Pet excrement treatment method

Abstract

The invention particularly relates to a pet excrement treatment method, which comprises the following steps of S1, setting and optimizing a flight path of an unmanned aerial vehicle by adopting an improved A * algorithm in combination with a community terrain and a building layout; s2, the system sets the unmanned aerial vehicle to take off regularly every day and automatically executes a cleaning task; s3, uploading an image shot by the high-definition camera to a central processing system in real time through a wireless transmission technology by using a 5G network technology; s4, the central processing system adopts an advanced image processing algorithm to analyze the image in real time and recognize the position of the pet excrement; s5, the unmanned aerial vehicle hovers above the excrement, and collection and treatment of the excrement are completed through a pickup device; and S6, when the unmanned aerial vehicle completes the cleaning task or does not detect excrement, the unmanned aerial vehicle automatically returns to a preset charging station for charging.

Description

Technical Field

[0001] The invention belongs to the technical field of image transmission and recognition, and in particular relates to a method for processing pet excrement. Background Art

[0002] In modern cities and residential areas, pet feces management has always been a prominent public health and environmental issue. Traditional pet feces cleaning methods mainly rely on manual collection, which is not only time-consuming and labor-intensive, but also poses health risks. In some cases, pet feces are not handled in a timely manner and may be washed into water bodies by rainwater, causing environmental pollution and affecting the balance of the ecosystem.

[0003] With the advancement of science and technology, the application of intelligent technology and automated equipment in various fields continues to expand. Drone technology has shown great potential in agriculture, logistics, monitoring and other fields with its efficient monitoring capabilities and flexible execution capabilities. Applying drone technology to pet feces management can not only improve cleaning efficiency, but also effectively avoid hygiene problems that occur during manual cleaning.

[0004] The present invention proposes an automatic detection and processing system for pet feces using a drone based on cloud-edge collaboration and an improved A* algorithm. The system uses a drone equipped with a high-definition camera and a robotic arm feces processing device, and realizes efficient detection and cleaning of pet feces through a path planning module and an image processing module. The drone can patrol the community according to a preset path, and transmit the images taken by the camera to the central processing system for analysis and processing through a real-time image transmission module. Summary of the invention

[0005] The purpose of the present invention is to provide a pet excrement treatment method to solve the problems existing in the background technology.

[0006] In order to achieve the above technical objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for treating pet feces comprises the following steps:

[0008] S1: Using the improved A* algorithm, combined with the community terrain and building layout, the flight path of the drone is set and optimized;

[0009] S2: The system sets the drone to take off at a fixed time every day and automatically perform cleaning tasks;

[0010] S3: Using 5G network technology, images captured by high-definition cameras are uploaded to the central processing system in real time through wireless transmission technology;

[0011] S4: The central processing system uses advanced image processing algorithms to analyze images in real time and identify the location of pet feces;

[0012] S5: The drone hovers above the feces and uses a picking device to complete the collection and processing of the feces;

[0013] S6: When the drone completes its cleaning task or detects no feces, it automatically returns to the preset charging station for charging.

[0014] Preferably, in step S1, the path planning system adopts a cloud-edge collaborative architecture, and the system consists of a cloud server, an edge server, a roadside unit RSU and a drone, wherein:

[0015] Cloud server: responsible for global path planning, storing the environmental map of the entire community, and assigning fine-grained path planning tasks to corresponding edge servers

[0016] Edge server: stores local environment maps of the cell, caches the optimal path in the coverage area, and provides local path planning services to reduce the response delay of path planning requests;

[0017] Roadside Unit (RSU): deployed on the ground, connecting all drones in its coverage area and communicating with the edge server;

[0018] Drone: As an end user, it sends a path planning request to the edge server and moves along the planned path.

[0019] Preferably, when the UAV sends a path request, the edge server first checks whether there is an available path in the cache; if not, the edge server forwards the request to the cloud server, and the cloud server performs coarse-grained path planning, calculates the entry and exit points of each edge area through which the path passes, and distributes this information to the corresponding edge server. The edge server then performs fine-grained path planning and uses the improved A* algorithm and grid cache structure to quickly calculate the optimal path.

[0020] Preferably, in step S3, 5G network technology is used for image transmission. The video captured by the high-definition camera is transmitted to the central processing system in real time through the wireless module on the drone; the image data is compressed using the H.265 encoding format to ensure data transmission efficiency under high bandwidth conditions. The transmission system also includes data caching and packet loss retransmission mechanisms to ensure the continuity and clarity of the image.

[0021] Preferably, in step S4, the central processing system uses the YOLOv4 deep learning algorithm for image processing and feces recognition:

[0022] Preprocessing: extract frames from the received video stream, extract 30 frames per second, convert them into grayscale images, and perform Gaussian filtering to remove noise;

[0023] Object detection: Use the pre-trained YOLOv4 model for object detection. The model is based on the CSPDarknet53 backbone network and has efficient feature extraction capabilities, which can quickly identify pet feces in images.

[0024] Advantages of the invention: The present invention discloses an automatic detection and processing system for pet feces by drones based on cloud-edge collaboration and improved A* algorithm. By combining drone path planning, real-time image transmission, image recognition processing and automated cleaning technology, it can efficiently and accurately detect and process pet feces, significantly improve the sanitation management efficiency of cities and residential areas, reduce labor costs, and improve environmental quality. It has broad application prospects in the fields of urban management and property services. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention is further illustrated by means of the non-limiting examples given in the accompanying drawings.

[0026] Figure 1 A flow chart of a pet excrement treatment method of the present invention; DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0028] A commercial quad-rotor drone, model DJI Matrice 200, was used in this embodiment. The drone is equipped with a Sony IMX477 high-definition camera, which has high resolution (12.3MP) and night vision function, and can provide clear images under various lighting conditions; a manipulator feces treatment device is installed on the drone, and the manipulator uses a Robotiq 2F-85 gripper, equipped with a suction nozzle with a vacuum function and a disinfectant sprayer. The suction nozzle has a built-in powerful suction motor that can efficiently absorb feces, and the disinfectant sprayer uses Clorox disinfectant to disinfect the cleaned area;

[0029] like Figure 1 As shown, a pet excrement treatment method of the present invention comprises the following steps:

[0030] S1: Using the improved A* algorithm, combined with the community terrain and building layout, the flight path of the drone is set and optimized;

[0031] Path planning is a core component for drones to perform automated tasks. Traditional path planning algorithms usually rely on centralized cloud servers to process path calculation requests, which leads to problems such as transmission delay, calculation delay, and queue delay, especially in large-scale environments. These delays not only reduce the response speed of path planning, but also increase the complexity and computational cost of the system. In order to solve these problems, this paper introduces a grid cache path planning algorithm based on cloud-edge collaboration to improve the efficiency and response speed of path planning.

[0032] This path planning system adopts a cloud-edge collaborative architecture. The system consists of a cloud server, an edge server, a roadside unit (RSU), and a drone.

[0033] Cloud server: responsible for global path planning, storing the environmental map of the entire community, and assigning fine-grained path planning tasks to corresponding edge servers

[0034] Edge server: stores local environment maps of the cell, caches the optimal path in the coverage area, and provides local path planning services to reduce the response delay of path planning requests;

[0035] Roadside Unit (RSU): deployed on the ground, connecting all drones in its coverage area and communicating with the edge server;

[0036] Drone: As an end user, it sends a path planning request to the edge server and moves along the planned path;

[0037] Specifically, the path cache includes dynamic cache and static cache; traditional path cache methods such as least recently used (LRU) and highest query frequency (HQF) do not consider path planning, query time and the space occupied by the path; in order to solve this problem, the present invention adopts a grid-based path cache structure and introduces a dynamic update method to improve the hit rate and query efficiency of the path cache;

[0038] Assume that the entire environment is divided into a grid G, where each grid cell g i,j Represents a location. The path cache can be represented as a set C, which contains all cached paths. The path cache hit rate H is defined as:

[0039]

[0040] By optimizing the path cache structure, the value of H can be increased, thereby improving the efficiency of path planning;

[0041] When the drone sends a path request, the edge server first checks whether there is an available path in the cache; if not, the edge server forwards the request to the cloud server, which performs coarse-grained path planning, calculates the entry and exit points of each edge area through which the path passes, and distributes this information to the corresponding edge server. The edge server then performs fine-grained path planning and uses the improved A* algorithm and grid cache structure to quickly calculate the optimal path.

[0042] The goal of coarse-grained path planning is to find a path P from the starting point s to the end point t. coarse , which can be expressed as:

[0043] P coarse ={s,e1,e2,...,t}

[0044] where e i Indicates the entry and exit points of each edge region;

[0045] In order to speed up the path planning process, the present invention improves the traditional A* algorithm to make it compatible with the path cache structure; the improved A* algorithm reduces the number of path points, shortens the total path length, and improves the smoothness of the path, thereby enhancing the practicality of the path cache;

[0046] The heuristic function h(n) of the improved A* algorithm is defined as the estimated cost from node n to the target node; h(n) of the traditional A* algorithm usually uses Euclidean distance or Manhattan distance. The improved heuristic function is used in the present invention, combined with the path cache information:

[0047] h(n)=α·d cache (n,t)+β·d real (n,t)

[0048] Among them, d cache (n, t) represents the distance from node n to target node t in the cache path, d real (n, t) represents the distance from node n to target node t in the actual environment. α and β are weight parameters used to balance the impact of cache path and actual path.

[0049] In the path planning process, through the collaborative work of cloud servers and edge servers, cloud-edge collaboration reduces the cloud load of high-resolution path planning tasks and improves the response speed of the system; based on the grid-based path cache structure, the grid cache supports efficient path search and sharing, reducing the amount of invalid calculations; a dynamic update mechanism is introduced to continuously optimize the path cache and improve the cache hit rate; the improved A* algorithm reduces the computational complexity and the number of path points in path planning, improving the efficiency of path planning and the utilization of the cache;

[0050] S2: The system sets the drone to take off at a fixed time every day and automatically perform cleaning tasks; the specific operations are as follows:

[0051] Mission plan: The central control system sets the take-off time and patrol route of the drone, with the drone flying once at 7 a.m. and 5 p.m. every day;

[0052] Automatic take-off: According to the preset time, the drone automatically takes off to perform the task, patrols the community through the path planning module, and transmits images in real time for feces detection and processing;

[0053] S3: Using 5G network technology, images captured by high-definition cameras are uploaded to the central processing system in real time through wireless transmission technology;

[0054] Use 5G network technology for image transmission. The video captured by the high-definition camera is transmitted to the central processing system in real time through the wireless module on the drone; the image data is compressed using the H.265 encoding format to ensure data transmission efficiency under high bandwidth conditions. The transmission system also includes data caching and packet loss retransmission mechanisms to ensure image continuity and clarity;

[0055] S4: The central processing system uses advanced image processing algorithms to analyze images in real time and identify the location of pet feces;

[0056] The central processing system uses the YOLOv4 (You Only Look Once) deep learning algorithm for image processing and feces recognition;

[0057] Preprocessing: extract frames from the received video stream, extract 30 frames per second, and convert them into grayscale images;

[0058] Object detection: Use the pre-trained YOLOv4 model for object detection. The model is based on the CSPDarknet53 backbone network and has efficient feature extraction capabilities, which can quickly identify pet feces in images.

[0059] In the process of feces target detection, the size of feces may be very small and may appear insignificant relative to the entire image. In this case, traditional detection methods may not be able to accurately detect these tiny defects. We optimize the above problems from the following perspectives:

[0060] ①. Improve image resolution: In the present invention, we first optimize this problem from the perspective of resolution. High-resolution images can make defects more obvious, thereby improving the accuracy of detection. We use Sony IMX477 high-definition camera to shoot in the community to capture more details;

[0061] ②. Image preprocessing: In this invention, we use Gaussian filtering to reduce the background noise of the image to highlight the defects;

[0062] ③. Multi-scale detection is an important technology in computer vision, which aims to improve the model's ability to detect targets of different sizes. In the present invention, we use the Feature Pyramid Network (FPN) in the multi-scale detection method for detection. Specifically, first, the Feature Pyramid Network creates a feature pyramid by generating feature maps in different convolutional layers. Each feature map corresponds to a different scale, so that targets of different sizes can be captured. Secondly, the feature pyramid can fuse features of different scales through top-down and bottom-up paths to generate multi-scale feature maps. In this way, the network can perform target detection at each scale and combine feature information of different sizes. In summary, the feature pyramid method significantly improves the detection accuracy, and because only operations need to be performed on feature maps, the computational overhead can be reduced.

[0063] Position calibration: Position calibration uses a machine vision and “partner” calibration method to ensure that the drone can accurately identify and handle pet feces; the specific process is as follows:

[0064] 1. Introducing the coordinate system conversion relationship, we set the position of feces in the pixel coordinate system, and finally determine its precise position in the world coordinate system through a series of coordinate system conversions; the specific conversion relationship is as follows:

[0065] Conversion from pixel coordinate system to camera coordinate system,

[0066]

[0067] Among them, K is the camera intrinsic parameter matrix;

[0068] Conversion from camera coordinate system to drone coordinate system,

[0069]

[0070] in, is the rotation matrix from camera to drone, is the translation vector from the camera to the drone;

[0071] Conversion from drone coordinate system to world coordinate system,

[0072]

[0073] in, is the rotation matrix from the drone to the world coordinate system, is the translation vector from the drone to the world coordinate system;

[0074] 2. Based on the partner calibration model of 6D pose estimation, a partner calibration method is introduced, and the gimbal camera of the follower UAV is used to track the leader UAV to achieve temporal and spatial calibration of sensors in the multi-UAV system; the specific steps are as follows:

[0075] 6D pose estimation: Obtain the 6D pose estimation of the target drone, including the rotation matrix and translation vector, by following the drone’s gimbal camera;

[0076] Time offset and rotation parameter calibration: Due to the delay in camera image data, we need to calibrate the time offset Δt and rotation parameters The calibration model is as follows:

[0077]

[0078] Among them, Δt L is the communication delay between the pilot drone and the follower drone;

[0079] 3. Multi-step solution method,In order to deal with measurement errors and initial value sensitivity, we adopted an error-compatible multi-step solution strategy, including:

[0080] Outlier screening: outliers are screened out through the random sampling consensus algorithm (RANSAC) to improve the robustness of the model;

[0081] Time offset and rotation parameter calculation: Based on the extracted 6D pose estimation data, the time offset Δt and rotation parameters are calculated respectively.

[0082] Multi-parameter joint optimization: Finally, the time offset and rotation parameters are further calibrated through a joint optimization strategy to improve the calibration accuracy;

[0083] S5: The drone hovers above the feces and uses a picking device to complete the collection and processing of the feces;

[0084] In order to achieve accurate hovering of the drone above the feces, this system adopts a hovering control strategy based on supervised reinforcement learning (WAC algorithm); the hovering height is set to 10 cm to ensure the stability and safety of the drone when performing feces treatment tasks; the specific steps are as follows:

[0085] Assume that the UAV system is a continuous dynamic system, and its state characteristic is s t ∈R s , the system input is a∈R a , then:

[0086] s t+1 =g(st ,a)+N t

[0087] Among them, s t +1 is based on the current status t and the next state prediction of action a, g() is the unknown dynamic transition function of the system state, N t is a random noise function, representing a small range of exploration near the control variable;

[0088] Assume that each time process of the drone has a Markov property and establish a Markov decision process quintuple<S,A,P,R,γ> , the state s and action a are defined as follows:

[0089]

[0090] a=[f p1 ,f p2 ,f p3 ,f p4 ]

[0091] Among them, the state includes information such as position, velocity, acceleration, angular velocity and attitude angle, P bx , P by , P bz is the position coordinate of the UAV in space, and h is the hovering height;

[0092] The WAC control strategy is adopted, combining the traditional PID controller and the neural network. In the specific implementation process, a three-layer neural network with parameters is used as the dynamic monitor of the PID controller, strong supervision is performed at the beginning of each learning stage, and the supervision intensity is gradually reduced, and finally unsupervised online learning is achieved.

[0093]

[0094] Among them, R(s t ,a t ) is the reward function, which reflects the quality of the current state and action, and γ is the discount factor;

[0095] Using the improved DDPG algorithm, the UAV can be precisely controlled through dynamic monitoring of the supervisory network to ensure that it hovers above the target location.

[0096] When the drone hovers over the detected feces, the robotic arm will start and locate the feces. The suction device will suck the feces into the collection container through a vacuum pump and filtration system, and disinfect it;

[0097] The robotic arm adopts a six-degree-of-freedom joint design, combined with precise motion control algorithms, to ensure that it can be flexibly positioned and moved; a suction device is installed at the end of the robotic arm, including a vacuum pump and a filtration system;

[0098] The suction device is guided by the robotic arm to approach the feces, and then the vacuum pump is started to generate negative pressure, sucking the feces into the filtration system. The feces are initially separated and disinfected through the filtration system and enter the collection container.

[0099] Then start the disinfectant sprayer and use Clorox disinfectant to spray and disinfect the ground. The disinfectant sprayer evenly sprays the disinfectant through a high-pressure nozzle to ensure thorough disinfection of the cleaned area and eliminate potential health risks.

[0100] After the feces are collected, the drone transports the feces to the designated treatment area; in the treatment area, the drone transfers the feces in the storage tank to the centralized treatment facility through an automatic unloading device to ensure that the waste is properly treated and disposed of;

[0101] S6: When the drone completes the cleaning task or detects no feces, it automatically returns to the preset charging station for charging;

[0102] To ensure that the drone can continue to perform cleaning tasks efficiently, we designed an automated fixed-point recovery and charging system. The system allows the drone to automatically return to a preset charging station for charging after completing a series of tasks or when the battery is low; the charging station is equipped with the following functions and features:

[0103] Multiple charging points support: The charging station has multiple charging points, supporting multiple drones to charge at the same time; each charging point is equipped with an independent automatic docking system, which optimizes charging efficiency and equipment utilization;

[0104] Automatic docking: When the drone returns to the charging station, it automatically docks through the visual and infrared guidance system. The drone uses the built-in RTK GPS system for precise navigation to ensure that it can accurately land at the designated location of the charging station. When landing, fine adjustments are made using high-precision sensors, infrared recognition systems and electromagnets to ensure that the drone can accurately dock with the charging port.

[0105] Intelligent Scheduling: The system has built-in intelligent energy management software that can intelligently schedule the charging and task execution order of drones according to the power status and mission requirements of each drone; this ensures that each drone can get enough power when needed and maximizes the operating efficiency of the overall system;

[0106] Battery health management: During the charging process, the system will detect and maintain the battery health status, detect and deal with battery problems in a timely manner, and extend the battery life; health management includes detecting battery temperature, charge and discharge cycle times, voltage changes and other parameters to ensure that the battery operates in the best condition;

[0107] Efficient Charging: The charging system uses efficient wireless charging technology (Qi standard), which can fully charge the drone in a short time, shortening the drone's downtime and improving mission continuity.

[0108] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical ideas disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A method for treating pet excrement, characterized in that: The following steps are involved: S1: Using the improved A* algorithm, combined with the community terrain and building layout, the flight path of the drone is set and optimized; S2: The system sets the drone to take off at a fixed time every day and automatically perform cleaning tasks; S3: Using 5G network technology, images captured by high-definition cameras are uploaded to the central processing system in real time through wireless transmission technology; S4: The central processing system uses advanced image processing algorithms to analyze images in real time and identify the location of pet feces; S5: The drone hovers above the feces and uses a picking device to complete the collection and processing of the feces; S6: When the drone completes its cleaning task or detects no feces, it automatically returns to the preset charging station for charging.

2. A pet excrement treatment method according to claim 1, characterized in that: In step S1, the path planning system adopts a cloud-edge collaborative architecture. The system consists of a cloud server, an edge server, a roadside unit RSU and a drone, where: Cloud server: responsible for global path planning, storing the environmental map of the entire community, and assigning fine-grained path planning tasks to corresponding edge servers Edge server: stores local environment maps of the cell, caches the optimal path in the coverage area, and provides local path planning services to reduce the response delay of path planning requests; Roadside Unit (RSU): deployed on the ground, connecting all drones in its coverage area and communicating with the edge server; Drone: As an end user, it sends a path planning request to the edge server and moves along the planned path.

3. A pet excrement treatment method according to claim 2, characterized in that: When a drone sends a path request, the edge server first checks whether there is an available path in the cache; If not, the edge server forwards the request to the cloud server, which performs coarse-grained path planning, calculates the entry and exit points of each edge area through which the path passes, and distributes this information to the corresponding edge server. The edge server then performs fine-grained path planning and uses the improved A* algorithm and grid cache structure to quickly calculate the optimal path.

4. A pet excrement treatment method according to claim 1, characterized in that: In step S3, 5G network technology is used for image transmission. The video captured by the high-definition camera is transmitted to the central processing system in real time through the wireless module on the drone; the image data is compressed using the H.265 encoding format to ensure data transmission efficiency under high bandwidth conditions. The transmission system also includes data caching and packet loss retransmission mechanisms to ensure image continuity and clarity.

5. A pet excrement treatment method according to claim 1, characterized in that: In step S4, the central processing system uses the YOLOv4 deep learning algorithm for image processing and feces recognition: Preprocessing: extract frames from the received video stream, extract 30 frames per second, convert them into grayscale images, and perform Gaussian filtering to remove noise; Object detection: Use the pre-trained YOLOv4 model for object detection. The model is based on the CSPDarknet53 backbone network and has efficient feature extraction capabilities, which can quickly identify pet feces in images.

Images