Intelligent collaborative water surface garbage collection method

Abstract

The application discloses an intelligent collaborative water surface garbage collection method, and relates to the technical field of water surface environmental protection management.The method cooperates unmanned aerial vehicles and ship machines to collect data, forms a global atlas and a comprehensive data set through modeling calculation, extracts features based on the data and identifies garbage target objects through an algorithm, plans an optimal path according to the positioning of the target objects and captures the target objects, breaks, compresses and classifies the collected target objects, automatically returns and discharges when the loading reaches a threshold, rechecks the unmanned aerial vehicles during the period, and finally updates the operation experience through an incremental algorithm.The method cooperates unmanned aerial vehicles and ship machines, constructs a global visual atlas, accurately identifies water area target objects, solves the problem of complex environment detection, cooperates path planning and real-time correction to ensure efficient, safe operation, improves resource utilization by combining volume reduction and classification design, iteratively optimizes the system performance in an incremental mode, promotes efficient and refined water surface garbage management, and has a wide application prospect.

Description

Technical Field

[0001] This invention relates to the field of water surface environmental management technology, specifically to an intelligent collaborative method for collecting surface waste. Background Technology

[0002] With the increasing emphasis on global aquatic ecological environment protection, the management of surface waste has become a key link in ecological environment construction and water resource protection. Surface waste not only damages the integrity of aquatic landscapes, but also encroaches on the living space of aquatic organisms, pollutes the aquatic environment, and even blocks waterways and affects the normal operation of water conservancy facilities and shipping systems, posing a dual threat to ecological security and human production and life. At present, surface waste management technology is gradually transforming from the traditional manual-dominated model to mechanization and intelligence. The rapid development of technologies such as UAV remote sensing monitoring, shipborne multi-sensor detection, and intelligent algorithm optimization has provided the possibility of breaking through the bottlenecks of traditional management. However, how to deeply integrate multi-source technologies and build a collaborative system covering the entire process of detection-identification-collection-processing to achieve precise control and efficient management of waste in complex aquatic environments has become the core problem that the industry urgently needs to solve.

[0003] Traditional surface waste collection technologies have significant shortcomings and are ill-suited to the diverse needs of water management. On the one hand, traditional methods rely heavily on manual inspections and single-vehicle operations. These vessels, specifically intelligent surface waste collection boats, possess autonomous navigation, sensor data collection, and actuator operation capabilities. Manual inspections, limited by manpower and environmental conditions, cannot achieve real-time monitoring of large water areas, leading to incomplete understanding of waste distribution and high rates of missed detection. Single-vehicle operations lack a comprehensive field of view, relying solely on localized perception, resulting in delayed target discovery, unscientific path planning, and frequent ineffective navigation, significantly reducing collection efficiency. On the other hand, traditional technologies often rely on single-feature judgments for target identification, failing to accurately distinguish between waste types and aquatic plants, leading to miscollection or missed collection. Post-collection waste treatment lacks a systematic volume reduction and classification mechanism, not only consuming substantial storage space but also reducing resource recycling rates. Furthermore, it fails to accumulate operational experience, requiring repeated environmental adaptation for each operation, resulting in high management costs, unstable results, and difficulty in coping with complex and ever-changing water scenarios. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent collaborative method for collecting surface waste. By coordinating the operation of drones and boats, the method achieves efficient collection and treatment of surface waste. The method covers steps such as collaborative data acquisition, environmental adaptation, target identification, path planning, capture, volume reduction, classified storage, and return discharge. It utilizes multi-source data fusion, dynamic path planning, and feature matching algorithms to accurately identify and capture waste, while simultaneously crushing, compressing, and classifying it for storage. Finally, it automatically returns for discharge and updates operational experience through incremental learning to improve operational efficiency and environmental adaptability.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent collaborative method for collecting surface waste, the specific steps of which are as follows:

[0006] S100, Collaborative Data Acquisition: UAVs cruise to capture the full-domain distribution of targets, while ships collect detailed data on targets and aquatic environmental parameters. The data are then fused through a collaborative multi-source data fusion algorithm to form a full-domain distribution map and a comprehensive dataset.

[0007] The overall distribution map is based on an electronic map of the water area as its geographical framework, overlaid with target distribution information collected by drones to form a visual map. The map includes geographic coordinates of the targets, delineation of distribution boundaries, and regional distribution density level indicators. The distribution density level is divided into three levels—dense, medium, and sparse—based on the number of targets per unit area, each distinguished by different colors. The map also integrates water area geographic feature information, including the location boundaries of shorelines, channels, shoals, and reefs. Target distribution information and geographic feature information are presented through layer overlay. The map is stored in vector map format, supports segmented loading by cruise area, and is dynamically updated with new data transmitted in real time by the drone. The update frequency is consistent with the drone data transmission frequency and is synchronously pushed to the ship's path planning module for use.

[0008] S200, Environmental Adaptation and Target Recognition: Based on the aquatic environmental parameters in the comprehensive dataset, a multi-dimensional feature dynamic matching algorithm is used to extract the morphological, spectral, temperature, and material features of the target object, and compare them with the preset feature template library to complete the target object recognition;

[0009] S300, Collaborative Path Planning and Capture: Combining global distribution map, target identification results and real-time water flow data, the operation path is planned through a collaborative dynamic path planning algorithm; the UAV updates the target position coordinates in real time and sends route correction commands to the ship-to-aircraft aircraft; the ship-to-aircraft aircraft switches actuators according to the target type, captures the target, and transfers it to the collection compartment via a guide channel;

[0010] The aforementioned vessel-machinery refers to an intelligent surface garbage collection vessel, equipped with autonomous navigation, sensor data acquisition, and actuator operation functions. The actuator of the vessel-machinery includes a flexible fishing net and a rotary cutting-and-retrieval integrated component. The flexible fishing net is made of high-strength nylon material with a mesh size of 0.3-0.8cm and a telescopic arm extension stroke of 3-6m. The rotary cutting-and-retrieval integrated component contains 3-6 tungsten carbide cutting blades with a stepless speed adjustment between 200-400r / min and is equipped with a negative pressure adsorption device.

[0011] S400, Volume Reduction Processing and Classified Storage: After receiving the target material, the collection chamber starts the dual-shaft crushing and compression integrated device to crush and compress the target material; the volume-reduced target material is separated by a multi-layer screening mechanism, and different types of target materials are introduced into independent sealed chambers respectively;

[0012] The dual-shaft crushing and compressing integrated device includes two sets of relatively rotating crushing blades, each set containing 4-6 blades, rotating in opposite directions. The particle size of the target material after crushing is controlled at 2-5cm. The compression process adopts a staged pressurization mode, with an initial pressure of 3-5MPa, which is increased to 6-8MPa after holding for 3-5 seconds. The multi-layer screening mechanism has a first-layer screen with a mesh size of 1.5-2.5cm, a second-layer magnetic attraction device with adjustable magnetic strength, and a third-layer horizontal water flow impact density screening with a water flow velocity controlled at 0.5-1m / s.

[0013] The volume ratio of each independent sealed compartment is as follows: 40%-50% for aquatic plant storage, 30%-35% for recyclable material storage, and 15%-20% for hazardous waste storage. Each compartment is equipped with a pressure sensor, a liquid level sensor, and a rubber sealing gasket. The sealing pressure is not less than 0.1MPa when the cover is closed. The inner wall of the compartment is coated with an anti-corrosion coating and equipped with a small exhaust valve.

[0014] S500: Return Discharge, Re-inspection, and Experience Update: The loading status is monitored by the weight and volume sensors built into the collection compartment, and the return route is automatically planned after the preset threshold is reached; after the ship and aircraft arrive at the designated discharge point, the target object is transferred by docking with the container on shore through the vision alignment system; during the discharge, the UAV returns to the original operation area to perform a patrol re-inspection; after the operation is completed, the key information of this operation is integrated into the knowledge base through an incremental experience iteration algorithm.

[0015] Furthermore, in S100, during collaborative data acquisition, the UAV, equipped with infrared and visible light sensors, performs a full-coverage scan of the water area to capture the full-area distribution of the target object. The acquired data includes the target object's geographic coordinates, distribution range, outline shape, and temperature correlation information with the surrounding environment. The ship-mounted aircraft is equipped with multispectral sensors, hyperspectral sensors, 3D lidar, millimeter-wave radar, and environmental perception sensors to acquire the target object's material spectral characteristics, three-dimensional spatial position, movement trajectory, and water area parameters such as illumination, water flow, water transparency, and ambient temperature. All acquired data is synchronously associated and stored according to timestamps.

[0016] Furthermore, in S100, during collaborative data acquisition, the mathematical expression of the collaborative multi-source data fusion algorithm is: ,in, For the fused target object-environment integrated dataset, Weighting of drone data As the weight of the ship's machinery data, and For the global distribution data of target objects collected by drones, Detailed data on target objects and environmental parameters collected by the ship's aircraft. This is the ambient temperature correction factor. Due to the temperature deviation in the aquatic environment, For material recognition enhancement factor, Confidence level of hyperspectral material characteristics.

[0017] Furthermore, in S200, the mathematical expression for the multi-dimensional feature dynamic matching algorithm in environment adaptation and target recognition is: Where Match represents the target object matching degree. For feature dimension, For feature dynamic weights, Let i be the i-th eigenvalue of the current target object. The i-th dimension standard value of the preset feature template, Environmental confidence factors; extracted morphological features include equivalent diameter, contour irregularity, and area ratio; spectral features include characteristic band absorption peaks and reflectivity; temperature features include the temperature difference between the target object and the water body; and material features include density correlation values ​​and dielectric constant correlation parameters.

[0018] Furthermore, in S200, during environmental adaptation and target recognition, the preset feature template library contains standard feature datasets for two core target objects: surface debris and aquatic plants. The surface debris feature templates cover feature data for plastic bags, foam, metal cans, plastic bottles, paper waste, textile fragments, rubber products, batteries, and oil stains. The aquatic plant feature templates cover feature data for water hyacinth, duckweed, water peanut, pondweed, and goldfish algae. Each type of target object's feature template includes morphological feature baseline values, spectral feature baseline curves, temperature feature baseline ranges, and material feature baseline parameters. All feature data are categorized and indexed according to the target object category. The template library supports offline updates via wired transmission interfaces or wireless communication, and the updated feature data automatically overwrites the original corresponding templates.

[0019] Furthermore, in S300, during cooperative path planning and capture, the mathematical expression of the cooperative dynamic path planning algorithm is: ,in, The total path cost, The cost weights are 1, For distance cost, For water flow costs, For barrier costs, Cost of the target object.

[0020] Furthermore, in S300, during cooperative path planning and capture, obstacle costs... The calculation formula is: ,in, For obstacle weighting coefficients, The straight-line distance between the ship / engine and the obstacle; the formula for calculating distance cost is: ,in, This represents the straight-line distance from the ship's current position to the target area. Here is the path deflection factor; the formula for calculating water flow cost is: ,in, For water flow velocity, The countercurrent coefficient is used; the formula for calculating the cost of the target product is: ,in, For the density of the target object, This is the priority coefficient for the target object. For the quantity of the target item, the cost weight The system dynamically adjusts the weighting based on the water area operation scenario and governance needs. Specifically, the system allows users to preset weighting combinations for different water area scenarios such as lakes, rivers, and ports through the scenario mode selection function of the ship engine control system; or users can directly adjust the values ​​of each cost weight according to the current governance needs through the manual parameter configuration interface. The adjusted weights are then loaded into the path cost calculation module in real time, so that the path planning results can be adapted to different operation priorities and environmental characteristics.

[0021] Furthermore, in the S400 process of capacity reduction and classified storage, the execution steps for automatically planning the return route are as follows: the GPS positioning module obtains the geographical coordinates of the ship's current position and the designated discharge point, and extracts the geographical information of the shoreline, waterway, shoals, and reefs from the pre-stored electronic map of the water area; the ship avoids areas with adverse currents and busy waterways by referring to the real-time collected data on water flow direction, water flow speed, and wind speed; and the ship generates a return route that includes turning nodes and sailing speed by combining the real-time obstacle information fed back by the ship's obstacle avoidance sensors. The route information is then synchronized to the ship's navigation control module.

[0022] Furthermore, in the S500 process of return-to-port discharge, re-inspection, and experience update, the preset threshold is the loading trigger threshold of the collection compartment, which includes two types: volume threshold and weight threshold. Meeting either condition triggers the return-to-port command. The volume threshold is set based on the rated volume of the collection compartment and is 80% of the rated volume. The volume ratio occupied by the target object in the collection compartment is detected in real time by a volume sensor. The weight threshold is set based on the rated load of the collection compartment and is 75%-85% of the rated load. The cumulative weight of the target object in the collection compartment is detected in real time by a weight sensor. Both types of thresholds can be adjusted through the local operation interface of the ship's engine control system or a remote communication terminal. The adjusted threshold is automatically synchronized to the loading status monitoring module, and the monitoring module determines in real time whether the return-to-port trigger condition is met based on the adjusted threshold.

[0023] Furthermore, in the S500 process of return emission, re-inspection, and experience update, the mathematical expression of the incremental experience iteration algorithm is: ,in, For the updated knowledge base, For the existing knowledge base, For update rate, For efficiency weighting coefficients, These are empirical data related to work efficiency. It is experience data related to environmental adaptation.

[0024] Compared with existing technologies, this intelligent collaborative method for collecting surface waste has the following advantages:

[0025] I. This invention integrates global distributed monitoring and detailed parameter perception capabilities through a collaborative data acquisition mode of UAVs and ship-based aircraft. It constructs a visualized global distribution map and a comprehensive dataset, enabling all-round and three-dimensional perception of targets in water areas. Combining environmental adaptive parameter optimization technology and multi-dimensional feature dynamic matching algorithms, it accurately identifies various targets and aquatic plants, overcoming the limitations of traditional single identification methods. The data fusion process incorporates environmental and material-related correction logic to improve data reliability and identification accuracy. It solves the detection challenges caused by uneven distribution and diverse types of targets in complex water environments. This deep integration of collaborative perception and intelligent identification significantly improves the comprehensiveness and accuracy of target detection, providing solid data support for subsequent operations, significantly reducing the probability of missed detections and false detections, and enhancing the overall intelligence level of operations.

[0026] Second, this invention optimizes the operation route by integrating multiple influencing factors through a collaborative dynamic path planning algorithm, coupled with a real-time dynamic correction mechanism, to ensure the efficiency and safety of the operation route, avoiding the risks of ineffective navigation and obstacle collisions. It employs a targeted execution mechanism and an integrated design of volume reduction and classification to achieve efficient capture, volume reduction, and classified storage of target objects, improving operational efficiency and resource recycling rates. Leveraging an incremental experience-based iterative algorithm, key information from each operation is integrated into a knowledge base, driving continuous optimization of system performance and enhancing adaptability to different aquatic environments. This intelligent design, encompassing the entire process from path planning and operation execution to system iteration, not only improves the efficiency and stability of surface waste collection but also achieves rational resource recycling and effective environmental protection, promoting the development of surface waste management towards high efficiency, sustainability, and refinement. It has broad application value and promising prospects for promotion.

[0027] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0029] Figure 1 A flowchart illustrating the steps of an intelligent, collaborative method for collecting surface waste.

[0030] Figure 2 This is a diagram showing the input-output relationship of each step in the intelligent collaborative water surface waste collection method. Detailed Implementation

[0031] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0032] Implementation Case:

[0033] An example of an intelligent collaborative method for collecting surface waste in urban rivers, implemented in a core section.

[0034] This embodiment applies to a core section of an urban river, 5 kilometers long and 100-200 meters wide. Along the riverbanks, due to residential sewage discharge, tourist waste, and upstream floating debris, the water surface is covered with man-made waste such as plastic bottles, foam blocks, and plastic bags, mixed with natural waste such as tree branches and leaves. In some areas, aquatic plants easily blend in with the waste. The river flow velocity is stable at 0.3 meters per second. On that day, there was ample sunlight, and the water transparency was moderate. The riverbanks contain shoals, reefs, and fixed waterways. The designated discharge point is located at a waste transfer station on the downstream bank of the river. Specific steps are as follows... Figure 1 As shown.

[0035] S100, Collaborative Data Acquisition:

[0036] After the operation commenced, drones equipped with infrared and visible light sensors took off from the shore landing point to conduct a full-coverage cruise scan of a 5-kilometer stretch of inland waterway. This accurately captured the geographical coordinates, distribution range, and outline of targets on the water surface, while simultaneously recording temperature correlation information between the targets and their surrounding environment, providing fundamental data support for understanding the overall distribution of waste in the area. Simultaneously, a surface-operation vessel set sail, using its multispectral sensors, hyperspectral sensors, 3D lidar, millimeter-wave radar, and environmental perception sensors to closely collect the material spectral characteristics, three-dimensional spatial location, and movement trajectory of targets. It also simultaneously acquired parameters such as light intensity, water flow velocity, water transparency, and ambient temperature, providing detailed support for subsequent accurate target identification and operational path planning. All data collected by the drones and the vessel were synchronously linked and stored with timestamps to ensure data timeliness and relevance. Subsequently, a collaborative multi-source data fusion algorithm was used to deeply integrate the overall distribution data collected by the drones with the detailed data and environmental parameters collected by the vessel. The mathematical expression for the collaborative multi-source data fusion algorithm is as follows: ,in, For the fused target object-environment integrated dataset, Weighting of drone data As the weight of the ship's machinery data, and For the global distribution data of target objects collected by drones, Detailed data on target objects and environmental parameters collected by the ship's aircraft. This is the ambient temperature correction factor. Due to the temperature deviation in the aquatic environment, For material recognition enhancement factor, To enhance the confidence level of hyperspectral material characteristics and effectively compensate for the limitations of data collected by a single device, a comprehensive distribution map of target objects and a comprehensive target-environment dataset for the inland river section are ultimately generated. This clearly presents the dense distribution areas of various types of waste and the environmental conditions of the surrounding waters, providing a comprehensive and accurate data foundation for subsequent operations.

[0037] S200, Environmental Adaptation and Target Recognition:

[0038] Based on the aquatic environmental parameters in the comprehensive dataset generated by S100, the operation system initiates a multi-dimensional feature dynamic matching algorithm to comprehensively extract four core features of each target object: morphology, spectrum, temperature, and material. The mathematical expression of the multi-dimensional feature dynamic matching algorithm is as follows: Where Match represents the target object matching degree. For feature dimension, For feature dynamic weights, Let i be the i-th eigenvalue of the current target object. The i-th dimension standard value of the preset feature template, Environmental confidence factors are used to ensure the comprehensiveness and relevance of feature extraction. The system's built-in preset feature template library contains standard feature datasets for two core target categories: surface debris and aquatic plants. Each target category has corresponding morphological feature baseline values, spectral feature baseline curves, temperature feature baseline ranges, and material feature baseline parameters. All feature data is categorized and indexed by target category for easy and rapid comparison. The system accurately compares the extracted feature values ​​of the current target with the standard values ​​in the template library, and corrects the matching results using environmental confidence factors. This allows for precise differentiation between man-made waste such as plastic bottles and foam blocks, natural waste such as tree branches and leaves, and aquatic plants that need to be removed, avoiding accidental capture of aquatic plants or missed waste. This efficient identification of target objects provides clear and accurate data for subsequent capture operations. This preset feature template library has been updated offline via a wired transmission interface. Updated feature data automatically overwrites the original corresponding templates, ensuring the timeliness and accuracy of the template library's feature data and further improving the success rate of target recognition.

[0039] S300, Cooperative Path Planning and Capture:

[0040] Combining the global distribution map generated by S100, the target object identification results from S200, and the real-time collected water flow data, the operation system plans the optimal operation path using a collaborative dynamic path planning algorithm. The mathematical expression of the collaborative dynamic path planning algorithm is: ,in, The total path cost, The cost weights are 1, For distance cost, For water flow costs, For barrier costs, The target cost, of which the barrier cost The calculation formula is: ,in, The obstacle weighting coefficient. The straight-line distance between the ship / engine and the obstacle; the formula for calculating distance cost is: ,in, This represents the straight-line distance from the ship's current position to the target area. Here is the path deflection factor; the formula for calculating water flow cost is: ,in, For water flow velocity, The countercurrent coefficient is used; the formula for calculating the cost of the target product is: ,in, For the density of the target object, This is the priority coefficient for the target object. To determine the number of target objects, the algorithm comprehensively considers distance cost, water flow cost, obstacle cost, and target object cost. This minimizes navigation distance, reduces the impact of water flow on operations, avoids navigation risks, and improves waste collection efficiency. During planning, it avoids coastal shoals, reefs, and busy waterways to reduce navigational obstacles and safety hazards. It prioritizes operations in densely populated waste areas and optimizes the path along the water flow direction to reduce energy consumption and improve overall operational efficiency. During operations, the drone continuously patrols the air, updating the target object's coordinates in real time. When the waste location is detected to have shifted due to water flow, it promptly sends route correction commands to the vessel to ensure the vessel always travels in the accurate target direction, preventing capture failure due to target movement. Based on the received path information and target object identification results, the ship-mounted aircraft switches to the corresponding actuators for different types of waste. For large, hard waste such as plastic bottles and foam blocks, a clamping actuator is used to ensure stable capture. For lightweight, scattered waste such as plastic bags and tree branches and leaves, a net-like retrieval actuator is used to achieve comprehensive retrieval. The captured waste is smoothly transferred to the collection compartment through the guide channel on the ship-mounted aircraft deck, avoiding secondary pollution caused by waste falling during the transfer process and ensuring that all captured waste safely enters the collection compartment.

[0041] S400, capacity reduction and categorized storage:

[0042] Once the waste enters the collection chamber via the guide channel, the chamber automatically activates its dual-shaft crushing and compressing integrated device to crush and compress various types of waste, significantly reducing its volume and effectively improving the chamber's loading efficiency. This allows the chamber to hold more waste, reducing the number of times it needs to return to port for discharge and improving overall operational efficiency. The volume-reduced waste then enters a multi-layer screening mechanism, which precisely separates and grades the waste based on its material and particle size. Different types of waste, such as plastics, foams, and wood, are then directed to separate, sealed compartments within the collection chamber. This separate storage facilitates subsequent waste recycling and reduces sorting costs. Each sealed compartment is leak-proof, preventing waste degradation products from polluting water bodies or producing odors, ensuring the aquatic environment is not subject to secondary pollution, and maintaining a clean environment around the operating vessel.

[0043] S500: Return Discharge, Re-inspection, and Experience Updates

[0044] The collection compartment's built-in weight and volume sensors monitor the loading status in real time. The volume threshold is set at 80% of the compartment's rated volume, and the weight threshold is set at 75%-85% of its rated load. Meeting either threshold triggers a return command, ensuring the collection compartment doesn't become overloaded and thus avoids navigational risks, while fully utilizing the compartment's loading space. The system uses a GPS positioning module to obtain the vessel's current location and the geographical coordinates of the downstream discharge point. Combined with a pre-stored electronic map of the waterway, it extracts geographical information such as the shoreline, channels, shoals, and reefs. Referring to real-time water flow direction, speed, and wind speed data, it avoids areas with adverse currents and busy channels, reducing navigation resistance and safety hazards. Simultaneously, it combines real-time obstacle information from the vessel's obstacle avoidance sensors to generate a return path including turning points and navigation speed, ensuring a safe and efficient return route. This path is synchronized to the vessel's navigation control module, providing precise guidance for the return journey. After the vessel-machinery arrives at the discharge point along the planned path, it precisely docks with the waste transfer containers on shore using a visual alignment system, quickly completing the sorting and transfer of various types of waste, improving discharge efficiency, and reducing the time waste spends on the vessel-machinery. During the discharge, the drone returns to the original operating section to perform a patrol and re-inspection, focusing on checking for any missed waste in previously densely distributed waste areas and around the operating path, ensuring that the operating area is thoroughly cleaned and free of residue. After the entire operation is completed, the system uses an incremental experience-iterative algorithm to integrate key information from this operation into a knowledge base, including the coordinates of densely distributed waste areas, the capture efficiency of different types of waste, the impact of water flow on the operating path, and the correlation between environmental parameters and target recognition accuracy. The mathematical expression of the incremental experience-iterative algorithm is: ,in, For the updated knowledge base, For the existing knowledge base, For update rate, For efficiency weighting coefficients, These are empirical data related to work efficiency. It involves using relevant experience data on environmental adaptation to continuously enrich and optimize the knowledge base, providing a more scientific and efficient reference for subsequent operations in this river section and similar water areas, and continuously improving the overall intelligence level and operational effectiveness of the operations.

[0045] In summary, the implementation of this intelligent collaborative surface waste collection method in the core sections of urban rivers utilizes collaborative data collection by drones and boats, combined with a collaborative multi-source data fusion algorithm to construct a comprehensive dataset, laying a solid foundation for subsequent operations. A multi-dimensional feature dynamic matching algorithm accurately identifies waste and aquatic plants, avoiding accidental capture and omission. A collaborative dynamic path planning algorithm optimizes the operational route, coupled with targeted actuators to achieve efficient capture. Dual-axis crushing and compression, along with multi-layer screening, completes volume reduction and sorting, improving loading efficiency and subsequent processing convenience. A return-to-base discharge mechanism ensures navigation safety and accurate discharge, drone re-inspection ensures thorough cleanup, and an incremental experience-iterative algorithm continuously optimizes the knowledge base, such as... Figure 2 As shown, the entire process of collecting surface waste is intelligent, efficient, and environmentally friendly, adapting to the complex environment of inland rivers and the needs of waste cleanup.

[0046] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart collaborative method for collecting surface waste, characterized in that, The specific steps of this method are as follows: S100, Collaborative Data Acquisition: UAVs cruise to capture the full-domain distribution of targets, while ships collect detailed data on targets and aquatic environmental parameters. The data are then fused through a collaborative multi-source data fusion algorithm to form a full-domain distribution map and a comprehensive dataset. S200, Environmental Adaptation and Target Recognition: Based on the aquatic environmental parameters in the comprehensive dataset, a multi-dimensional feature dynamic matching algorithm is used to extract the morphological, spectral, temperature, and material features of the target object, and compare them with the preset feature template library to complete the target object recognition; S300, Collaborative Path Planning and Capture: Combining global distribution map, target identification results and real-time water flow data, the operation path is planned through a collaborative dynamic path planning algorithm; The drone updates the target's position coordinates in real time and sends route correction commands to the ship-to-aircraft aircraft. The ship's machinery switches actuators according to the type of target object, captures the target object, and transfers it to the collection compartment via a guide channel; S400, Volume Reduction Processing and Classified Storage: After receiving the target object, the collection chamber starts the dual-shaft crushing and compression integrated device to crush and compress the target object. The reduced-volume target material is separated by a multi-layer screening mechanism, and different types of target materials are introduced into independent sealed chambers. S500, return discharge, re-inspection and experience update: The loading status is monitored by the weight and volume sensors built into the collection compartment. After reaching the preset threshold, the return route is automatically planned. After the ship and aircraft reach the designated discharge point, the target object is transferred by docking with the container on shore through the vision alignment system. During the discharge, the UAV returns to the original operation area to perform a patrol re-inspection. After the assignment is completed, the key information of this assignment is integrated into the knowledge base through an incremental experience iteration algorithm.

2. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In S100, during collaborative data acquisition, the UAV, equipped with infrared and visible light sensors, performs a full-coverage scan of the water area to capture the full distribution of the target object. The acquired data includes the target object's geographic coordinates, distribution range, outline shape, and temperature correlation information with the surrounding environment. The ship-mounted aircraft is equipped with multispectral sensors, hyperspectral sensors, 3D lidar, millimeter-wave radar, and environmental perception sensors to acquire the target object's material spectral characteristics, three-dimensional spatial position, movement trajectory, and water area parameters such as illumination, water flow, water transparency, and environmental temperature. All acquired data is synchronously associated and stored according to timestamps.

3. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In S100, during collaborative data acquisition, the mathematical expression of the collaborative multi-source data fusion algorithm is: ,in, For the fused target object-environment integrated dataset, Weighting of drone data As the weight of the ship's machinery data, and For the global distribution data of target objects collected by drones, Detailed data on target objects and environmental parameters collected by the ship's aircraft. This is the ambient temperature correction factor. Due to the temperature deviation in the aquatic environment, For material recognition enhancement factor, Confidence level of hyperspectral material characteristics.

4. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In S200, during environment adaptation and target recognition, the mathematical expression for the multi-dimensional feature dynamic matching algorithm is: Where Match represents the target object matching degree. For feature dimension, For feature dynamic weights, Let i be the i-th eigenvalue of the current target object. The i-th dimension standard value of the preset feature template, Environmental confidence factors.

5. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In S200, during environmental adaptation and target recognition, the preset feature template library contains standard feature datasets for two core target objects: surface debris and aquatic plants. Each target object's feature template includes morphological feature baseline values, spectral feature baseline curves, temperature feature baseline ranges, and material feature baseline parameters. All feature data are classified and indexed according to the target object category. The template library supports offline updates via wired transmission interface or wireless communication. Updated feature data automatically overwrites the original corresponding template.

6. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In S300, during cooperative path planning and capture, the mathematical expression of the cooperative dynamic path planning algorithm is: ,in, The total path cost, The cost weights are 1, For distance cost, For water flow costs, For barrier costs, Cost of the target object.

7. The intelligent collaborative water surface waste collection method according to claim 6, characterized in that, In S300, during cooperative path planning and capture, obstacle cost... The calculation formula is: ,in, The obstacle weighting coefficient. The straight-line distance between the ship / engine and the obstacle; the formula for calculating distance cost is: ,in, This represents the straight-line distance from the ship's current position to the target area. Here is the path deflection factor; the formula for calculating water flow cost is: ,in, For water flow velocity, The countercurrent coefficient is used; the formula for calculating the cost of the target product is: ,in, For the density of the target object, This is the priority coefficient for the target object. The quantity of the target object.

8. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In the S400, the automatic planning of the return route in the capacity reduction and classification storage process is as follows: the geographical coordinates of the ship's current position and the designated discharge point are obtained through the GPS positioning module, and the geographical information of the shoreline, waterway, shoal and reef in the map is extracted by combining the pre-stored electronic map of the water area. Referring to real-time collected data on water flow direction, water flow speed, and wind speed, the system avoids areas with adverse currents and busy waterways. Combining real-time obstacle information fed back by the ship's obstacle avoidance sensors, a return path including turning points and sailing speed is generated, and the path information is synchronized to the ship's navigation control module.

9. The intelligent collaborative water surface waste collection method according to claim 1, characterized in that, In the S500, during return discharge, re-inspection, and experience update, the preset threshold is the loading trigger threshold of the collection compartment, which includes two types: volume threshold and weight threshold. If either condition is met, the return command is triggered. The volume threshold is set based on the rated volume of the collection chamber, which is 80% of the rated volume. The volume ratio occupied by the target object in the collection chamber is detected in real time by the volume sensor. The weight threshold is set based on the rated load of the collection compartment, which is 75%-85% of the rated load. The cumulative weight of the target object in the collection compartment is detected in real time by a weight sensor. Both types of thresholds can be adjusted through the local operation interface of the ship's engine control system or a remote communication terminal. The adjusted threshold is automatically synchronized to the loading status monitoring module, and the monitoring module determines in real time whether the return trigger condition is met based on the adjusted threshold.

10. The intelligent collaborative method for collecting surface waste according to claim 1, characterized in that, The mathematical expression for the incremental iterative algorithm in the S500's return-to-base emissions, re-inspection, and experience update processes is: ,in, For the updated knowledge base, For the existing knowledge base, For update rate, For efficiency weighting coefficients, These are empirical data related to work efficiency. It is experience data related to environmental adaptation.

Images