Image recognition guided waste site mosquito comprehensive management system

Abstract

The present application relates to the field of environmental sanitation treatment technology, specifically to a garbage site mosquito comprehensive treatment system guided by image recognition and targeted by probiotics, comprising: a multi-modal perception module that collects mosquito flight dynamic images, breeding point three-dimensional distribution images, ground surface environment texture images and underground leachate microbial data; a mosquito characteristic analysis module that generates dynamic characteristic data sets through an improved YOLO fusion algorithm; a probiotic delivery module that generates probiotic delivery area, path and dosage scheme through a deep reinforcement learning hybrid algorithm, and drives a delivery device to release composite mosquito inhibiting probiotics in a targeted manner; a traceability monitoring module that collects feedback data after release, generates adaptive feedback results through correlation analysis; and a collaborative decision hub that builds a mosquito population spatio-temporal sequence prediction model, optimizes probiotic delivery dosage, concentration ratio, injection angle and delivery frequency. Thus, the problems of insufficient treatment targeting and low comprehensive treatment efficiency in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of environmental sanitation management technology, specifically to a comprehensive mosquito control system for landfills that uses image recognition to guide the targeted action of probiotics. Background Technology

[0002] Landfills, as centralized sites for the disposal of urban solid waste, are breeding grounds for mosquitoes due to their high organic content, accumulation of leachate, and complex surface environment. Mosquitoes not only cause nuisance but also spread infectious diseases such as dengue fever and malaria, seriously threatening the health and public health safety of surrounding residents. They also disrupt the ecological balance and hinder the standardized operation of landfills. Current mosquito control methods at landfills largely rely on traditional approaches such as chemical spraying and physical trapping. Chemical agents easily pollute the soil and leachate, disrupt the microbial community structure, and mosquitoes are prone to developing resistance. Physical trapping methods have drawbacks, including limited coverage and low efficiency against low-density mosquitoes, making it difficult to address the dynamic changes and large-scale migrations of mosquito populations at landfills.

[0003] Current mosquito control technologies still have significant shortcomings: In terms of mosquito identification, there is a lack of multi-dimensional data fusion capabilities, making it difficult to accurately distinguish mosquito species, life cycle stages, and risk levels of breeding sites, resulting in insufficient targeted control; in the probiotic delivery process, most are delivered in a fixed area and at a fixed dose, without optimizing the path based on mosquito distribution dynamics and terrain features, resulting in wasted probiotics and poor colonization effects; at the same time, there is a lack of full-process traceability monitoring and dynamic optimization mechanisms, making it impossible to quantitatively link environmental changes, probiotic effects, and mosquito population fluctuations, and making it difficult to adjust control strategies based on real-time feedback, ultimately leading to low control efficiency, high costs, and an inability to fundamentally curb mosquito breeding and spread. Summary of the Invention

[0004] This application provides an integrated mosquito control system for landfills that uses image recognition to guide the targeted action of probiotics, in order to solve the problems of insufficient targeting and low efficiency of integrated control in existing technologies.

[0005] The first aspect of this application provides an image recognition-guided probiotic-targeting integrated mosquito control system for landfills, comprising: a multimodal perception module, a mosquito feature analysis module, a probiotic delivery module, a source tracing and monitoring module, and a collaborative decision-making center; wherein, the multimodal perception module is used to collect dynamic images of mosquito flight, three-dimensional distribution images of breeding sites, surface environmental texture images, and groundwater leachate microbial data; the mosquito feature analysis module is used to accurately identify mosquito species, peak density, migration paths, life cycle stages, and breeding site risk levels using an improved YOLO fusion algorithm, and generate a dynamic feature dataset; the probiotic delivery module is based on dynamic... The feature dataset, through a deep reinforcement learning hybrid algorithm, generates probiotic delivery areas, paths, and dosage schemes, and drives the delivery device to target and release compound mosquito-inhibiting probiotics. The source tracking and monitoring module is used to collect feedback data after the release of the compound mosquito-inhibiting probiotics. Through multi-dimensional data correlation analysis, it matches the correspondence between changes in environmental conditions and the probiotic colonization effect and mosquito population size, extracts core influencing factors, and generates adaptive feedback results. Based on the dynamic feature dataset and adaptive feedback results, the collaborative decision-making center constructs a spatiotemporal sequence prediction model for mosquito populations, dynamically optimizing the probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency.

[0006] Preferably, the multimodal perception module includes a multi-source image acquisition unit, a microbial data acquisition unit, and a data processing unit. The multi-source image acquisition unit is used to acquire dynamic images of mosquito flight in the landfill, three-dimensional distribution images of breeding points, and surface environmental texture images using a panoramic camera, lidar, and infrared thermal imager. The microbial data acquisition unit is used to acquire real-time data on the abundance of microbial communities in underground leachate using an embedded microbial sensor. The data processing unit is used to perform noise reduction, enhancement, and format standardization processing on the acquired image data, remove outliers from the microbial data, and generate a raw dataset in a unified format.

[0007] Preferably, the mosquito feature analysis module includes an image feature extraction unit, a multi-dimensional recognition unit, and a dynamic dataset generation unit. The image feature extraction unit receives the original dataset, inputs an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, optimizes the extraction accuracy of features at different scales through a feature pyramid, simultaneously captures key features of mosquito morphology, flight trajectory, breeding point texture, and spatiotemporal correlation, and strengthens the weight allocation of these key features. The multi-dimensional recognition unit, based on the... NarrativeKey features are used to accurately identify mosquito species, peak density, migration routes, and life cycle stages, and risk levels are determined based on mosquito density at breeding sites, reproduction rate, and spatiotemporal diffusion trends, resulting in multi-dimensional identification results. The dynamic dataset generation unit is used to integrate the multi-dimensional identification results and accurately associate them with timestamps and spatial coordinates to generate a dynamic feature dataset.

[0008] Preferably, the probiotic delivery module includes a delivery plan planning unit and a delivery equipment driving unit. The delivery plan planning unit, based on the dynamic feature dataset, generates an optimal delivery area and obstacle-avoidance delivery path by combining a deep reinforcement learning hybrid algorithm that integrates DQN and PPO with slope and obstacle distribution information in the landfill terrain data. At the same time, it initially matches a probiotic delivery dosage scheme based on mosquito density and breeding point risk level. The delivery equipment driving unit is used to drive drones and ground tracked delivery robots to target and release compound mosquito-inhibiting probiotics according to the delivery area, delivery path, and delivery dosage scheme.

[0009] Preferably, the traceability monitoring module includes a feedback data acquisition unit, a correlation analysis unit, and an adaptability result generation unit. The feedback data acquisition unit, through distributed environmental sensors and microbial colonization detection sensors, collects real-time data on changes in environmental parameters, probiotic colonization, and mosquito population changes after the release of the compound mosquito-inhibiting probiotics. The correlation analysis unit performs multi-dimensional data comparison and trend correlation on the environmental parameter change data, probiotic colonization data, and mosquito population change data to establish a quantitative correspondence between environmental condition changes and the probiotic colonization effect and mosquito population size. Based on this correspondence and combined with data fluctuation patterns, the adaptability result generation unit extracts the core factors affecting the probiotic effect and generates a feedback result on the adaptability between the landfill environmental conditions and the probiotic effect.

[0010] Preferably, the collaborative decision-making center includes a prediction model construction unit and a parameter optimization unit. The prediction model construction unit constructs a mosquito population spatiotemporal sequence prediction model based on the dynamic feature dataset and the adaptation feedback results, using the XGBoost extreme gradient boosting algorithm to generate prediction results of mosquito density change trends, life cycle stage evolution, and breeding point risk level migration over the next 72 hours. The parameter optimization unit is used to dynamically optimize the probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency based on the prediction results and the adaptation feedback results, and transmits the data to the probiotic delivery module in real time.

[0011] The second aspect of this application provides a method for a comprehensive mosquito control system for landfills using image recognition-guided probiotic targeting, comprising: acquiring dynamic images of mosquito flight, three-dimensional distribution images of breeding sites, surface texture images, and groundwater leachate microbial data; processing the aforementioned dynamic images of mosquito flight, three-dimensional distribution images of breeding sites, surface texture images, and groundwater leachate microbial data, and then inputting them into an improved YOLO fusion algorithm to accurately identify mosquito species, peak density, migration paths, life cycle stages, and breeding site risk levels, thereby generating a dynamic feature dataset. Based on the dynamic feature dataset, a hybrid deep reinforcement learning algorithm is used to generate probiotic delivery areas, paths, and dosage schemes. Simultaneously, the delivery device is driven to target and release compound mosquito-inhibiting probiotics, and feedback data after release is collected. Multi-dimensional data correlation analysis is performed to generate adaptation feedback results. Based on the adaptation feedback results and the dynamic feature dataset, a mosquito population spatiotemporal sequence prediction model is constructed using the XGBoost extreme gradient boosting algorithm. The probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency are dynamically optimized and transmitted to the delivery device for execution in real time.

[0012] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for a landfill mosquito integrated management system that uses image recognition to guide the targeted action of probiotics, as described in the above embodiments.

[0013] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for a landfill mosquito integrated management system with image recognition-guided probiotic targeting as described in the above embodiments.

[0014] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method for a landfill mosquito integrated management system that uses image recognition to guide the targeted action of probiotics, as described in the above embodiments.

[0015] Therefore, this application has the following beneficial effects: This application integrates image and microbial data through a multimodal perception module to comprehensively and accurately collect mosquito and environmental characteristics of landfills, providing complete data support for subsequent management. Relying on an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, it significantly improves the accuracy of mosquito species, density, and breeding site risk levels, enhancing the comprehensiveness and targeting of identification. A deep reinforcement learning hybrid algorithm is used to plan targeted delivery schemes, driving multiple types of equipment to accurately release compound probiotics, avoiding resource waste and improving probiotic colonization efficiency. A source tracking and monitoring module analyzes the correlation between the environment, probiotics, and mosquitoes, extracting core influencing factors and providing precise data feedback for strategy optimization. The collaborative decision-making center uses the XGBoost algorithm to build a predictive model, dynamically optimizing delivery parameters to achieve precise, dynamic, and intelligent mosquito management, improving the adaptability and efficiency of management. Therefore, it solves the problems of insufficient targeting and low comprehensive management efficiency in existing technologies.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the structure of the integrated mosquito control system for landfills, which uses image recognition to guide the targeted action of probiotics, according to an embodiment of this application. Figure 2 This is a schematic diagram of a multimodal sensing module according to an embodiment of this application; Figure 3 This is a schematic diagram of a mosquito feature parsing module provided according to an embodiment of this application; Figure 4 This is a schematic diagram of a probiotic delivery module according to an embodiment of this application; Figure 5 This is a schematic diagram of a traceability monitoring module according to one embodiment of this application; Figure 6 This is a schematic diagram of a collaborative decision-making center provided according to an embodiment of this application; Figure 7 This is a schematic diagram of a landfill mosquito integrated management system based on image recognition-guided probiotic targeting, provided in an embodiment of this application. Figure 8 This is a flowchart of a method for an image recognition-guided probiotic-targeting integrated mosquito control system for landfills, according to an embodiment of this application. Figure 9This is a schematic diagram of a method for a landfill mosquito integrated management system that uses image recognition to guide the targeted action of probiotics, according to an embodiment of this application. Figure 10 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The following description, with reference to the accompanying drawings, describes an embodiment of the integrated mosquito control system for landfills using image recognition-guided probiotic targeting. Addressing the low efficiency of integrated control mentioned in the background section, this application provides an integrated mosquito control system for landfills using image recognition-guided probiotic targeting. In this system, a multimodal perception module integrates image data and microbial data to comprehensively and accurately collect mosquito and environmental characteristics from the landfill, providing complete data support for subsequent control. Relying on an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, the system significantly improves the accuracy of mosquito species, density, and breeding site risk levels, enhancing the comprehensiveness and targeting of identification. A deep reinforcement learning hybrid algorithm is used to plan targeted delivery schemes, driving multiple types of equipment to accurately release compound probiotics, avoiding resource waste and improving probiotic colonization efficiency. A source tracking and monitoring module analyzes the correlation between the environment, probiotics, and mosquitoes, extracting core influencing factors and providing accurate data feedback for strategy optimization. A collaborative decision-making center uses the XGBoost algorithm to build a predictive model, dynamically optimizing delivery parameters to achieve precise, dynamic, and intelligent mosquito control, improving the adaptability and efficiency of control. This solves the problems of insufficient targeted governance and low efficiency of comprehensive governance in existing technologies.

[0020] Figure 1 This is a schematic diagram of the structure of the integrated mosquito control system for landfills, which uses image recognition to guide the targeted action of probiotics, as provided in an embodiment of this application.

[0021] This application provides an image recognition-guided probiotic targeted mosquito control system for landfills. The system 10 includes: a multimodal sensing module 100, a mosquito feature analysis module 200, a probiotic delivery module 300, a source tracing and monitoring module 400, and a collaborative decision-making center 500.

[0022] The multimodal perception module 100 is used to collect dynamic images of mosquito flight in the landfill, three-dimensional distribution images of breeding sites, surface environmental texture images, and microbial data of groundwater leachate. The mosquito feature analysis module 200 is used to accurately identify mosquito species, peak density, migration paths, life cycle stages, and breeding site risk levels using an improved YOLO fusion algorithm, and generate a dynamic feature dataset. The probiotic delivery module 300, based on the dynamic feature dataset, uses a deep reinforcement learning hybrid algorithm to generate probiotic delivery areas, paths, and dosage schemes. It drives the delivery device to target and release compound mosquito-inhibiting probiotics; the traceability monitoring module 400 is used to collect feedback data after the release of compound mosquito-inhibiting probiotics, and through multi-dimensional data correlation analysis, matches the correspondence between changes in environmental conditions and the colonization effect of probiotics and the number of mosquito populations, extracts core influencing factors, and generates adaptive feedback results; the collaborative decision center 500 constructs a spatiotemporal sequence prediction model of mosquito populations based on dynamic feature datasets and adaptive feedback results, and dynamically optimizes the probiotic delivery dosage, concentration ratio, spray angle and delivery frequency.

[0023] It is understood that in this embodiment, image data and microbial data are integrated through a multimodal perception module to comprehensively and accurately collect mosquito and environmental characteristics of landfills, providing complete data support for subsequent management. Relying on an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, the accuracy of mosquito species, density, and breeding site risk levels is significantly improved, enhancing the comprehensiveness and targeting of identification. A deep reinforcement learning hybrid algorithm is used to plan targeted delivery schemes, driving multiple types of equipment to accurately release compound probiotics, avoiding resource waste and improving probiotic colonization efficiency. A source tracking and monitoring module analyzes the correlation between the environment, probiotics, and mosquitoes, extracting core influencing factors and providing accurate data feedback for strategy optimization. The collaborative decision-making center uses the XGBoost algorithm to build a predictive model, dynamically optimizing delivery parameters to achieve precise, dynamic, and intelligent mosquito management, improving the adaptability and efficiency of management. Thus, it solves the problems of insufficient targeting and low comprehensive management efficiency in existing technologies.

[0024] In this embodiment of the application, the multimodal sensing module 100 includes: as follows Figure 2 As shown, there are a multi-source image acquisition unit, a microbial data acquisition unit, and a data processing unit.

[0025] The multi-source image acquisition unit is used to acquire dynamic images of mosquitoes flying in the landfill, three-dimensional distribution images of breeding points, and texture images of the surface environment through panoramic cameras, lidar, and infrared thermal imagers; the microbial data acquisition unit is used to acquire microbial community abundance data in underground leachate in real time through embedded microbial sensors; and the data processing unit is used to denoise, enhance, and standardize the format of the acquired image data, remove outliers from the microbial data, and generate a raw dataset in a unified format.

[0026] It is understood that this application embodiment uses a multi-source image acquisition unit, with the help of panoramic cameras, lidar, and infrared thermal imagers, to acquire dynamic images of mosquito flight, three-dimensional distribution images of breeding points, and surface environmental texture images of landfills. This comprehensively covers visual information of different scenes in landfills, effectively compensating for the limited field of view of traditional single-device acquisition and accurately capturing the intuitive characteristics of mosquito activity and breeding environment. Through the microbial data acquisition unit, embedded microbial sensors are used to collect microbial community abundance data in underground leachate in real time, breaking through the monitoring blind spot of focusing only on surface mosquitoes, deeply exploring the potential impact of underground microbial environment on mosquito breeding, and providing more comprehensive environmental dimension support for governance decisions. Through the data processing unit, image data is denoised, enhanced, and format standardized, and outlier values ​​are removed from microbial data, effectively filtering data noise and abnormal interference, ensuring the purity and format uniformity of the original data, and laying a solid data foundation for accurate analysis in subsequent modules.

[0027] For example, at a large suburban municipal solid waste landfill (with a daily processing capacity of 800 tons), during the peak mosquito season in the summer evenings, a multi-source image acquisition unit was used. Three panoramic cameras were deployed in the landfill operation area, leachate collection area, and surrounding green belt to capture the dynamic flight of mosquitoes in different areas in 360°, clearly recording the flight trajectory of Culex pipiens pallens swarms. Two lidar sensors scanned the low-lying waterlogged areas and waste accumulation slopes of the landfill, generating three-dimensional distribution images of breeding points and accurately marking three core breeding areas with a depth exceeding 10 cm. Four infrared thermal imagers identified high-density mosquito areas at night by using temperature differences, compensating for the nighttime monitoring limitations of visible light cameras. At the same time, six embedded microbial sensors were evenly distributed in the leachate collection trench to collect real-time data on the abundance of microbial communities such as bacteria and fungi in the underground leachate, capturing the abnormal feature that the microbial abundance around the breeding points was three times higher than that in other areas. The data processing unit then started, performing Gaussian denoising, contrast enhancement, and format standardization on the image data. It removed two sets of outliers in the microbial data caused by sensor fluctuations, and finally generated a unified format raw dataset containing mosquito dynamics, breeding point locations, environmental texture, and microbial abundance, providing high-quality, interference-free basic data support for subsequent mosquito feature analysis.

[0028] In this embodiment of the application, the mosquito feature analysis module 200 includes: as follows Figure 3 As shown, there are image feature extraction unit, multi-dimensional recognition unit, and dynamic dataset generation unit.

[0029] The image feature extraction unit receives the original dataset and inputs it into an improved YOLO fusion algorithm that incorporates the Transformer attention mechanism. It optimizes the extraction accuracy of features at different scales through a feature pyramid, simultaneously capturing key features such as mosquito morphology, flight trajectory, breeding point texture, and spatiotemporal correlation, and strengthens the weight allocation of key features. The multi-dimensional recognition unit accurately identifies mosquito species, density peaks, migration paths, and life cycle stages based on key features, and classifies risk levels according to mosquito density at breeding points, reproduction rate, and spatiotemporal diffusion trends, obtaining multi-dimensional recognition results. The dynamic dataset generation unit integrates the multi-dimensional recognition results and accurately correlates them with timestamps and spatial coordinates to generate a dynamic feature dataset.

[0030] It is understood that the embodiments of this application, through the image feature extraction unit, receive the original dataset and input it into the improved YOLO fusion algorithm that integrates the Transformer attention mechanism. The feature pyramid optimizes the extraction accuracy of features at different scales, simultaneously capturing key features such as mosquito morphology and flight trajectory and strengthening their weights, thereby improving the targeting and accuracy of feature extraction at different scales and solving the problem of insufficient feature capture by traditional algorithms for small-sized mosquitoes and complex environments. Through the multi-dimensional recognition unit, information such as mosquito species and density peaks is accurately identified based on key features, and the risk level of breeding points is determined according to preset indicators, completing the upgrade from single-feature recognition to multi-dimensional comprehensive analysis. This accurately distinguishes different mosquito species, clarifies the differences in density distribution and breeding risk, and provides precise targeting basis for subsequent governance strategy formulation. Through the dynamic dataset generation unit, the multi-dimensional recognition results are integrated and associated with timestamps and spatial coordinates to generate a dynamic feature dataset, deeply binding the recognition results with spatiotemporal information and intuitively presenting the dynamic change trajectory of mosquito populations, providing real-time and traceable data support for the dynamic adjustment of probiotic delivery schemes.

[0031] It should be noted that the image feature extraction unit inputs the original dataset into the improved YOLO fusion algorithm, optimizes the extraction accuracy of features at different scales through feature pyramid, and at the same time uses the Transformer attention mechanism to filter out mosquito morphology, flight trajectory, breeding point texture and spatiotemporal correlation information from the original data, and strengthens the weight of these key features in a targeted manner to form a feature set that focuses on the core information of mosquitoes and their breeding environment. The multi-dimensional recognition unit calls upon the recognition module of the improved YOLO fusion algorithm, using the extracted and enhanced key feature set as the core input. It accurately matches the mosquito morphological information in the key features with a preset mosquito species morphology database to quickly identify specific mosquito species. Then, it performs frame-by-frame statistical analysis on the mosquito targets corresponding to the key features, and calculates the peak mosquito density in a unit area by combining continuous data in the time dimension. By tracking the spatiotemporal coordinate changes of the key features in multi-frame data, it fits the migration path of the mosquito population. Based on the details of mosquito size and morphological maturity reflected by the key features, it determines the life cycle stage of the mosquito: egg, larva, pupa, or adult. Finally, it integrates the mosquito density data of the breeding points associated with the key features, and calculates the reproduction rate and diffusion trend by combining the analysis of spatiotemporal features. It also classifies the risk level of the breeding points according to the preset quantitative evaluation standards. Finally, it integrates all the recognition results of mosquito species, peak density, migration path, life cycle stage, and breeding point risk level to form a complete multi-dimensional recognition result.

[0032] The pre-set mosquito species morphology database refers to a database of key morphological characteristic parameters of different mosquito species, including wing veins, antennae, and body shape, which are collected in advance and stored in a structured manner.

[0033] The preset quantitative assessment standards are as follows: low-risk areas correspond to mosquito density ≤ 50 mosquitoes / m², reproduction rate ≤ 1.2 times / day, and spatiotemporal spread limited to within 5 meters of the breeding site with a spread speed ≤ 1 meter / hour; medium-risk areas correspond to mosquito density 51-200 mosquitoes / m², reproduction rate 1.3-2.0 times / day, spatiotemporal spread covering 5-20 meters of the breeding site with a spread speed of 1.1-3 meters / hour; high-risk areas correspond to mosquito density ≥ 201 mosquitoes / m², reproduction rate ≥ 2.1 times / day, spatiotemporal spread exceeding 20 meters of the breeding site with a spread speed ≥ 3.1 meters / hour.

[0034] The improved YOLO fusion algorithm refers to a visual algorithm that integrates the Transformer attention mechanism and feature pyramids within the YOLO architecture. It is specifically designed to accurately extract multi-scale features of mosquitoes from complex environments and directly output multi-dimensional identification results such as species, density, path, and risk level. The formula is: ; ; ; ; ; ; ; ; ; in, This is the l-th layer feature map after feature pyramid fusion; This is a convolution operation; This is an upsampling operation; This is the feature map of the (l+1)th layer of the feature pyramid; For feature splicing operations; This is the original feature map of the l-th layer output by the backbone network; This represents the number of output channels for the convolution operation. The query matrix for the attention mechanism; The multi-scale feature set output by the feature pyramid; The dimensions of the query matrix and the key matrix; The key matrix for the attention mechanism; The value matrix for the attention mechanism; The dimension of the value matrix; To scale the dot product attention output features; The normalized activation function; This is the transpose of the key matrix; This is the scaling factor; The output features are those obtained after multi-head attention fusion. For feature splicing operations; The output feature of the i-th attention head; For the number of attention heads; The projection matrix for multi-head attention output; This is the set of key features after weight enhancement; The feature fusion coefficient; is the peak mosquito density per unit area; is the number of mosquitoes counted at time t. The area of ​​the statistical region is expressed in square meters (m²). This represents the total duration of the statistical time window; Mosquito reproduction rate; Let t2 be the number of mosquitoes. Let be the number of mosquitoes at time t1; The time interval between two statistical analyses; The rate at which mosquitoes spread; This represents the incremental distance for mosquito migration. For time increments; This is the total loss function of the YOLO algorithm; Taxonomic loss due to mosquito species / life cycle stage; These are the weighting coefficients for the coordinate regression loss; The regression loss is for mosquito / breeding point coordinates; The weighting coefficients for the target confidence loss; Confidence loss for distinguishing mosquito targets from the background.

[0035] For example, when conducting precise mosquito control monitoring at a city's landfill, the image feature extraction unit receives raw images of key areas such as the landfill leachate pool and landfill area captured by a panoramic camera, along with auxiliary data such as temperature, humidity, and light intensity recorded by environmental sensors. This raw dataset is then input into an improved YOLO fusion algorithm that incorporates the Transformer attention mechanism. The feature pyramid optimizes the extraction accuracy of features at different scales, simultaneously capturing key features such as mosquito wing vein morphology, swarm flight trajectories, muddy textures at breeding sites, and the spatiotemporal correlation of mosquito activity at different times. Furthermore, the feature fusion coefficient α (optimized through multiple experiments in the landfill mosquito scenario and verified by core indicators such as mosquito species identification accuracy and density calculation error) strengthens the weight allocation of key features. Based on the aforementioned key features, the multi-dimensional identification unit matches mosquito morphological information with a pre-set morphology database to accurately identify dominant mosquito species such as Culex pipiens pallens and Anopheles sinensis. It statistically analyzes mosquito targets frame-by-frame in consecutive frames, calculating the peak mosquito density per unit area based on the statistical region area. It tracks changes in mosquito coordinates across multiple frames, fitting a migration path from the leachate pond to the surrounding green belt of the landfill. Based on mosquito size and morphological maturity, it determines the mosquito's life cycle stage (egg, larva, pupa, adult). Simultaneously, it classifies risk levels according to mosquito density at breeding sites, reproduction rate, and spatiotemporal diffusion trends, referencing pre-set low, medium, and high quantification standards, thus obtaining multi-dimensional identification results. The dynamic dataset generation unit integrates these results, precisely linking them with the timestamp of each frame and the spatial coordinates of the landfill zones to generate a dynamic feature dataset, providing accurate data support for the subsequent development of targeted pest control strategies.

[0036] In this embodiment of the application, the probiotic delivery module 300 includes: Figure 4 As shown, the delivery scheme planning unit and the delivery equipment drive unit are shown.

[0037] The delivery plan planning unit, based on a dynamic feature dataset, uses a deep reinforcement learning hybrid algorithm that integrates DQN and PPO, combined with slope and obstacle distribution information from landfill terrain data, to generate the optimal delivery area and obstacle-avoidance delivery path. At the same time, it preliminarily matches the probiotic delivery dosage scheme based on mosquito density and breeding point risk level. The delivery equipment driving unit is used to drive drones and ground tracked delivery robots to target and release compound mosquito-inhibiting probiotics according to the delivery area, delivery path, and delivery dosage scheme.

[0038] It is understood that, through the delivery plan planning unit, based on a dynamic feature dataset, a deep reinforcement learning hybrid algorithm integrating DQN and PPO is used to generate the optimal delivery area, obstacle avoidance path, and dosage plan by combining the slope and obstacle information of the landfill terrain. This ensures that the delivery plan closely matches the actual terrain conditions and mosquito distribution characteristics of the landfill, significantly improving the targeting and rationality of the management action. Through the delivery equipment driving unit, drones and ground tracked delivery robots are driven to release probiotics in a targeted manner according to the plan, giving full play to the synergistic advantages of drones and ground robots to complete full coverage delivery in multiple scenarios, both in the air and on the ground. This ensures that the probiotics are accurately delivered to the target area, reducing resource waste while lowering the labor intensity and safety risks of manual delivery.

[0039] It should be noted that the deep reinforcement learning hybrid algorithm is an intelligent decision-making model that combines the advantages of DQN (Deep Q-Network) and PPO (Proximal Policy Optimization), and the formula is: ; ; ; ; ; in, Action decision-making strategy for the release of probiotics into landfills; The strategy fusion weights are 0.6 to 0.8. The distribution is based on a random delivery strategy; The action of dispensing probiotics; The environmental condition of the landfill; This is the optimal delivery action; The action value function; Update the objective function for the delivery strategy; These are the network parameters for the current delivery strategy; The network parameters are those of the old delivery strategy before the update. For expectation operators; The distribution of landfill disposal actions for the old strategy; This represents the distribution of landfill disposal actions under the current strategy. The dominant function; This is the clipping function; The cropping factor (default 0.2); Let the mean squared error loss function be used. These are the current Q network parameters; For instant rewards; To determine the environmental condition of the landfill at the next moment; The discount factor is (0.9~0.99). For the next delivery action; The action value function of the target Q-network; The target Q network parameters; The environmental state of the landfill at time t+1; This is the state transition function; Let t represent the environmental state of the landfill. The probiotic dispensing action performed at time t; The environmental observation noise at time t; Let be the total reward value at time t; The reward weight coefficient is matched to the delivery area; (the sum is 1, determined by the experiment and optimization in the landfill scenario, highlighting the delivery of key breeding areas and the priority of obstacle avoidance). Match rewards to the designated areas; Optimize the reward weighting coefficient for the investment path; Rewards for route optimization; Dose-matched reward weighting coefficient For dose matching rewards; The obstacle avoidance reward weighting coefficient; Rewards for obstacle avoidance.

[0040] For example, in mosquito control operations at a suburban landfill, the delivery plan planning unit first acquires a dynamic feature dataset, including the coordinates of high-risk breeding areas around the leachate pond, peak mosquito density per unit area (up to 80 mosquitoes / m²), and the risk level of the breeding point (high risk). Simultaneously, it accesses landfill topography data (including a 30° slope area, three large garbage heaps, and the distribution of obstacles such as the leachate pond). Then, the unit calls a deep reinforcement learning hybrid algorithm combining DQN and PPO: the PPO's random strategy explores feasible delivery paths in the slope area, while the optimal action selection from DQN locks in the core delivery area within 50 meters of the leachate pond, ultimately generating an obstacle avoidance path (bypassing the garbage heaps). Simultaneously, the dosage of compound probiotics (200 mL / m²) is matched according to the high-risk level. Next, the delivery equipment drive unit starts two multi-rotor drones and one ground tracked delivery robot: the drones cover the slope area along the planned path and spray probiotics precisely; the tracked robot enters the complex terrain between the garbage piles and completes the targeted release in the area near the leachate pool according to the dosage plan.

[0041] In this embodiment of the application, the traceability monitoring module 400 includes, as follows: Figure 5 As shown, there are a feedback data acquisition unit, a correlation analysis unit, and an adaptation result generation unit.

[0042] The feedback data acquisition unit, through distributed environmental sensors and microbial colonization detection sensors, collects real-time data on changes in environmental parameters, probiotic colonization, and mosquito population after the release of the compound mosquito-inhibiting probiotics. The correlation analysis unit performs multi-dimensional data comparison and trend correlation on the data on changes in environmental parameters, probiotic colonization, and mosquito population to establish a quantitative correspondence between changes in environmental conditions and the probiotic colonization effect and mosquito population. The adaptability result generation unit, based on the correspondence and combined with the data fluctuation pattern, extracts the core factors affecting the effect of probiotics and generates the adaptability feedback results between the landfill environmental conditions and the effect of probiotics.

[0043] It is understood that, through the feedback data acquisition unit, and with the help of distributed environmental sensors and microbial colonization detection sensors, this application embodiment collects environmental parameters, colonization amount, and mosquito population change data in real time after the release of probiotics. This comprehensively captures the environmental dynamics, colonization effect, and mosquito population changes after the release of probiotics, ensuring that the treatment effect is traceable and assessable. Through the correlation analysis unit, the three types of data are compared and correlated in multiple dimensions to establish quantitative correspondences. This clearly establishes the quantitative correlation between environmental conditions, probiotic effects, and mosquito population changes, solving the problem of ambiguous attribution of effects in traditional treatment and providing a clear causal basis for optimizing treatment effects. Through the adaptability result generation unit, core influencing factors are extracted and adaptability feedback results are output. This accurately identifies the key environmental factors affecting the effect of probiotics and outputs adaptability feedback that fits the actual situation of the landfill, providing a precise and implementable adjustment basis for the strategy optimization of the collaborative decision-making center. It should be noted that the correlation analysis unit automatically aligns environmental parameter change data, probiotic colonization data, and mosquito population change data according to the time dimension through a preset multi-dimensional data correlation rule library, synchronously associating the fluctuation characteristics and change sequence of data in each dimension; based on preset threshold ranges and quantitative mapping rules, it establishes a quantitative correspondence between the magnitude of environmental condition changes and the attenuation rate of probiotic colonization effect and the growth rate of mosquito population. The suitability result generation unit, based on the quantitative correspondence, automatically captures the environmental factor with the highest correlation in data fluctuations as the core influencing factor, and combines it with preset suitability judgment criteria to generate suitability feedback results between landfill environmental conditions and the effect of probiotics.

[0044] The pre-set multi-dimensional data association rule base is a set of pre-set rules based on professional experience in landfill management, historical monitoring data, and experimental verification. It clarifies the cross-correspondence between the change thresholds of each data dimension and the fluctuations of data in different dimensions. Specifically, it includes environmental humidity (≤65% is low humidity, 65%~80% is medium humidity, and >80% is high humidity), probiotic colonization (≥1.0×10⁻⁶), and other relevant rules. 6 CFU / mL represents effective colonization, 0.5 × 10⁻⁶. 6 ~1.0×10 6 CFU / mL indicates attenuation and colonization, <0.5×10 6 CFU / mL represents the threshold range for changes in probiotic colonization and mosquito population (≥20% indicates a significant decrease, ±10% indicates a slight change, and ≥15% indicates a significant increase), as well as the quantitative mapping relationship of data from different dimensions: when humidity ≤65%, it corresponds to a significant decrease in effective probiotic colonization and mosquito population; when humidity is 65%~80%, it corresponds to a slight change in probiotic colonization and mosquito population; and when humidity >80%, it corresponds to a significant increase in ineffective probiotic colonization and mosquito population.

[0045] The pre-set suitability assessment criteria are based on the environmental remediation goals of landfills, clarifying the quantitative judgment rules for whether the effects of probiotics are suitable for environmental conditions. Specifically, they are divided into three levels: suitable (excellent), basically suitable (good), and unsuitable (poor), with a probiotic colonization rate ≥ 1.0 × 10⁻⁶. 6 A complete fit is defined as a CFU / mL concentration and a significant decrease in mosquito population (≥20%), with a probiotic colonization rate of 0.5 × 10⁻⁶. 6 ~1.0×10 6 A CFU / mL concentration and a small change (±10%) in mosquito population size indicate a basic fit, with a probiotic colonization rate <0.5×10⁻⁶. 6 A mosquito population with a concentration of CFU / mL and a significant increase in mosquito population (≥15%) is considered an incompatible match and requires adjustment of the release program.

[0046] For example, after a suburban landfill in a certain city administered a compound probiotic for mosquito control, the feedback data acquisition unit used distributed environmental sensors to capture real-time changes in environmental parameters such as humidity (72%) and temperature (28°C). Using a microbial colonization detection sensor, the probiotic colonization rate was determined to be 0.8 × 10⁻⁶. 6 The data included CFU / mL, and image acquisition was used to monitor a 5% decrease in mosquito population compared to before the release. The correlation analysis unit called a pre-set multi-dimensional data association rule library, aligned the three types of data by time dimension, and compared trends. Based on the quantitative mapping relationship that "65%~80% humidity corresponds to the decline and colonization of probiotics and a slight change in mosquito population," a quantitative correspondence relationship was established between changes in environmental humidity and the colonization effect of probiotics and changes in mosquito population. Based on this correspondence relationship and combined with the data fluctuation pattern, the suitability result generation unit extracted environmental humidity as the core factor affecting the effect of probiotics. According to the pre-set suitability judgment criteria, it was determined that the current environmental conditions and the effect of probiotics were basically suitable, and finally generated the suitability feedback result: "The current environmental humidity of the landfill is 72%, the probiotic colonization is in the decline and colonization range, the mosquito population has decreased slightly, no adjustment of the release plan is required, and it is recommended to continue monitoring humidity changes."

[0047] In this embodiment of the application, the collaborative decision-making center 500 includes, as follows: Figure 6 As shown, the prediction model construction unit and parameter optimization unit are shown.

[0048] The prediction model building unit, based on the dynamic feature dataset and adaptation feedback results, constructs a spatiotemporal sequence prediction model for mosquito populations using the XGBoost extreme gradient boosting algorithm, generating prediction results for the mosquito density change trend, life cycle stage evolution, and breeding point risk level migration over the next 72 hours. The parameter optimization unit is used to dynamically optimize the probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency based on the prediction results and adaptation feedback results, and transmits the data to the probiotic delivery module in real time.

[0049] It is understood that, through the prediction model construction unit, based on the dynamic feature dataset and the adaptation feedback results, the XGBoost algorithm is used to construct a mosquito population spatiotemporal sequence prediction model to complete the relevant prediction results for the next 72 hours. This realizes the transformation from passively responding to mosquito problems to actively predicting population changes, and allows for the understanding of mosquito density trends, life cycle evolution, and risk migration directions 72 hours in advance, thus providing sufficient preparation time for control actions. Through the parameter optimization unit, key parameters such as probiotic delivery dosage and concentration ratio are dynamically optimized based on the prediction results and adaptation feedback, and transmitted to the delivery module in real time. This ensures that the probiotic delivery strategy is always synchronized with mosquito population changes and environmental adaptation needs, guaranteeing the sustainability, efficiency, and stability of integrated mosquito control in landfills.

[0050] It should be noted that the mosquito population spatiotemporal sequence prediction model is an intelligent analysis model based on the XGBoost algorithm, capable of jointly predicting the evolutionary trends of mosquito density, population structure, and risk level in the temporal and spatial dimensions within a specific future period. The formula is as follows: ; ; ; in, For the first A spatial grid in the future Hourly multi-objective prediction output vector; For output mapping function; Index for historical time steps; The total number of integrated decision trees; For the first The predicted output of each decision tree; The input feature vector; For model bias terms; For future prediction time steps (values ​​from 1 to 72); This is the total loss function of the model; The set of all parameters of the model; For training sample index; This represents the number of training samples; For multi-objective loss functions; For the first One sample in The true label vector at any given moment; Regularization terms for the decision tree; This is the regularization coefficient (which controls the penalty for loss based on the number of leaf nodes in the decision tree). For the first The number of leaf nodes in a decision tree; This is the regularization coefficient (the penalty for controlling the weight of leaf nodes in the decision tree). For the first The weight vector of the leaf nodes of a decision tree.

[0051] The parameter optimization unit receives the mosquito population spatiotemporal sequence prediction model's output for the next 72 hours, including mosquito density trends, life cycle stage evolution, and breeding site risk level migration results. Simultaneously, it correlates this with current suitability feedback results. Through a pre-set lightweight decision-making logic, it delineates key control areas based on breeding site risk level migration. Then, combining the mosquito density growth rate with key life cycle stages, it adjusts the probiotic delivery parameters accordingly: if a high-risk area is predicted and the current suitability is "unsuitable," the delivery dose is increased by 15%-20%, and the delivery time is shortened. The delivery frequency is increased to 70%-80% of the original cycle, the concentration ratio is optimized to enhance environmental adaptability, and the spray angle is adjusted to focus on the core area of ​​breeding sites. If a medium-risk area is predicted and it is basically adaptable, the delivery frequency is only slightly adjusted to 90% of the original cycle, while the dosage and concentration ratio remain unchanged to adapt to the slight fluctuation trend of mosquitoes. If a low-risk area is predicted and it is fully adaptable, the original delivery parameters are maintained, and the spray angle is only slightly adjusted according to the slight fluctuation of environmental parameters to ensure uniform coverage. Finally, the optimized parameters generated by the lightweight decision-making logic are transmitted to the probiotic delivery module in real time.

[0052] For example, the predictive model building unit for a city's waste transfer station is based on a dynamic feature dataset (including spatiotemporal and environmental features such as current ambient humidity of 85%, temperature of 30℃, mosquito density over the past 72 hours, and waste decomposition degree) and adaptation feedback results (current probiotic colonization rate of 0.4 × 10⁻⁶). 6 (The mosquito population increased by 18% compared to before the release, indicating "incompatibility"). A spatiotemporal sequence prediction model for the mosquito population, constructed using the XGBoost extreme gradient boosting algorithm, was used to generate predictions for the northwest area of ​​the transfer station as a high-risk breeding area, with larvae entering a peak period and the risk level migrating to surrounding green areas within the next 72 hours. The parameter optimization unit then combined the prediction results with the compatibility feedback of "incompatibility" to dynamically optimize the probiotic delivery parameters: increasing the delivery dose in the northwest area and surrounding green areas by 18%, reducing the delivery frequency from once every 24 hours to once every 17 hours (71% of the original cycle), adjusting the concentration ratio to increase the proportion of high humidity-tolerant components to enhance environmental compatibility, and adjusting the spray angle from 30° horizontally to 45° to focus on the core area of ​​the migrated breeding points. Finally, the optimized parameters were transmitted to the probiotic delivery module in real time.

[0053] The image recognition-guided probiotic-targeting integrated mosquito control system for landfills proposed in this application integrates image and microbial data through a multimodal perception module, comprehensively and accurately collecting mosquito and environmental characteristics from the landfill, providing complete data support for subsequent control. Relying on an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, it significantly improves the accuracy of mosquito species, density, and breeding site risk levels, enhancing the comprehensiveness and targeting of identification. A deep reinforcement learning hybrid algorithm is used to plan targeted delivery schemes, driving multiple types of equipment to accurately release compound probiotics, avoiding resource waste and improving probiotic colonization efficiency. A source tracking and monitoring module analyzes the correlation between the environment, probiotics, and mosquitoes, extracting core influencing factors and providing precise data feedback for strategy optimization. The collaborative decision-making center uses the XGBoost algorithm to build a predictive model, dynamically optimizing delivery parameters to achieve precise, dynamic, and intelligent mosquito control, improving the adaptability and efficiency of the control. Therefore, it solves the problems of insufficient targeting and low efficiency of integrated control in existing technologies.

[0054] The following is a specific example of an image recognition-guided probiotic-targeting integrated mosquito control system for landfills, such as... Figure 7 As shown, it includes: A suburban landfill in a certain city experiences high temperatures and humidity during the summer due to the subtropical monsoon climate (average daily temperature 32℃, humidity 85%). This rapid decomposition of waste leads to rampant mosquito breeding, with the dominant species being Culex pipiens quinquefasciatus and Aedes albopictus. Breeding sites are concentrated in waterlogged depressions within the landfill area, leachate drainage ditches, and surrounding weed beds, severely impacting the lives of residents in the surrounding communities. To address this, the landfill deployed an integrated mosquito control system using image recognition to guide the targeted action of probiotics. Upon system activation, the multimodal sensing module initiated comprehensive data collection across the entire area. Based on the landfill's 100-acre area and its functional zoning of the "core landfill area - leachate treatment area - buffer green area," staff deployed 18 panoramic cameras, 6 LiDAR sensors, and 12 infrared thermal imagers to form a multi-source image acquisition unit. Simultaneously, 20 embedded microbial sensors were buried in key locations such as the leachate drainage ditches and beneath the landfill structure to form a microbial data acquisition unit. All equipment was connected to the data processing unit. The multi-source image acquisition unit operated continuously for 72 hours. A panoramic camera captured dynamic images of mosquito swarms flying in the core landfill area. LiDAR scanning generated a three-dimensional distribution image of breeding points, clearly marking the coordinates and depths of eight key waterlogged breeding sites. An infrared thermal imager identified surface textures and hidden mosquito gathering areas through temperature differences. Buried microbial sensors collected real-time data on the abundance of microbial communities in the underground leachate, detecting a putrefactive bacteria abundance of 1.2 × 10⁻⁶ in the leachate. 9CFU / mL; the data processing unit then performs Gaussian denoising, contrast enhancement, and format standardization on the acquired image data, removes three sets of outliers caused by sensor malfunctions in the microbial data, and finally generates a unified format raw dataset containing image data and microbial data, with each data point accompanied by a collection timestamp and spatial coordinate information.

[0055] After the original dataset is generated, the mosquito feature analysis module immediately begins its analysis. The image feature extraction unit first receives the original dataset and inputs the image data into an improved YOLO fusion algorithm that incorporates the Transformer attention mechanism. This algorithm optimizes the extraction accuracy of features at different scales through a feature pyramid structure, accurately capturing subtle features such as mosquito wing vein morphology and flight trajectory. Simultaneously, it strengthens the weight allocation of key features such as breeding point texture and spatiotemporal correlation, effectively avoiding interference from the complex background of the landfill on identification. Based on the extracted key features, the multi-dimensional recognition unit successfully identified the main mosquito species in the landfill as Culex pipiens quinquefasciatus (78%) and Aedes albopictus (22%), and calculated the core landfill... The peak mosquito density in the burial area reached 65 mosquitoes / m². The migration path of mosquitoes from the burial area to the northeast buffer green area was mapped. It was determined that most mosquitoes were in the larval and pupa stages of their life cycle. Based on the mosquito density, reproduction rate, and spatiotemporal diffusion trend at the breeding sites, 8 key breeding sites were designated as high-risk areas and 12 secondary areas were designated as medium-risk areas. The dynamic dataset generation unit then integrated the above multi-dimensional identification results, accurately correlated information such as mosquito species, density, and migration path with timestamps and spatial coordinates, and generated a dynamic feature dataset with dimensions of [time step × area number × 15 features], which was transmitted to subsequent modules in real time.

[0056] After the dynamic feature dataset is output, the probiotic delivery module quickly initiates the scheme planning and equipment driving process. The delivery scheme planning unit receives the dynamic feature dataset, calls a deep reinforcement learning hybrid algorithm that integrates DQN and PPO, and combines the distribution information of obstacles such as 35° steep slopes and large garbage piles in the landfill terrain data. After iterative optimization, the algorithm generates the optimal delivery area—prioritizing the coverage of 8 high-risk breeding sites and areas along mosquito migration paths. At the same time, it plans 3 obstacle avoidance delivery paths to avoid steep slopes and large obstacles. The algorithm also preliminarily matches the delivery dosage scheme according to the mosquito density and risk level of different areas: the delivery dosage is set at 2.5L / ㎡ for high-risk breeding sites, 1.8L / ㎡ for medium-risk areas, and 1.2L / ㎡ for low-risk buffer zones. The delivery equipment drive unit responded immediately, launching 6 agricultural drones and 8 ground-tracked delivery robots. The drones sprayed the mosquitoes in the high-altitude mosquito-gathering areas and steep slopes of the landfill area according to the planned path, while the ground robots carried out precise delivery to low-lying water breeding points, leachate diversion ditches and other ground areas. The two worked together to release a compound mosquito-inhibiting probiotic containing Bacillus thuringiensis and Bacillus spheroidis. The delivery trajectory was corrected in real time through image recognition throughout the process to ensure that the probiotic accurately covered the target area.

[0057] After the probiotics were added, the traceability monitoring module immediately entered the feedback monitoring and analysis phase. The feedback data acquisition unit activated a distributed environmental sensor and microbial colonization detection sensor network. Thirty distributed environmental sensors collected data on changes in environmental parameters such as temperature, humidity, wind speed, and leachate pH every hour, monitoring that the humidity in the core landfill area remained at 83% and the temperature at 31°C 24 hours after addition. Fifteen microbial colonization detection sensors collected probiotic colonization samples every three hours, and the data showed that the probiotic colonization at high-risk breeding sites was 0.6 × 10⁶. 6 CFU / mL, the medium-risk zone is 0.9×10 6 CFU / mL; Simultaneously, mosquito population changes were continuously monitored using an infrared camera. Statistics 48 hours after release showed a 12% decrease in mosquito numbers in high-risk areas and an 18% decrease in medium-risk areas. The correlation analysis unit performed multi-dimensional comparisons of the three types of data, aligning environmental parameters, colonization rate, and mosquito numbers for the same area according to time series, establishing a correlation that "when humidity > 80%, probiotic colonization rate < 1.0 × 10⁻⁶". 6 The quantitative correlation between "CFU / mL and mosquito population decrease of <15%" was established. Based on this correlation and the data fluctuation pattern, the adaptability result generation unit extracted environmental humidity as the core factor affecting the effect of probiotics, and finally generated the adaptability feedback results: high-risk breeding sites had average probiotic colonization effect due to high humidity, and the adaptability level was "basic adaptability"; medium-risk areas had suitable environmental conditions and the adaptability level was "adaptable"; low-risk areas had good adaptability.

[0058] After the source tracing and monitoring module outputs the adaptability feedback results, the collaborative decision-making center initiates prediction and optimization. The prediction model construction unit calls upon the dynamic feature dataset and the adaptability feedback results, and uses the XGBoost extreme gradient boosting algorithm to construct a spatiotemporal sequence prediction model for mosquito populations. Using the dynamic feature dataset (including historical mosquito data, environmental parameters, mosquito species, migration routes, breeding site risk levels, and other multi-dimensional features) and the adaptability feedback results (including adaptability level and core influencing factors) as core inputs, it generates a prediction result for the next 72 hours: the humidity in the core landfill area will drop to 78% in the next 24 hours, and the probiotic colonization rate is expected to increase to 1.2 × 10⁻⁶. 6 With a target concentration of CFU / mL, mosquito density will further decrease by 25%. However, due to impending rainfall, the risk level in the leachate treatment area will upgrade from medium to high, and mosquito larvae will enter their peak period. Based on this prediction and adaptation feedback, the parameter optimization unit immediately and dynamically optimized the delivery parameters: increasing the delivery dose in the leachate treatment area from 1.8 L / m² to 2.2 L / m² (a 22% increase), reducing the delivery frequency from once every 24 hours to once every 18 hours, adjusting the concentration ratio, increasing the proportion of moisture-resistant strains to 40%, and adjusting the drone spray angle from 30° to 45° to cope with the spread of accumulated water after rainfall; maintaining the same dose in high-risk breeding sites, only slightly adjusting the delivery frequency to once every 20 hours. All optimized parameters are transmitted in real time to the probiotic delivery module, driving the equipment to start a new round of precise delivery. At the same time, the multimodal sensing module restarts data acquisition, initiating a closed-loop treatment process of "collection-analysis-delivery-monitoring-optimization" to ensure the continuous and stable mosquito suppression effect in the landfill.

[0059] In summary, this application's embodiments utilize a multimodal perception module to collect comprehensive and accurate images and microbial data, providing a solid foundation for subsequent mosquito feature analysis. A dynamic feature dataset is generated by the mosquito feature analysis module, enabling precise identification and quantification of key information such as mosquito species, density, and breeding site risk, thus providing a scientific basis for targeted application. A probiotic delivery module plans the optimal path and targets the delivery of probiotics, improving their effectiveness while significantly reducing ineffective resource consumption. A source tracking and monitoring module collects feedback data and generates adaptation results, monitoring the probiotic's effectiveness and environmental compatibility in real time, providing reliable feedback for subsequent optimization. A collaborative decision-making center builds a model based on the dynamic feature dataset and adaptation results to predict trends and optimize delivery parameters, achieving dynamic adjustment and precise adaptation of the control strategy. The coordinated operation of these modules significantly improves the accuracy and efficiency of mosquito control in landfills, ensuring continuous and stable control results.

[0060] Next, referring to the accompanying drawings, a method for a comprehensive mosquito control system for landfills based on image recognition-guided probiotic targeting, according to an embodiment of this application, is described.

[0061] like Figure 8 As shown, the method for an integrated mosquito control system for landfills guided by image recognition to target probiotics includes the following steps: In step S101, dynamic images of mosquitoes flying in the landfill, three-dimensional distribution images of breeding points, texture images of the surface environment, and microbial data of groundwater leachate are acquired.

[0062] It is understood that the embodiments of this application comprehensively capture the mosquito activity status, core breeding areas, surface environmental characteristics, and underground microbial substrate information by acquiring dynamic images of mosquito flight in landfills, three-dimensional distribution images of breeding points, surface environmental texture images, and underground leachate microbial data. This breaks through the limitations of traditional monitoring that only focuses on surface mosquitoes, providing comprehensive basic information for subsequent management and improving the pertinence and effectiveness of integrated management from the source.

[0063] In step S102, the dynamic images of mosquito flight in the landfill, the three-dimensional distribution images of breeding sites, the texture images of the surface environment, and the microbial data of groundwater leachate are processed and then input into the improved YOLO fusion algorithm to accurately identify mosquito species, peak density, migration path, life cycle stage, and risk level of breeding sites, and generate a dynamic feature dataset.

[0064] Among them, the dynamic feature dataset refers to a structured data set that integrates multi-dimensional identification results such as mosquito species, density, migration path, life cycle and breeding site risk, and is linked with time and space information in real time.

[0065] It is understood that the embodiments of this application, by generating a dynamic feature dataset, transform the originally isolated and static environmental information into a unified data view that integrates mosquito species, density, migration routes, life cycle stages, and breeding site risk levels, and is linked to spatiotemporal coordinates in real time. This provides accurate and structured factual evidence for subsequent intelligent decision-making, enabling governance actions to shift from an experience-driven, extensive model to a data-driven, precise, and dynamic intervention model.

[0066] In step S103, based on the dynamic feature dataset, a deep reinforcement learning hybrid algorithm is used to generate the probiotic delivery area, path, and dosage scheme. At the same time, the delivery device is driven to release the compound mosquito-inhibiting probiotics in a targeted manner, and feedback data after release is collected to perform multi-dimensional data correlation analysis and generate adaptation feedback results.

[0067] Among them, the adaptation feedback results refer to the quantitative correspondence between changes in environmental parameters, the effect of probiotic colonization and the effect of mosquito population control, and the analytical conclusions that accurately extract the core influencing factors.

[0068] It is understood that the embodiments of this application generate adaptive feedback results, transforming the complex relationship between deployment actions and environmental responses into clear quantitative evidence. This provides direct causal insights for strategy optimization, clearly revealing the actual efficacy of probiotics under different environmental conditions. It enhances the adaptability of the governance strategy, enabling each deployment decision to be precisely corrected based on objective effectiveness feedback, thereby continuously optimizing resource allocation and steadily promoting the overall mosquito control effect towards high efficiency, precision, and sustainability.

[0069] For example, in a suburban landfill in a certain city during the summer, the low-lying, waterlogged areas in the core landfill zone are high-breeding areas for mosquitoes. Monitoring showed that the dynamic characteristic dataset showed that Culex pipiens quinquefasciatus accounted for 80% of the mosquito population, with a mosquito density of 60 mosquitoes / m², and the mosquitoes were spreading along the main road of the landfill towards the northeast green buffer zone. Based on this data, a deep reinforcement learning hybrid algorithm combining DQN and PPO was used to generate a dosage of 2.4 L / m² for high-risk waterlogged areas and 1.7 L / m² for medium-risk areas along the main road. The deployment area was delineated to cover the core waterlogged area and migration paths, and obstacle avoidance paths were planned to avoid the garbage accumulation. Subsequently, agricultural drones were used to target steep slopes, and ground spraying equipment was used to target low-lying waterlogged areas, synergistically releasing a compound mosquito-inhibiting probiotic containing Bacillus thuringiensis and Bacillus sphaeroides. Continuous data collection after deployment revealed that in the high-risk waterlogged areas, due to the high temperature and humidity of summer (humidity consistently maintained at 82% and temperature at 31℃), the probiotic colonization rate was only 0.55 × 10⁻⁶. 6 CFU / mL, mosquito numbers decreased by only 13% compared to before release; humidity along the main road in medium-risk area was 76%, and probiotic colonization was 0.9×10⁻⁶. 6 CFU / mL, mosquito population decreased by 19%. Multi-dimensional data correlation analysis generated adaptation feedback results: the high humidity environment in the core waterlogged area of ​​the landfill significantly reduced the colonization ability of probiotics, resulting in a weaker mosquito suppression effect compared to medium humidity areas. The core influencing factor was environmental humidity. High-risk waterlogged areas require targeted adjustments to enhance the moisture tolerance of probiotics. This result provides a direct basis for subsequent deployment strategy adjustments. In subsequent deployments, the proportion of moisture-tolerant probiotic strains in high-risk waterlogged areas was increased to 42%, and the dosage was fine-tuned to 2.7 L / m², precisely matching the high-humidity environmental characteristics of the waterlogged areas and improving mosquito suppression effectiveness.

[0070] In step S104, based on the adaptation feedback results and combined with the dynamic feature dataset, a spatiotemporal sequence prediction model for mosquito populations is constructed using the XGBoost extreme gradient boosting algorithm. The probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency are dynamically optimized and transmitted to the delivery device for execution in real time.

[0071] Understandably, the embodiments of this application construct a mosquito population prediction model, transforming historical data and real-time feedback into accurate predictions of future breeding trends. This enables control actions to shift from passive response to proactive intervention, allowing for advance resource allocation and preventative suppression in high-risk areas. Based on dynamically adjusted delivery parameters according to the prediction results, the utilization efficiency and timing of probiotic resources are improved, thereby achieving a more sustained and stable regional mosquito control effect while reducing the overall dosage.

[0072] The method for integrated mosquito control system in landfills, based on image recognition-guided probiotic targeting proposed in this application, integrates image and microbial data through a multimodal perception module to comprehensively and accurately collect mosquito and environmental characteristics from the landfill, providing complete data support for subsequent control. It leverages an improved YOLO fusion algorithm incorporating the Transformer attention mechanism to significantly improve the accuracy of mosquito species, density, and breeding site risk levels, enhancing the comprehensiveness and targeting of identification. A deep reinforcement learning hybrid algorithm is used to plan targeted delivery schemes, driving multiple types of equipment to accurately release compound probiotics, avoiding resource waste and improving probiotic colonization efficiency. A source tracking and monitoring module analyzes the correlation between the environment, probiotics, and mosquitoes, extracting core influencing factors and providing precise data feedback for strategy optimization. A collaborative decision-making center uses the XGBoost algorithm to build a predictive model, dynamically optimizing delivery parameters to achieve precise, dynamic, and intelligent mosquito control, improving the adaptability and efficiency of the control. This solves the problems of insufficient targeting and low efficiency in integrated control in existing technologies.

[0073] The following will illustrate a specific embodiment of a method for integrated mosquito control in landfills using image recognition-guided probiotic targeting, such as... Figure 9 As shown, it includes: A garbage transfer station in the main urban area of ​​a certain city receives nearly a thousand tons of municipal solid waste daily. The high temperature and humidity in the transfer workshop and open storage area during the summer lead to severe mosquito breeding, particularly Culex pipiens pallens and Aedes albopictus. Breeding sites are concentrated in the gaps of the garbage storage tanks, leachate collection pools, and surrounding drainage ditches, impacting the health of transfer station staff and the lives of nearby residents. To address this problem, an integrated pest management method for garbage dumps, using image recognition to guide the targeted action of probiotics, was adopted. The first step was to acquire comprehensive data across the entire area. Workers deployed 12 high-definition infrared cameras and 4 lidar sensors in the transfer workshop, and 8 panoramic cameras around the open-air temporary storage area and leachate collection pool. Additionally, 10 embedded microbial sensors were buried at the bottom of the leachate collection pool. These devices collected data continuously for 48 hours, capturing dynamic images of mosquito swarms in the transfer workshop, three-dimensional distribution images of breeding points in the gaps of the temporary storage tanks and the leachate pool, textures of surface waste residue, and environmental textures of damp areas. The data also showed that the abundance of putrefactive bacteria in the leachate reached 1.5 × 10⁻⁶.9 The CFU / mL microbial data provided a complete set of basic data on mosquito activity and the surrounding environment.

[0074] The collected image data underwent Gaussian denoising, brightness enhancement, and format standardization. After removing three sets of abnormal microbial data caused by sensor malfunctions, the processed data was input into an improved YOLO fusion algorithm incorporating the Transformer attention mechanism. The algorithm optimized feature extraction accuracy at different scales through a feature pyramid, accurately identifying that 82% of the mosquitoes at the transfer station were Culex pipiens pallens and 18% were Aedes albopictus. The peak mosquito density in the transfer workshop reached 55 mosquitoes / m², indicating that mosquitoes were migrating from the leachate collection pool into the transfer workshop, with most currently in adult or larval stages. Based on mosquito density, reproduction rate, and diffusion trend, the area surrounding the leachate collection pool and temporary storage tank was classified as high-risk, the interior of the transfer workshop as medium-risk, and the open passageway as low-risk. These identification results were then integrated and precisely correlated with the collection timestamps and spatial coordinates of each area to generate a dynamic feature dataset with dimensions of [time step × region number × 16 features].

[0075] Based on the generated dynamic feature dataset, a hybrid deep reinforcement learning algorithm combining DQN and PPO was used, along with terrain information such as the internal equipment layout and passage width of the transfer station. Through iterative algorithm optimization, an optimal deployment plan was generated: the deployment dose around the high-risk leachate collection pool and temporary storage tank was set at 2.2 L / m², 1.6 L / m² inside the medium-risk transfer workshop, and 1.1 L / m² in the low-risk open passages. The deployment area prioritized coverage of the core breeding sites and mosquito migration routes, and three obstacle-avoidance deployment paths were planned to avoid obstacles such as transfer machinery and garbage cans. Subsequently, two small agricultural drones were deployed to target the high-altitude areas of the transfer workshop, while four ground-based tracked sprayers targeted the ground and crevices, releasing a compound mosquito-inhibiting probiotic containing Bacillus thuringiensis and Bacillus subtilis. Continuous data collection after deployment showed that the humidity in the high-risk area remained at 84%, the temperature at 32℃, and the probiotic colonization rate was 0.5 × 10⁻⁶. 6 CFU / mL, the number of mosquitoes decreased by 14% after 48 hours compared to before release; in medium-risk areas with 78% humidity and 30℃ temperature, the planting density was 0.95×10 6 CFU / mL, mosquito numbers decreased by 21%. Through multi-dimensional data comparison and trend correlation analysis, the following adaptation feedback results were generated: high humidity environment leads to the decline of probiotic colonization effect, and poor ventilation in the transfer workshop exacerbates this phenomenon. The core influencing factors are environmental humidity and ventilation conditions. In high-risk areas, it is necessary to strengthen the moisture resistance of probiotics and optimize the distribution coverage.

[0076] Combining the adaptation feedback results with the previously generated dynamic feature dataset, a spatiotemporal sequence prediction model for mosquito populations was constructed using the XGBoost extreme gradient boosting algorithm, generating predictions for the next 72 hours: In the next 24 hours, due to rainfall, humidity in the transfer workshop will rise to 86%, and mosquito density may rebound to 48 mosquitoes / m², maintaining a high risk level; humidity around the leachate collection pool will remain high, and mosquito larvae will enter their peak period. Based on these predictions, delivery parameters were dynamically optimized: the delivery dosage around the transfer workshop and leachate collection pool was increased to 2.5 L / m², the delivery frequency was reduced from once every 24 hours to once every 16 hours, the probiotic concentration ratio was adjusted, increasing the proportion of moisture-resistant strains from 35% to 43%, and the spray angle of the ground spraying equipment was adjusted from 30° to 40° to enhance coverage of the gaps in the temporary storage tank. The optimized parameters were transmitted to the delivery equipment in real time, and the equipment immediately started a new round of precise delivery, while simultaneously restarting data collection and entering the next round of treatment to ensure stable mosquito control.

[0077] In summary, this application's embodiments, targeting the main urban waste transfer station scenario, accurately acquire multiple types of data, including mosquito dynamics, breeding point distribution, surface environment, and leachate microorganisms, thus solidifying the data foundation for governance. An improved YOLO fusion algorithm is used to process and generate a dynamic feature dataset with associated spatiotemporal information, enabling accurate identification and quantification of key mosquito information. A deep reinforcement learning hybrid algorithm is used to plan an obstacle-avoidance delivery scheme adapted to the transfer station equipment layout and to target the delivery of probiotics. Simultaneously, feedback data is collected to generate adaptation results covering core influencing factors such as humidity and ventilation. Combining the dynamic feature dataset and adaptation results, the XGBoost algorithm is used to model and predict mosquito trends under environmental changes such as rainfall, dynamically optimizing delivery parameters to effectively address the mosquito breeding problem at transfer stations and ensure stable and controllable governance results.

[0078] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0079] When the processor 1002 executes the program, it implements the method of the integrated mosquito control system for landfills with image recognition-guided probiotic targeting provided in the above embodiments.

[0080] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0081] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0082] The memory 1001 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0083] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0084] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0085] The processor 1002 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0086] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for a comprehensive mosquito control system for landfills using image recognition-guided probiotic targeting.

[0087] Furthermore, this application also provides a computer program product, including a computer program or instructions, which, when executed, enables the implementation of the above-mentioned method for a comprehensive mosquito control system for landfills that guides probiotics to target specific areas using image recognition.

[0088] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0089] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0090] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0091] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0092] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0093] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A comprehensive mosquito control system for landfills using image recognition-guided probiotic targeting, characterized in that: include: The system comprises a multimodal perception module, a mosquito feature analysis module, a probiotic delivery module, a source tracing and monitoring module, and a collaborative decision-making center; among which, The multimodal sensing module is used to collect dynamic images of mosquito flight in landfills, three-dimensional distribution images of breeding points, texture images of the surface environment, and microbial data of groundwater leachate. The mosquito feature analysis module is used to accurately identify mosquito species, peak density, migration path, life cycle stage, and breeding site risk level through an improved YOLO fusion algorithm, and generate a dynamic feature dataset. The probiotic delivery module, based on a dynamic feature dataset, uses a deep reinforcement learning hybrid algorithm to generate probiotic delivery areas, paths, and dosage schemes, and drives the delivery device to target and release compound mosquito-inhibiting probiotics. The source tracking and monitoring module is used to collect feedback data after the release of the compound mosquito-inhibiting probiotics. Through multi-dimensional data correlation analysis, it matches the correspondence between changes in environmental conditions and the colonization effect of probiotics and the number of mosquito populations, extracts core influencing factors, and generates adaptive feedback results. Based on the dynamic feature dataset and the adaptation feedback results, the collaborative decision-making center constructs a spatiotemporal sequence prediction model for mosquito populations and dynamically optimizes the probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency.

2. The image recognition-guided probiotic-targeting integrated mosquito control system for landfills according to claim 1, characterized in that, The multimodal perception module includes a multi-source image acquisition unit, a microbial data acquisition unit, and a data processing unit. The multi-source image acquisition unit is used to acquire dynamic images of mosquito flight, three-dimensional distribution images of breeding points, and surface environmental texture images of the landfill using a panoramic camera, lidar, and infrared thermal imager. The microbial data acquisition unit is used to acquire microbial community abundance data in groundwater in real time using an embedded microbial sensor. The data processing unit is used to perform noise reduction, enhancement, and format standardization processing on the acquired image data, remove outliers from the microbial data, and generate a raw dataset in a unified format.

3. The image recognition-guided probiotic-targeting integrated mosquito control system for landfills according to claim 1, characterized in that, The mosquito feature analysis module includes an image feature extraction unit, a multi-dimensional recognition unit, and a dynamic dataset generation unit. The image feature extraction unit receives the original dataset, inputs an improved YOLO fusion algorithm incorporating the Transformer attention mechanism, optimizes the extraction accuracy of features at different scales through a feature pyramid, simultaneously captures key features of mosquito morphology, flight trajectory, breeding point texture, and spatiotemporal correlation, and strengthens the weight allocation of these key features. The multi-dimensional recognition unit, based on the... Narrative Key features are used to accurately identify mosquito species, peak density, migration routes, and life cycle stages, and risk levels are determined based on mosquito density at breeding sites, reproduction rate, and spatiotemporal diffusion trends, resulting in multi-dimensional identification results. The dynamic dataset generation unit is used to integrate the multi-dimensional identification results and accurately associate them with timestamps and spatial coordinates to generate a dynamic feature dataset.

4. The image recognition-guided probiotic-targeting integrated mosquito control system for landfills according to claim 1, characterized in that, The probiotic delivery module includes a delivery plan planning unit and a delivery equipment driving unit. The delivery plan planning unit, based on the dynamic feature dataset, uses a deep reinforcement learning hybrid algorithm that integrates DQN and PPO, combined with slope and obstacle distribution information in the landfill terrain data, to generate the optimal delivery area and obstacle-avoidance delivery path. At the same time, it preliminarily matches the probiotic delivery dosage scheme according to mosquito density and breeding point risk level. The delivery equipment driving unit is used to drive drones and ground tracked delivery robots to target and release compound mosquito-inhibiting probiotics according to the delivery area, delivery path, and delivery dosage scheme.

5. The image recognition-guided probiotic-targeting integrated mosquito control system for landfills according to claim 1, characterized in that, The source tracing and monitoring module includes a feedback data acquisition unit, a correlation analysis unit, and an adaptability result generation unit. The feedback data acquisition unit, through distributed environmental sensors and microbial colonization detection sensors, collects real-time data on changes in environmental parameters, probiotic colonization, and mosquito population changes after the release of the compound mosquito-inhibiting probiotics. The correlation analysis unit performs multi-dimensional data comparison and trend correlation on the environmental parameter changes, probiotic colonization, and mosquito population changes to establish a quantitative correspondence between environmental condition changes and the probiotic colonization effect and mosquito population size. Based on this correspondence and combined with data fluctuation patterns, the adaptability result generation unit extracts the core factors affecting the probiotic effect and generates a feedback result on the adaptability between the landfill environmental conditions and the probiotic effect.

6. The image recognition-guided probiotic-targeting integrated mosquito control system for landfills according to claim 1, characterized in that, The collaborative decision-making center includes a prediction model construction unit and a parameter optimization unit. The prediction model construction unit constructs a mosquito population spatiotemporal sequence prediction model based on the dynamic feature dataset and the adaptation feedback results, using the XGBoost extreme gradient boosting algorithm to generate prediction results of mosquito density change trends, life cycle stage evolution, and breeding point risk level migration in the next 72 hours. The parameter optimization unit is used to dynamically optimize the probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency according to the prediction results and the adaptation feedback results, and transmits the data to the probiotic delivery module in real time.

7. A method for an image recognition-guided probiotic-targeting integrated pest management system for landfill mosquitoes, applicable to any one of claims 1-6, characterized in that, include: Acquire dynamic images of mosquitoes flying in landfills, three-dimensional distribution images of breeding sites, texture images of the surface environment, and microbial data of groundwater leachate; After processing the dynamic images of mosquito flight in the landfill, the three-dimensional distribution images of breeding sites, the surface environmental texture images, and the microbial data of groundwater leachate, the data are input into the improved YOLO fusion algorithm to accurately identify mosquito species, peak density, migration path, life cycle stage, and risk level of breeding sites, and generate a dynamic feature dataset. Based on the dynamic feature dataset, a deep reinforcement learning hybrid algorithm is used to generate probiotic delivery areas, paths, and dosage schemes. At the same time, the delivery device is driven to release compound mosquito-inhibiting probiotics in a targeted manner, and feedback data after release is collected. Multi-dimensional data correlation analysis is performed to generate adaptation feedback results. Based on the adaptation feedback results and the dynamic feature dataset, a spatiotemporal sequence prediction model for mosquito populations is constructed using the XGBoost extreme gradient boosting algorithm. The probiotic delivery dosage, concentration ratio, spray angle, and delivery frequency are dynamically optimized and transmitted to the delivery device for execution in real time.

8. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method of the image recognition-guided probiotic-targeting integrated mosquito control system for landfills as described in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they implement the method of the integrated mosquito control system for landfills with image recognition-guided probiotic targeting as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed, they implement the method of the integrated mosquito control system for landfills with image recognition-guided probiotic targeting as described in claim 7.

Images