbage concentration of each bounding box.

[0022] The cleaning radius is used as the cluster boundary of the cluster center, and the adjacent bounding box of the cluster center is clustered within the minimum number of points to form a cleaning area;

[0023] The cleaning area is divided into multiple sub-areas, and the local garbage concentration of each sub-area is determined based on the area ratio and garbage concentration of each bounding box in the sub-area;

[0024] Determine the distance between each sub-region and the cluster center, select the sub-region with the largest product of local garbage concentration and distance as the new cluster center, and determine whether the change in distance to the cluster center is below a distance threshold. If so, execute S450; if not, execute S420;

[0025] The garbage concentration of each bounding box in the cleaning area is calculated by gradient, and multiple bounding boxes with larger gradients are used as reference boxes to generate a circumscribed curve containing multiple reference boxes. With the cluster center as the center of the circle, a cleaning area containing the circumscribed curve is generated within the cleaning radius and the minimum number of points.

[0026] Optionally, the method uses the cleaning radius as the cluster boundary of the cluster center, clustering the adjacent bounding box of the cluster center within a minimum number of points to form a cleaning area, including:

[0027] Filter neighboring bounding boxes based on the cluster center and cleaning radius to ensure that the number of bounding boxes in each cluster is not less than the minimum number of points to form a cleaning unit;

[0028] If there are overlapping bounding boxes on the edge of the cleaning unit, the bounding box is divided into the cleaning unit with a larger pixel ratio, the position of the cleaning unit is adjusted based on the bounding box in the cleaning unit, and the adjusted cleaning unit is used as the cleaning area.

[0029] Optionally, determining the garbage concentration in the cleaning area and controlling the angle and rotation speed of the cleaning brush in each cleaning area according to the garbage concentration may include:

[0030] The garbage concentration of each bounding box in the cleaning area is accumulated and then normalized exponentially to obtain the cleaning difficulty of the cleaning area;

[0031] Acquiring a range value of a bristle angle, establishing a first functional relationship between cleaning difficulty and bristle angle within the range value; and adjusting the bristle angle within the range value based on the cleaning difficulty;

[0032] Obtain the speed of the sweeping brush, the driving speed of the sweeping vehicle, the baseline speed of the sweeping vehicle, and the basic speed of the sweeping brush, and determine the speed of the sweeping brush based on the difficulty of garbage cleaning.

[0033] On the other hand, an embodiment of the present invention provides a sweeper brush control system based on image recognition, comprising:

[0034] at least one processor;

[0035] at least one memory for storing at least one program;

[0036] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0037] On the other hand, an embodiment of the present invention provides a computer-readable storage medium storing a program executable by a processor. When the program is executed by the processor, it is used to perform the above method.

[0038] The beneficial effects of the present invention are as follows: the present invention discloses a method, system, and medium for controlling the sweeping brush of a road sweeper based on image recognition. The method first acquires a road surface image and extracts bounding boxes of various types of garbage in the road image. The method then determines the probability distribution of garbage types for pixels within each bounding box and calculates the intersection of each type of garbage based on the probability distribution of garbage types and the pixel positions of each type of garbage, thereby accurately identifying the garbage distribution. Next, a pollution weight is established for the bounding box based on the garbage type of each pixel. The garbage concentration of each bounding box is determined based on the intersection and pollution weight of each type of garbage within each bounding box, thereby accurately dividing the cleaning area. Cluster analysis is performed on each bounding box based on the garbage concentration and position distribution of each bounding box to obtain the cleaning area. This maximizes cleaning efficiency. The cleaning difficulty of the cleaning area is determined based on the garbage concentration of each bounding box within the cleaning area. The angle and speed of the sweeping brush in each cleaning area are controlled according to the cleaning difficulty, thereby accurately adjusting the cleaning force, ensuring efficient garbage removal, and improving cleaning quality. The present invention can achieve efficient and accurate cleaning and improve cleaning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a flow chart of a method for controlling a sweeping brush of a sweeper based on image recognition according to an embodiment of the present invention;

[0041] Figure 2 It is a structural diagram of a cleaning brush control system for a cleaning vehicle based on image recognition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects disclosed in the present invention, so as to fully understand the purpose, scheme and effect disclosed in the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other unless there is a conflict.

[0043] This invention aims to achieve intelligent control of the sweeping brush through image recognition technology, improving cleaning efficiency and reducing manual intervention. Through real-time image analysis, the sweeping brush angle and speed are automatically adjusted to ensure comprehensive and even coverage, improving operational convenience and achieving efficient and intelligent cleaning.

[0044] refer to Figure 1 ,like Figure 1The present invention provides a method for controlling a sweeping brush of a sweeper based on image recognition, which includes the following steps:

[0045] S100, acquiring a road image and extracting bounding boxes of various types of garbage in the road image using an object detection model;

[0046] Specifically, a trained object detection algorithm is used to extract features of trash in road images. These features are then classified to determine the type of trash (such as plastic bags, paper scraps, cigarette butts, and leaves) and locate the trash. Bounding boxes are then output for the trash in the road image. Object detection algorithms such as YOLO (You Only Look Once) and Faster R-CNN can be used.

[0047] S200, determining the probability distribution of garbage types for pixels within each bounding box, and calculating the intersection degree of each type of garbage within the bounding box based on the probability distribution of garbage types and the pixel positions of each type of garbage;

[0048] Specifically, after obtaining bounding boxes for all garbage in the road image, any intersecting bounding boxes indicate garbage accumulation and require special treatment, marking them as high-priority cleaning areas. Subsequently, the sweeper dynamically adjusts the rotation speed and downward pressure of the sweeping brush based on the type of garbage and the degree of accumulation to ensure thorough removal. In this way, the sweeper can intelligently identify and efficiently handle various types of road debris, improving cleaning results. Bounding boxes that are distributed farther apart are considered to represent scattered garbage distribution areas, and the cleaning intensity is appropriately reduced to avoid wasting resources.

[0049] S300 , establishing a pollution weight for the bounding box based on the garbage type of each pixel within the bounding box, and determining the garbage concentration of each bounding box based on the intersection degree and pollution weight of each type of garbage within each bounding box;

[0050] Specifically, a dynamically updated garbage concentration field is established through real-time collected road surface images, which comprehensively considers the physical characteristics, garbage concentration and mutual superposition relationship of different garbage types to provide accurate data support for cleaning strategies.

[0051] Pollution weights are dynamically adjusted based on the type, volume, and difficulty of cleaning. For example, difficult-to-clean trash like cigarette butts and candy wrappers are assigned a higher weight, while easier-to-clean trash like fallen leaves is assigned a lower weight. By continuously optimizing the weight allocation strategy, cleaning decisions become more accurate and efficient.

[0052] During the subsequent cleaning process, the cleaning brush speed and downward pressure will be automatically increased for high-concentration areas; for low-concentration areas, energy-saving mode will be adopted to achieve optimal resource allocation.

[0053] S400 , performing cluster analysis on each bounding box based on the garbage concentration and location distribution of each bounding box to obtain a cleaning area;

[0054] Specifically, the spatial position data of each bounding box in the road image is input into the selected clustering algorithm. The clustering algorithm clusters the bounding boxes according to the set parameters, grouping bounding boxes with similar spatial positions into the same cluster. Each cluster corresponds to a potential cleaning area.

[0055] S500: Determine the cleaning difficulty of the cleaning area based on the garbage concentration of each boundary box in the cleaning area, and control the angle and rotation speed of the cleaning brush in each cleaning area according to the cleaning difficulty.

[0056] In the embodiment provided by the present invention, by real-time monitoring of road surface images and dynamic adjustment of cleaning strategies, it is ensured that garbage in each area can be effectively removed, thereby greatly improving cleaning efficiency and road cleanliness.

[0057] As an improvement to the above embodiment, in S200, determining the probability distribution of garbage types for pixels within each bounding box and calculating the intersection degree of garbage within the bounding box based on the probability distribution of garbage types and the pixel positions of each type of garbage include:

[0058] S210 , summing the ratios of the frequencies of each pixel in the bounding box belonging to the target type of garbage to the total number of pixels to obtain the probability of the target type of garbage in the bounding box;

[0059] S220, calculating the pixel position difference of the target type of garbage in the two bounding boxes to obtain the pixel difference of the target type of garbage in the two bounding boxes;

[0060] S230 , multiplying the probability of each target type of garbage in the two bounding boxes by the pixel difference and summing the results to obtain the intersection degree of the target type of garbage in the two bounding boxes, and summing the intersection degrees of each target type of garbage in the bounding boxes to obtain the intersection degree of each type of garbage in the bounding boxes.

[0061] Specifically, the intersection degree of garbage types within the bounding box is accurately calculated by combining the location information. The intersection degree calculation formula is: , where P(i,j) is the intersection between the i-th bounding box and the j-th bounding box, p(k) is the probability of the k-th type of garbage, and d(k,i,j) is the pixel difference between the k-th type of garbage in the i-th and j-th bounding boxes.

[0062] d(k,i,j) is obtained by calculating the pixel position difference of each type of garbage in the two bounding boxes, comprehensively evaluating the garbage distribution in the overlapping area to ensure the accuracy of the intersection calculation.

[0063] , where f(m,k) is the frequency at which the mth pixel belongs to the kth type of garbage, and T is the total number of pixels. Using this formula, the system can accurately assess the degree of overlap between garbage types within each bounding box, thereby optimizing region division and cleaning strategies.

[0064] By classifying and identifying each pixel and counting its frequency of occurrence in each type of garbage, the accuracy of f(m,k) is ensured. Combined with the total number of pixels T, p(k) is precisely calculated. The intersection degree is then evaluated using the P(i,j) formula, optimizing garbage area division and improving cleaning efficiency. The ratio of f(m,k) to T reflects the distribution of the kth type of garbage in the overall image and directly affects the accuracy of the intersection degree calculation. Accurate f(m,k) and T values ​​ensure the reliability of p(k), making bounding box merging and area division more rational, ultimately achieving efficient and accurate cleaning path planning.

[0065] As an improvement to the above embodiment, in S300, establishing a pollution weight for a bounding box based on the garbage type of each pixel within the bounding box, and determining the garbage concentration of each bounding box based on the intersection degree of each type of garbage within each bounding box and the pollution weight, includes:

[0066] S310 , setting a basic weight coefficient based on the type of garbage, multiplying the garbage concentration of each type of garbage within the bounding box by the corresponding basic weight coefficient to obtain the pollution weight of each type of garbage;

[0067] Specifically, a basic weight coefficient is set based on the type of garbage. Real-time image analysis is used to calculate the concentration of each type of garbage within the bounding box. This concentration is multiplied by the basic weight coefficient to generate a dynamic pollution weight. This pollution weight accurately reflects the impact of different types of garbage on the difficulty of cleaning, providing a reliable basis for subsequent garbage concentration calculations.

[0068] S320 , performing a weighted summation of the pollution weights of each type of garbage within the bounding box and their intersections to obtain a comprehensive pollution index for the bounding box;

[0069] Specifically, by introducing an intersection correction factor, the pollution index of overlapping areas is nonlinearly amplified to ensure that the pollution level in areas with high garbage concentrations is fully reflected. At the same time, a linear weighting method is used for isolated garbage areas to maintain the rationality of the pollution index.

[0070] S330 , calculating a garbage concentration value per unit area based on the comprehensive pollution index of the bounding box and the area of ​​the bounding box as the garbage concentration of the bounding box.

[0071] Specifically, the comprehensive pollution index is divided by the area of ​​the bounding box to perform area normalization. This eliminates the influence of bounding box size on concentration assessment and ensures that garbage concentrations in areas of different sizes are comparable. The resulting garbage concentration field accurately reflects the pollution level of each area of ​​the road, providing a quantitative basis for intelligent cleaning decisions.

[0072] As an improvement to the above embodiment, in S400, cluster analysis is performed on each bounding box based on the garbage concentration and location distribution of each bounding box to obtain a cleaning area, including:

[0073] S410, setting a cleaning radius based on the coverage area of ​​the cleaning brush and the average value of the garbage concentration in each bounding box, setting a minimum number of points based on the minimum amount of garbage in the road surface image to be considered as a covered area, and determining a cluster center based on the garbage concentration in each bounding box;

[0074] Specifically, set an appropriate sweeping radius and minimum number of points. These two parameters need to be adjusted based on the actual distribution of the bounding box and the desired coverage area granularity. If the garbage is densely distributed, the sweeping radius can be appropriately reduced, and the minimum number of points is set based on the minimum amount of garbage in the road image that can be considered a covered area.

[0075] S420, using the cleaning radius as the cluster boundary of the cluster center, clustering the adjacent bounding boxes of the cluster center within the minimum number of points to form a cleaning area;

[0076] Specifically, each bounding box is arranged in descending order according to the garbage concentration, and the first N high-concentration bounding boxes are selected as cluster centers. It can be understood that the cluster boundary divided in the area with the cluster center as the center and the cleaning radius as the radius contains multiple bounding boxes.

[0077] S430: Divide the cleaning area into multiple sub-areas, and determine the local garbage concentration of each sub-area based on the area ratio and garbage concentration of each bounding box in the sub-area;

[0078] Specifically, the area proportion of each bounding box in the sub-area is determined, and the weighted average of each bounding box in the sub-area according to the area proportion and garbage concentration is calculated to obtain the local garbage concentration of the sub-area, ensuring the reasonable setting of the subsequent cluster center.

[0079] S440: Determine the distance between each sub-region and the cluster center, select the sub-region with the largest product of local garbage concentration and distance as the new cluster center, and determine whether the change in distance between the cluster centers is below a distance threshold. If so, execute S450; if not, execute S420.

[0080] Specifically, the system calculates the product of the garbage concentration in each sub-area and the distance to the cluster center. The sub-area with the largest product is selected as the new cluster center to ensure maximum cleaning efficiency. Through continuous iterative optimization, the optimal cleaning area in the road image is ultimately determined, achieving efficient and intelligent cleaning.

[0081] S450, performing gradient calculation on the garbage concentration of each boundary box in the cleaning area, taking multiple boundary boxes with larger gradients as reference boxes, generating a circumscribed curve containing multiple reference boxes, taking the cluster center as the center of the circle, and generating a cleaning area containing the circumscribed curve within the cleaning radius and the minimum number of points.

[0082] By calculating the gradient of garbage concentration within the cleaning area, the greater the gradient change, the clearer the boundary, making it easier to focus on cleaning. Based on the cluster center, cleaning radius, and minimum number of points, the cleaning area boundaries are precisely delineated to avoid missed or duplicate cleaning. By iteratively optimizing clustering parameters, a comprehensive cleaning area division scheme with clear boundaries is ultimately established, laying the foundation for subsequent route planning.

[0083] As an improvement to the above embodiment, in S420, the method of using the cleaning radius as the cluster boundary of the cluster center and clustering the adjacent bounding boxes of the cluster center within the minimum number of points to form a cleaning area includes:

[0084] S421, filtering neighboring bounding boxes based on the cluster center and the cleaning radius to ensure that the number of bounding boxes in each cluster is not less than the minimum number of points to form a cleaning unit;

[0085] Specifically, by checking the bounding boxes around the cluster center one by one, the unqualified bounding boxes that are too far away or have too low garbage concentration are eliminated to ensure that the number of bounding boxes in each cleaning unit meets the minimum point requirement, thereby accurately dividing the cleaning area.

[0086] S422: If there are overlapping bounding boxes on the edge of the cleaning unit, the bounding box is divided into a cleaning unit with a larger pixel ratio, the position of the cleaning unit is adjusted based on the bounding box in the cleaning unit, and the adjusted cleaning unit is used as the cleaning area.

[0087] The formed cleaning units are optimized and adjusted to ensure that the boundaries of each unit are clear and non-overlapping. By calculating the overlapping area of ​​the boundary boxes of each cleaning unit, the boundary position is adjusted to ensure that each cleaning unit is independent and has complete coverage. Finally, an accurate cleaning area map is generated to guide the efficient operation of the intelligent sweeper.

[0088] As an improvement to the above embodiment, in S500, determining the garbage concentration in the cleaning area and controlling the angle and speed of the cleaning brush in each cleaning area according to the garbage concentration include:

[0089] S510 , accumulating the garbage concentrations of each bounding box in the cleaning area and performing exponential normalization processing to obtain the cleaning difficulty of the cleaning area;

[0090] S520, obtaining a range of bristle angles, establishing a first functional relationship between cleaning difficulty and bristle angle within the range; and adjusting the bristle angle within the range based on the cleaning difficulty.

[0091] Specifically, the angle of the cleaning brush bristles is adjusted based on the difficulty of cleaning the garbage. For example, if there are large pieces of garbage or a large amount of garbage that is difficult to clean in the cleaning area, the bristle angle is increased to enable the bristles to better pick up the garbage and sweep it into the trash can. If there are small particles or other garbage that are easier to clean in the cleaning area, the bristle angle is decreased to increase the contact area between the bristles and the ground, thereby improving the cleaning effect.

[0092] The relationship between the bristle angle θ and the cleaning difficulty S is as follows: ;

[0093] in, and are the maximum and minimum values ​​of the bristle angle, is the base step length of the bristle angle, is the threshold of cleaning difficulty, is the cleaning difficulty, k is the adjustment coefficient, 0<k<1.

[0094] S530, obtaining the rotation speed of the cleaning brush, the driving speed of the cleaning vehicle, the reference speed of the cleaning vehicle and the basic rotation speed of the cleaning brush, and determining the rotation speed of the cleaning brush in combination with the difficulty of cleaning the garbage.

[0095] Specifically, the speed of the sweeping brush is adjusted in real time based on the difficulty of garbage removal and the speed of the sweeper. In areas with dense garbage or when the sweeper is traveling slowly, the speed is increased to enhance cleaning power; in areas with sparse garbage or when the sweeper is traveling quickly, the speed is reduced to conserve energy.

[0096] The relationship between the sweeping brush speed n, the cleaning difficulty S and the sweeper speed v is as follows:

[0097] n=n0×(1+a×S b×v / v0)

[0098] Among them, n is the speed of the sweeping brush, S is the difficulty of garbage cleaning, v is the driving speed of the sweeper, v0 is the baseline speed of the sweeper, n0 is the basic speed of the sweeping brush, a and b are the weight coefficients of cleaning difficulty and sweeper speed respectively.

[0099] During the sweeper's operation, the control strategy for the sweeping brushes is dynamically adjusted based on the real-time changes in the difficulty of garbage removal. When the sweeper is about to reach an area with greater cleaning difficulty, the speed, angle, and position of the sweeping brushes are adjusted in advance to ensure that the brushes can remove garbage in a timely and effective manner. When the sweeper leaves the garbage area, the operating state of the sweeping brushes is adjusted accordingly to reduce energy consumption.

[0100] refer to Figure 2 The embodiment of the present invention further provides a sweeping brush control system for a sweeping vehicle based on image recognition, comprising:

[0101] at least one processor;

[0102] at least one memory for storing at least one program;

[0103] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0104] The contents of the above method embodiments are all applicable to this embodiment. The functions specifically implemented by this embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments, which will not be repeated here.

[0105] In addition, an embodiment of the present application further discloses a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to execute the above method.

[0106] In addition, the embodiments of the present application further disclose a computer program product or computer program, which is stored in a computer-readable storage medium. The processor of a computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device performs the above-mentioned method. Similarly, the contents of the above-mentioned method embodiment are all applicable to the present storage medium embodiment, and the functions specifically implemented by the present storage medium embodiment are the same as those of the above-mentioned method embodiment, and the beneficial effects achieved are also the same as those achieved by the above-mentioned method embodiment.

[0107] Those skilled in the art will appreciate that all or some of the systems in the methods disclosed above can be implemented as software, firmware, hardware, or any suitable combination thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on computer-readable media, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is well known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVDs) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. Furthermore, as is well known to those skilled in the art, communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0108] Although the description of the present disclosure has been quite detailed and particularly describes several embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be considered to provide a broad possible interpretation of these claims by reference to the appended claims in view of the prior art, thereby effectively covering the intended scope of the present disclosure. In addition, the above description of the present disclosure is based on the embodiments foreseen by the inventors, which is intended to provide a useful description, and those non-substantial changes to the present disclosure that have not yet been foreseen may still represent equivalent changes to the present disclosure.

Claims

1. A method for controlling a sweeping brush of a sweeper based on image recognition, characterized in that: The method comprises the following steps: Obtain road images and use the object detection model to extract the bounding boxes of various types of garbage in the road images; Determine the probability distribution of garbage types for pixels within each bounding box, and calculate the intersection degree of each garbage type within the bounding box based on the garbage type probability distribution and the pixel position of each garbage type; The pollution weight of the bounding box is established based on the garbage type of each pixel in the bounding box, and the garbage concentration of each bounding box is determined based on the intersection degree and pollution weight of each type of garbage in each bounding box; Perform cluster analysis on each bounding box based on its garbage concentration and location distribution to obtain the cleaning area; The cleaning difficulty of the cleaning area is determined based on the garbage concentration of each bounding box in the cleaning area, and the angle and speed of the cleaning brush in each cleaning area are controlled according to the cleaning difficulty.

2. The method according to claim 1, characterized in that Determining the probability distribution of garbage types for pixels within each bounding box and calculating the intersection degree of garbage within the bounding box based on the probability distribution of garbage types and the pixel positions of each type of garbage includes: The probability of the target type of garbage in the bounding box is obtained by summing the ratio of the frequency of each pixel belonging to the target type of garbage to the total number of pixels in the bounding box; Calculate the pixel position difference of the target type garbage in the two bounding boxes to obtain the pixel difference of the target type garbage in the two bounding boxes; The probability of each target type of garbage in the two bounding boxes is multiplied by the pixel difference and then added together to obtain the intersection degree of the target type of garbage in the two bounding boxes. The intersection degree of each target type of garbage in the bounding box is accumulated to obtain the intersection degree of each type of garbage in the bounding box.

3. The method according to claim 1, characterized in that The method of establishing a pollution weight of a bounding box based on the garbage type of each pixel in the bounding box and determining the garbage concentration of each bounding box based on the intersection degree and pollution weight of each type of garbage in each bounding box includes: Set a basic weight coefficient based on the type of garbage, and multiply the garbage concentration of each type of garbage within the boundary box by the corresponding basic weight coefficient to obtain the pollution weight of each type of garbage; The pollution weights of each type of garbage in the bounding box and their intersection are weighted and summed to obtain the comprehensive pollution index of the bounding box; Based on the comprehensive pollution index of the bounding box and the area of ​​the bounding box, the garbage concentration value per unit area is calculated as the garbage concentration of the bounding box.

4. The method according to claim 1, wherein The cluster analysis of each bounding box based on the garbage concentration and location distribution of each bounding box is performed to obtain the cleaning area, including: The sweeping radius is set based on the average value of the brush coverage area and the garbage concentration of each bounding box. The minimum number of points in the road image is set based on the minimum amount of garbage to be considered as a covered area. The cluster center is determined based on the garbage concentration of each bounding box. The cleaning radius is used as the cluster boundary of the cluster center, and the adjacent bounding box of the cluster center is clustered within the minimum number of points to form a cleaning area; The cleaning area is divided into multiple sub-areas, and the local garbage concentration of each sub-area is determined based on the area ratio and garbage concentration of each bounding box in the sub-area; Determine the distance between each sub-region and the cluster center, select the sub-region with the largest product of local garbage concentration and distance as the new cluster center, and determine whether the change in distance to the cluster center is below a distance threshold. If so, execute S450; if not, execute S420; The garbage concentration of each bounding box in the cleaning area is calculated by gradient, and multiple bounding boxes with larger gradients are used as reference boxes to generate a circumscribed curve containing multiple reference boxes. With the cluster center as the center of the circle, a cleaning area containing the circumscribed curve is generated within the cleaning radius and the minimum number of points.

5. The method according to claim 4, characterized in that The sweeping radius is used as the cluster boundary of the cluster center, and the adjacent bounding box of the cluster center is clustered within the minimum number of points to form a sweeping area, including: Filter neighboring bounding boxes based on the cluster center and cleaning radius to ensure that the number of bounding boxes in each cluster is not less than the minimum number of points to form a cleaning unit; If there are overlapping bounding boxes on the edge of the cleaning unit, the bounding box is divided into the cleaning unit with a larger pixel ratio, the position of the cleaning unit is adjusted based on the bounding box in the cleaning unit, and the adjusted cleaning unit is used as the cleaning area.

6. The method according to claim 1, characterized in that The method of determining the garbage concentration in the cleaning area and controlling the angle and speed of the cleaning brush in each cleaning area according to the garbage concentration includes: The garbage concentration of each bounding box in the cleaning area is accumulated and then normalized exponentially to obtain the cleaning difficulty of the cleaning area; Acquiring a range value of a bristle angle, establishing a first functional relationship between cleaning difficulty and bristle angle within the range value; and adjusting the bristle angle within the range value based on the cleaning difficulty; Obtain the speed of the sweeping brush, the driving speed of the sweeping vehicle, the baseline speed of the sweeping vehicle, and the basic speed of the sweeping brush, and determine the speed of the sweeping brush based on the difficulty of garbage cleaning.

7. A sweeper brush control system based on image recognition, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 6.

8. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is configured to perform the method according to any one of claims 1 to 6 when executed by the processor.

Images