Image recognition-based wasteyard mosquito suppression system for precisely throwing probiotics

Abstract

The invention relates to the technical field of disease and pest control, in particular to an image recognition-based garbage dump mosquito suppression system for precisely throwing probiotics, which comprises an image acquisition and sensing module for acquiring image data, environmental parameters, mosquito characteristic data and garbage decomposition state data; the mosquito feature analysis center extracts mosquito species, density and gathering area features, and constructs a mosquito breeding risk assessment model in combination with environmental parameters; the breeding risk assessment module fuses landform, garbage distribution and mosquito activity data, assesses a mosquito breeding risk level and a key breeding area, and generates a breeding risk probability cloud picture; the probiotic delivery decision-making unit simulates inhibition effects of different probiotic formulas and delivery strategies, generates a delivery scheme, and drives the ground robot to perform targeted delivery; the self-adaptive regulation and control module monitors mosquito density change and probiotic activity in real time, and dynamically adjusts a putting scheme through a swarm intelligence algorithm. Therefore, the problems of high pollution cost, poor accuracy and the like in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of pest and disease control technology, specifically to a probiotic precise delivery system for mosquito control in landfills based on image recognition. Background Technology

[0002] Landfills are high-risk breeding grounds for mosquitoes, which not only transmit various diseases but also pose a serious threat to the living environment and health of surrounding residents. Currently, mosquito control in landfills mainly relies on chemical spraying, physical trapping, and traditional probiotic application. While traditional probiotic application is environmentally friendly and pollution-free, it is mostly done manually at fixed times and in fixed quantities, lacking real-time monitoring and accurate assessment of mosquito breeding conditions. Due to differences in garbage type, pile thickness, and humidity, mosquito density varies in different areas of a landfill. Blindly applying probiotics manually can easily lead to insufficient application in densely populated areas and excessive application in sparsely populated areas, resulting in wasted probiotics and poor control effects.

[0003] However, traditional probiotic application methods, which inhibit mosquito larvae growth by administering probiotics containing mosquito pathogens, have the advantages of being environmentally friendly and pollution-free. However, existing methods mostly involve manual, timed, and quantitative application, which cannot accurately measure mosquito breeding density in different areas of the landfill. Environmental factors such as the type of waste, pile thickness, and humidity vary in different areas of the landfill, leading to uneven mosquito breeding density. Blindly applying probiotics manually wastes them; insufficient application in densely populated areas results in poor suppression, while excessive application in sparsely populated areas increases costs.

[0004] Furthermore, existing technologies lack real-time monitoring and precise assessment methods for mosquito breeding in landfills, making it impossible to obtain timely information on the distribution range and density of mosquito larvae. This further affects the accuracy of probiotic application and the mosquito-suppressing effect. Therefore, there is an urgent need for a landfill mosquito suppression system capable of real-time monitoring of mosquito breeding and precise probiotic application to address the problems of high pollution costs and poor accuracy in existing technologies. Summary of the Invention

[0005] This application provides a probiotic precise delivery system for suppressing mosquitoes in landfills based on image recognition, in order to solve the problems of high pollution costs and poor accuracy in existing technologies.

[0006] The first aspect of this application provides a probiotic-based precision delivery system for mosquito control in landfills, comprising: an image acquisition and sensing module, a mosquito feature analysis center, a breeding risk assessment module, a probiotic delivery decision unit, and an adaptive control module. The image acquisition and sensing module collects landfill image data, environmental parameters, mosquito population characteristic data, and landfill decomposition status data. The mosquito feature analysis center extracts mosquito species, density, and aggregation area characteristics, and constructs a mosquito breeding risk assessment model based on environmental parameters. The breeding risk assessment module, based on the mosquito breeding risk assessment model, integrates topographic data, landfill distribution data, and mosquito activity data to assess the mosquito breeding risk level and key breeding areas, generating a breeding risk probability cloud map. The probiotic delivery decision unit, based on the breeding risk probability cloud map, simulates the inhibition effects of different probiotic formulations and delivery strategies, generates a delivery plan, and drives a ground robot for targeted delivery. The adaptive control module monitors mosquito density changes and probiotic activity in real time, and dynamically adjusts the delivery plan using a swarm intelligence algorithm.

[0007] Preferably, the image acquisition and sensing module includes an image acquisition unit, an environmental parameter acquisition unit, a mosquito population characteristic acquisition unit, and a garbage decomposition status acquisition unit. The image acquisition unit is used to acquire panoramic and close-up images of the landfill at different times, covering areas with high mosquito activity. The environmental parameter acquisition unit is used to acquire data on temperature, humidity, light intensity, precipitation, and wind speed, and simultaneously record the acquisition time and location information. The mosquito population characteristic acquisition unit collects data on mosquito morphological characteristics, activity trajectories, and habitat preferences. The garbage decomposition status acquisition unit collects data on the humidity, pH, odor level, and decomposition stage of the garbage, and, combined with the appearance of the garbage in the images, comprehensively judges the degree of decomposition.

[0008] Preferably, the mosquito feature analysis center includes an image recognition unit, a feature extraction unit, and a model building unit. The image recognition unit uses the GhostYOLO algorithm to accurately identify mosquito species and distinguish between adults and larvae. The feature extraction unit extracts core features such as mosquito density, spatial distribution of aggregation areas, and peak activity periods. The model building unit correlates the extracted mosquito features with environmental parameters to construct a multi-factor coupled mosquito reproductive risk assessment model and outputs a quantitative value of reproductive potential.

[0009] Preferably, the breeding risk assessment module includes a risk level assessment unit, a key area positioning unit, and a cloud map generation unit. The risk level assessment unit integrates topographic data, garbage distribution data, and mosquito activity data, and uses the analytic hierarchy process (AHP) to determine the weights of each factor, classifying mosquito breeding risk into four levels: low, medium, high, and extremely high. The key area positioning unit uses risk value overlay analysis to determine key breeding areas and marks area boundaries and core locations. The cloud map generation unit uses a kernel density estimation algorithm to map risk probability values ​​to a three-dimensional geographic information model of the landfill, and uses color gradients to represent the risk intensity distribution, generating a breeding risk probability cloud map that includes risk level, key area coordinates, and risk diffusion trends.

[0010] Preferably, the probiotic delivery decision unit includes a formula screening module, an inhibition effect simulation unit, a delivery plan generation unit, and a targeted delivery drive unit. The formula screening module selects suitable probiotic strain combinations based on mosquito species and the state of waste decomposition. The inhibition effect simulation unit, based on the mosquito inhibition mechanism of different probiotic formulas and combined with a breeding risk probability cloud map, simulates the decrease in mosquito density under different formulas, dosages, delivery times, and delivery frequencies using a multi-objective optimization algorithm, evaluating the inhibition effect and duration. The delivery plan generation unit determines the specific probiotic formula, single delivery dosage, delivery frequency, precise delivery area, and optimal delivery time based on the simulation results, forming a standardized delivery plan. The targeted delivery drive unit plans the delivery path of the ground robot, generating a target movement path based on the coordinates of key breeding areas and combined with landfill terrain and obstacle information, driving the robot to reach the delivery area and complete the targeted delivery operation.

[0011] Preferably, the adaptive control module includes a real-time monitoring unit and a scheme adjustment unit. The real-time monitoring unit is used to continuously monitor changes in mosquito density, probiotic activity, and fluctuations in environmental parameters. The scheme adjustment unit analyzes the monitoring data, explores the correlation between changes in mosquito density and probiotic administration, and dynamically adjusts the administration scheme from three dimensions: administration dosage, administration frequency, and probiotic formula, to ensure a continuous and stable inhibitory effect.

[0012] The second aspect of this application provides a method for precise probiotic delivery to suppress mosquitoes in landfills based on image recognition, comprising: acquiring landfill image data, environmental parameters, mosquito population characteristic data, and landfill decomposition status data; based on the landfill image data and mosquito population characteristic data, extracting mosquito species, density, and aggregation area characteristics using the GhostYOLO algorithm, and constructing a mosquito breeding risk assessment model by combining the environmental parameters and the landfill decomposition status data; according to the mosquito breeding risk assessment model, integrating topographic data, landfill distribution data, and mosquito activity data to assess the mosquito breeding risk level and key breeding areas, and generating a breeding risk probability cloud map; based on the breeding risk probability cloud map, simulating the suppression effects of different probiotic formulations and delivery strategies, generating a delivery plan, driving a ground robot to perform targeted delivery, and simultaneously monitoring changes in mosquito density and probiotic activity in real time, dynamically adjusting the delivery plan.

[0013] 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 precise delivery of probiotics to suppress mosquitoes in a landfill based on image recognition, as described in the above embodiment.

[0014] 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 precisely delivering probiotics to a landfill to suppress mosquitoes, as described in the above embodiments.

[0015] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method for precisely delivering probiotics to a landfill to suppress mosquitoes, as described in the above embodiments.

[0016] Therefore, this application has the following beneficial effects: This application's embodiments comprehensively collect landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data through an image acquisition and perception module, providing multi-dimensional foundational data for mosquito analysis and risk assessment. The mosquito feature analysis center extracts core mosquito features using an improved image recognition algorithm and constructs a multi-factor coupled reproductive risk assessment model based on environmental parameters, achieving precise mosquito feature analysis and overcoming the limitations of traditional analysis relying solely on single data points. The reproductive risk assessment module integrates multi-source data to classify risk levels, locate key areas, and generate a breeding risk probability cloud map, improving the accuracy and intuitiveness of risk assessment and enhancing the ability to identify key areas. The probiotic delivery decision unit simulates and selects the optimal probiotic formula and delivery strategy based on the risk cloud map, driving a ground robot for targeted delivery, ensuring accuracy and efficiency. The adaptive control module monitors mosquito density and probiotic activity in real time and dynamically optimizes the delivery plan through a swarm intelligence algorithm, effectively improving the sustainability and stability of mosquito suppression effects in landfills and reducing the environmental impact of mosquito breeding. Thus, it solves the problems of high pollution costs and poor accuracy in existing technologies.

[0017] 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

[0018] 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 a probiotic-based precision delivery system for mosquito control in landfills, provided according to an embodiment of this application. Figure 2 This is a schematic diagram of an image acquisition and sensing module according to an embodiment of this application; Figure 3 This is a schematic diagram of a mosquito feature analysis center according to an embodiment of this application; Figure 4 This is a schematic diagram of a reproductive risk assessment module provided according to an embodiment of this application; Figure 5 This is a schematic diagram of a probiotic dispensing decision unit according to an embodiment of this application; Figure 6 This is a schematic diagram of an adaptive control unit provided according to an embodiment of this application; Figure 7 This is a flowchart of a probiotic precise delivery system for mosquito control in landfills based on image recognition, according to an embodiment of this application. Figure 8This is a flowchart illustrating a method for precisely distributing probiotics based on image recognition to suppress mosquitoes in a landfill, according to an embodiment of this application. Figure 9 This is a schematic diagram of a method for precisely delivering probiotics to a landfill to suppress mosquitoes, based on image recognition, 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

[0019] 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.

[0020] The following description, with reference to the accompanying drawings, illustrates an image recognition-based system for precisely delivering probiotics to landfills to suppress mosquitoes. Addressing the high pollution costs mentioned in the background section, this application provides an image recognition-based system for precisely delivering probiotics to landfills to suppress mosquitoes. In this system, an image acquisition and sensing module comprehensively collects landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data, providing multi-dimensional foundational data for mosquito analysis and risk assessment. A mosquito feature analysis center uses an improved image recognition algorithm to extract core mosquito features and combines them with environmental parameters to construct a multi-factor coupled reproductive risk assessment model, achieving precise mosquito feature analysis and overcoming the limitations of traditional analysis relying solely on single data sources. The reproductive risk assessment module integrates multi-source data to classify risk levels, locate key areas, and generate a breeding risk probability cloud map, improving the accuracy and intuitiveness of risk assessment and enhancing the ability to identify key areas. The probiotic delivery decision unit simulates and selects the optimal probiotic formula and delivery strategy based on the risk cloud map, driving a ground robot for targeted delivery, ensuring accurate and efficient delivery. The adaptive control module monitors mosquito density and probiotic activity in real time and dynamically optimizes the distribution plan through a swarm intelligence algorithm, effectively improving the sustainability and stability of mosquito suppression in landfills and reducing the environmental impact of mosquito breeding. This solves the problems of high pollution costs and poor accuracy in existing technologies.

[0021] Figure 1 This is a schematic diagram of a probiotic precise delivery system for mosquito control in landfills based on image recognition, provided as an embodiment of this application.

[0022] This application provides an image recognition-based system for precise delivery of probiotics to suppress mosquitoes in landfills. The system 10 includes: Image acquisition and sensing module 100, mosquito feature analysis center 200, reproductive risk assessment module 300, probiotic delivery decision unit 400, and adaptive control module 500.

[0023] The system comprises the following components: an image acquisition and perception module 100 for collecting landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data; a mosquito feature analysis center 200 for extracting mosquito species, density, and aggregation area characteristics based on image recognition algorithms, and constructing a mosquito breeding risk assessment model by combining environmental parameters; a breeding risk assessment module 300 for assessing mosquito breeding risk levels and key breeding areas based on the mosquito breeding risk assessment model, integrating topographic data, landfill distribution, and mosquito activity data, and generating a breeding risk probability cloud map; a probiotic delivery decision unit 400 for simulating the inhibitory effects of different probiotic formulations and delivery strategies based on the breeding risk probability cloud map, generating a delivery plan, and driving a ground robot to perform targeted delivery; and an adaptive control module 500 for real-time monitoring of mosquito density changes and probiotic activity, and dynamically adjusting the delivery plan through a swarm intelligence algorithm.

[0024] It is understood that in this embodiment, the image acquisition and perception module comprehensively collects landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data, providing multi-dimensional basic data for mosquito analysis and risk assessment. The mosquito feature analysis center uses an improved image recognition algorithm to extract core mosquito features and combines them with environmental parameters to construct a multi-factor coupled breeding risk assessment model, achieving accurate mosquito feature analysis and breaking through the limitations of traditional analysis that relies on only single data. The breeding risk assessment module integrates multi-source data to classify risk levels, locate key areas, and generate a breeding risk probability cloud map, improving the accuracy and intuitiveness of risk assessment and enhancing the ability to identify key areas. The probiotic delivery decision unit simulates and selects the optimal probiotic formula and delivery strategy based on the risk cloud map, driving the ground robot to deliver the probiotics in a targeted manner, ensuring the accuracy and efficiency of delivery. The adaptive control module monitors mosquito density and probiotic activity in real time and dynamically optimizes the delivery plan through a swarm intelligence algorithm, effectively improving the sustainability and stability of mosquito suppression effects in the landfill and reducing the environmental impact of mosquito breeding. Thus, it solves the problems of high pollution costs and poor accuracy in existing technologies.

[0025] In this embodiment of the application, the image acquisition and sensing module 100 includes: Figure 2 As shown, there are image acquisition unit, environmental parameter acquisition unit, mosquito population characteristic acquisition unit, and garbage decomposition status acquisition unit.

[0026] The image acquisition unit is used to collect panoramic and close-up images of the landfill at different times, covering areas with high mosquito activity; the environmental parameter acquisition unit is used to collect data on temperature, humidity, light intensity, precipitation, and wind speed, and simultaneously record the collection time and location information; the mosquito population characteristic acquisition unit collects data on the morphological characteristics, activity trajectories, and habitat preferences of mosquitoes; and the landfill decomposition status acquisition unit is used to collect data on the humidity, pH, odor level, and decomposition stage of the landfill, and, combined with the appearance of the landfill in the images, comprehensively judge the degree of decomposition.

[0027] It is understood that, through the image acquisition unit, this application's embodiments acquire panoramic and close-up images of the landfill at different times, covering high-frequency areas of mosquito activity, comprehensively capturing mosquito activity scenes, and providing image basis for mosquito species and density identification; the environmental parameter acquisition unit collects data such as temperature and humidity and simultaneously records time and location information, analyzes the impact of the environment on mosquito breeding, and reduces analysis bias caused by environmental factors; the mosquito population characteristic acquisition unit collects data such as mosquito morphological characteristics, accurately extracts key information about mosquito populations, and provides core data for the feature analysis center; the waste decomposition status acquisition unit collects multiple parameters and combines them with the appearance of the waste to judge the degree of decomposition, clarifying the environmental conditions of the waste for mosquito breeding, providing basic support for the formulation of disposal plans, and improving the comprehensiveness and reliability of data collection.

[0028] For example, a waste decomposition status collection unit deployed at a landfill uses distributed humidity sensors, pH probes, and odor sensors embedded at key locations in the waste pile. Simultaneously, high-definition cameras capture images of the waste's appearance, collecting real-time data on internal humidity, pH, concentrations of odor-related gases such as hydrogen sulfide, surface color, and looseness. Through fusion analysis, it is determined that waste in area A is in a semi-decomposed, high-humidity state (easy for mosquito breeding), while waste in area B is fresh (low decomposition). This data is transmitted to the mosquito characteristic analysis center, where a probiotic delivery decision unit matches a high-humidity-resistant compound probiotic formula to area A and increases the dosage, while matching a basic probiotic formula to area B. After targeted delivery by a robot, the mosquito density in area A decreased by 62% within 15 days, effectively verifying the core role of this collection unit in providing crucial environmental data for precise mosquito control.

[0029] In this embodiment of the application, the mosquito feature analysis center 200 includes: such as Figure 3 As shown, the image recognition unit, feature extraction unit, and model building unit are included.

[0030] The image recognition unit uses the GhostYOLO algorithm to accurately identify mosquito species and distinguish between adults and larvae; the feature extraction unit extracts core features such as mosquito density, spatial distribution of aggregation areas, and peak activity periods; the model building unit associates the extracted mosquito features with environmental parameters to construct a multi-factor coupled mosquito breeding risk assessment model and outputs a quantitative value of breeding potential.

[0031] It is understood that the embodiments of this application use the GhostYOLO algorithm in the image recognition unit to accurately identify mosquito species and distinguish between adults and larvae, and to classify mosquito populations in a refined manner, providing species basis for targeted suppression; the feature extraction unit extracts core features such as mosquito density, spatial distribution of aggregation areas, and peak activity periods, comprehensively capturing key patterns of mosquito activity and reducing analytical bias caused by feature omissions; the model building unit deeply correlates mosquito features with environmental parameters, constructs a multi-factor coupled mosquito breeding risk assessment model, and outputs a quantitative value of breeding potential, breaking through the limitations of traditional qualitative analysis, accurately quantifying breeding risks, and improving the scientificity and accuracy of mosquito breeding risk assessment.

[0032] It should be noted that the GhostYOLO algorithm is sensitive to the input feature map. Perform a small number of standard convolutions (with a kernel size of 10 ... ),generate Each intrinsic feature map .

[0033] ; in, The intrinsic feature map generated by standard convolution; X is the input feature map; * is the convolution operator; The kernel corresponding to the standard convolution; m is the intrinsic feature map. The number of feature maps included; n is the number of original output channels (m is usually much smaller than n).

[0034] For each intrinsic feature map Applying linear operations (such as depthwise convolution) to generate An additional phantom feature map.

[0035] ; in, The j-th phantom feature map is generated for the i-th intrinsic feature map; This is the j-th cheap linear operation applied to the i-th intrinsic feature map; For intrinsic feature maps The i-th feature map in the matrix; j is the index of the phantom feature map; s is the total number of feature maps corresponding to each intrinsic feature map (including itself).

[0036] The intrinsic feature map is concatenated with all phantom feature maps to obtain the final result. Each output feature map.

[0037] ; Where Y is the final output feature map obtained by concatenating the intrinsic feature map and the phantom feature map; Concat is the feature map concatenation operation; For intrinsic feature maps The first feature map in; The first phantom feature map generated for the first intrinsic feature map; The first intrinsic feature map generated by the first intrinsic feature map A phantom feature map; For intrinsic feature maps The second feature map in; The generation of the m-th intrinsic feature map There are 1 phantom feature map; n is the final output channel number (satisfying n=m×s).

[0038] The core characteristics of mosquito density, spatial distribution of aggregation areas, and peak activity periods were extracted using a time-series data analysis model. These core characteristics include: 1) mosquito density characteristics, specifically the number of individual mosquitoes per unit monitoring area (e.g., per square kilometer, coverage area per monitoring point) and density gradient changes (e.g., density differences between different areas, density fluctuation trends over time); 2) spatial distribution characteristics of aggregation areas, encompassing the geographical coordinates of concentrated mosquito activity, the shape and size of aggregation patches, and the spatial relationship between aggregation areas and environmental factors (e.g., water sources, vegetation, buildings) (e.g., whether they are distributed in a strip along water sources, whether they form high-aggregation cores in densely vegetated areas); and 3) peak activity period characteristics, which, through analysis of time-series monitoring data, determine the specific time periods (e.g., 5-7 am, 6-8 pm) during which mosquito activity is highest within a day (or a specific period), and the duration of these peak periods. The curves showing the changes in mosquito activity intensity (such as the rate of increase and decrease in mosquito activity intensity during peak periods) provide a comprehensive and three-dimensional capture of mosquito activity patterns. On the one hand, this avoids the one-sidedness of information caused by single-dimensional feature extraction (such as focusing only on density while ignoring spatial distribution or temporal rhythms), ensuring that all key patterns related to mosquito control decisions (such as "a high-gathering area within 100 meters of a certain water source, with peak activity between 6-8 pm") are included in the analysis. On the other hand, through the comprehensiveness and accuracy of feature extraction, it reduces subsequent analysis biases caused by the omission of key information (such as failure to identify the spatial boundaries of gathering areas or misjudging peak activity periods). For example, if the spatial distribution characteristics of gathering areas are missing, the allocation of control resources may deviate from high-risk areas; if the peak activity period is misjudged, the timing of mosquito control operations may be mismatched with the peak mosquito activity, thus affecting the control effect. The extracted mosquito characteristics are correlated with environmental parameters to construct a multi-factor coupled mosquito reproductive risk assessment model, outputting a quantitative value of reproductive potential. First, the core mosquito characteristics obtained by the feature extraction unit (including key information reflecting mosquito population status such as mosquito density, spatial distribution of aggregation areas, and peak activity periods) are systematically correlated with multi-dimensional environmental parameters affecting mosquito reproduction (such as temperature, humidity, precipitation, water source distribution, vegetation coverage, and altitude, which directly or indirectly regulate mosquito oviposition, larval development, and adult survival). This is achieved by integrating the interaction mechanisms between biological characteristics and environmental factors (e.g., high temperature and humidity accelerate mosquito larval development). (The development of mosquitoes, and the combination of high mosquito density and concentrated water sources significantly increases the probability of reproduction). A multi-factor coupled mosquito reproduction risk assessment model is constructed. By quantifying the influence weight of each related factor on mosquito reproduction (such as the influence weight of water source distribution being higher than that of vegetation coverage), the risk of mosquito reproduction in complex environments is comprehensively considered. The model outputs quantitative values ​​(such as specific numerical values, risk level quantitative scores, etc.) that can accurately reflect the mosquito reproduction potential of a specific area. It can intuitively present the intensity and probability of future mosquito population reproduction, and provide accurate decision-making basis for subsequent targeted mosquito control strategies (such as prioritizing water source cleanup and egg eradication operations in areas with high quantitative values).

[0039] In this embodiment of the application, the reproductive risk assessment module 300 includes: Figure 4 As shown, there are risk level assessment units, key area positioning units, and cloud map generation units.

[0040] The risk level assessment unit integrates topographic data, garbage distribution data, and mosquito activity data, and uses the analytic hierarchy process (AHP) to determine the weight of each factor, classifying mosquito breeding risk into four levels: low, medium, high, and extremely high. The key area location unit uses risk value overlay analysis to determine key breeding areas and marks area boundaries and core locations. The cloud map generation unit uses a kernel density estimation algorithm to map risk probability values ​​onto a three-dimensional geographic information model of the landfill, and uses color gradients to represent the risk intensity distribution, generating a breeding risk probability cloud map that includes risk level, key area coordinates, and risk diffusion trends.

[0041] It is understood that the embodiments of this application integrate topographic data, garbage distribution data, and mosquito activity data through a risk level assessment unit, and use the analytic hierarchy process (AHP) to classify four levels of breeding risk. The key area positioning unit accurately determines key breeding areas and core locations, and the cloud map generation unit generates a three-dimensional visualized risk probability cloud map using a kernel density estimation algorithm. This achieves accurate classification of mosquito breeding risk, accurate positioning of key areas, and intuitive presentation of risk distribution, effectively improving the scientific nature and visualization of risk assessment. It provides strong support for subsequent targeted development of prevention and control strategies, optimization of prevention and control resource allocation, and improvement of mosquito control efficiency and accuracy.

[0042] It should be noted that the formula for the Analytic Hierarchy Process (AHP) is as follows: ; ; ; ; Where A is the constructed judgment matrix; To determine the importance scale of the i-th factor relative to the j-th factor in the matrix; n is the number of factors involved in the weighting analysis; is the largest eigenvalue of the judgment matrix; W is the eigenvector of the largest eigenvalue of the judgment matrix, which, after normalization, becomes the weight of each factor; CI is the consistency index; CR is the random consistency ratio; RI is the average random consistency index.

[0043] Risk value superposition analysis formula:

[0044] in, Total risk value; is the index of the factor; n is the number of factors participating in the risk overlay analysis; Let be the weight of the i-th factor; Let be the standardized risk value of the i-th factor.

[0045] Kernel density estimation algorithm formula: ; in, Let x be the risk density at the location to be calculated; n be the number of risk samples; h be the bandwidth; d be the spatial dimension; and i be the sample number. Here, x is the kernel function; x is the spatial location where the risk density needs to be calculated. Let be the spatial coordinates of the i-th risk sample.

[0046] The risk level assessment unit first integrates three types of related data: First, topographic data, covering the terrain undulations, distribution of low-lying waterlogged areas, and the presence of easily water-storing pits within the monitoring area (these topographic conditions directly affect the formation of the waterlogged environment required for mosquito breeding); second, waste distribution data, including the specific location, scale, and type of waste (especially kitchen waste or loose waste that easily retains moisture and breeds microorganisms); and third, mosquito activity data, namely the core feature data such as mosquito density and activity frequency obtained by the feature extraction unit previously. The analytic hierarchy process (AHP) is used to determine the factor weights corresponding to each data point: "Mosquito breeding risk" is first set as the target layer, and topography, waste distribution, and mosquito activity are set as criteria layer factors. The importance of each criteria layer factor is compared pairwise (e.g., judging the importance of waste...). The influence of distribution on mosquito breeding is compared to the strength of topography. A corresponding judgment matrix is ​​constructed, and the initial weights of each factor are obtained by calculating the maximum eigenvalue and eigenvector of the matrix. At the same time, the rationality of the weight allocation is verified by consistency test. Finally, the contribution weight of each type of data to the risk of mosquito breeding is determined. Then, the actual indicators of each type of data are combined with the corresponding weights to calculate the comprehensive risk score of the region. Based on the preset score interval threshold, the risk of mosquito breeding is accurately divided into four levels: low, medium, high, and very high. Low risk corresponds to areas with dry terrain without water accumulation, little garbage, and sparse mosquito activity, while very high risk corresponds to low-lying areas with water accumulation, large accumulation of garbage, and extremely high mosquito density. This achieves a refined risk classification based on multi-dimensional data and reasonable weights, avoiding risk misjudgment caused by single-factor assessment.

[0047] For example, taking a large urban waste treatment plant and its surrounding 3-kilometer radius as the monitoring target, the key area positioning unit first obtains the weighted risk values ​​of three factors output by the risk level assessment unit: topography (e.g., the low-lying area on the east side prone to water accumulation), waste distribution (e.g., the concentrated food waste dumping point in the southwest corner), and mosquito activity (e.g., the high mosquito density area on the northern edge). Through risk value superposition analysis, it is found that the low-lying area on the east side of the waste treatment plant has a high risk of water accumulation (topography weight 0.35) + large amount of waste accumulation (waste distribution weight 0.4) + high mosquito density. The combined condition of "exceeding the standard by 2 times (mosquito activity weight 0.25)" resulted in a risk value far exceeding the preset threshold, thus accurately identifying the area as a key breeding area for mosquitoes. The area boundaries (113°XX′XX″-113°XX′XX″ E, 23°XX′XX″-23°XX′XX″ N) and core locations (center of low-lying waterlogged areas and within 50 meters of the edge of the kitchen waste dumping area) were marked by spatial coordinate calibration, providing clear targets for subsequent targeted waterlogging cleanup, pesticide spraying and other precise control operations.

[0048] In this embodiment of the application, the probiotic dispensing decision unit 400 includes, as follows: Figure 5 As shown, the module includes a formula screening module, an inhibition effect simulation unit, an administration plan generation unit, and a targeted administration drive unit.

[0049] The system comprises several modules: a formula selection module, which selects suitable probiotic strain combinations based on mosquito species and the state of waste decomposition; a suppression effect simulation unit, which simulates the mosquito suppression mechanism of different probiotic formulas and combines a breeding risk probability cloud map with a multi-objective optimization algorithm to simulate the decrease in mosquito density under different formulas, dosages, timings, and frequencies, and evaluates the suppression effect and duration; a distribution plan generation unit, which determines the specific probiotic formula, single dosage, frequency, precise distribution area, and optimal distribution time based on the simulation results, forming a standardized distribution plan; and a targeted distribution drive unit, which plans the distribution path of the ground robot, generates a target movement path based on the coordinates of key breeding areas and the terrain and obstacle information of the landfill, drives the robot to the distribution area, and completes the targeted distribution operation.

[0050] It is understood that the embodiments of this application use a formula screening module to accurately select probiotic strain combinations that are adapted to the mosquito species and the state of garbage decomposition. The inhibition effect simulation unit uses a multi-objective optimization algorithm to scientifically evaluate the mosquito inhibition effect under different placement conditions. The placement scheme generation unit formulates standardized and refined placement schemes. The targeted placement drive unit plans the target path of the robot and completes the targeted placement. This not only ensures the targeting and effectiveness of the placement and improves the mosquito inhibition efficiency, but also reduces labor costs and avoids resource waste through standardized schemes and targeted robot operations. It provides an efficient, accurate and feasible intelligent solution for mosquito control in landfills.

[0051] It should be noted that the formulation screening module identifies dominant mosquito species (such as Culex and Aedes) through mosquito monitoring data from the target landfill (such as adult mosquito morphology identification and larval species typing results). Different mosquito species have different reproductive habits, larval feeding preferences, and intestinal microbial structures (e.g., Aedes larvae rely more on microorganisms in stagnant water environments for food, while Culex larvae are more sensitive to specific antibacterial substances), and their sensitivity to probiotic strains also varies. At the same time, by detecting indicators such as the humidity, temperature, degree of organic matter degradation, pH value, and microbial community composition of the waste, the module accurately assesses the state of waste composting (uncomposted, semi-composted, and fully composted). Waste at different stages of composting has different nutrient supply, acid-base environment, and intensity of competition with native microorganisms (e.g., semi-composted waste is rich in organic matter but has a high acidity, requiring the screening of acid-resistant strains that can efficiently utilize nutrients at this stage). Based on this, the module calls a pre-set strain matching database (containing the mechanism of action, host specificity, and environmental adaptability parameters of different probiotic strains). Through matching analysis, it selects a combination of probiotic strains (rather than a single strain) that can specifically inhibit target mosquito species (such as by secreting antibacterial substances to interfere with mosquito larval development and compete for the nutrients needed by mosquitoes) and can stably colonize, reproduce, and exert their effects in the current state of garbage decomposition. This avoids inhibition failure due to strain mismatch with mosquito species, or poor survival ability and reduced effectiveness due to strains' inability to adapt to the garbage environment. This provides core strain support for subsequent mosquito suppression effects, ensuring the targeted and basic effectiveness of probiotic delivery.

[0052] Multi-objective optimization algorithm formula: ; ; ; in, The first objective function is the magnitude of the decrease in mosquito density. The second objective function is the duration of mosquito suppression. A vector of decision variables; Decision variables The corresponding decrease in mosquito density; Decision variables The corresponding duration of mosquito suppression; It is a probiotic formula type; This is a single-dose administration. For delivery time; For delivery frequency; This represents the lower limit of the selectable range of probiotic formulation types. This represents the upper limit of the selectable range for probiotic formulation types; This represents the lower limit of the safe and effective range for a single dose. This represents the upper limit of the safe and effective range for a single dose. This represents the lower limit of a reasonable range for the delivery time. This represents the upper limit of a reasonable range for the delivery time. This represents the lower limit of the feasible range for the frequency of deployment; This represents the upper limit of the feasible range for the frequency of deployment; Decision variables Corresponding environmental adaptability indicators for probiotics; This represents the minimum threshold for the environmental adaptability of probiotics.

[0053] The targeted delivery unit, based on the precise coordinates of key mosquito breeding areas and taking into full account the actual terrain of the landfill (such as slope, ground hardness, and distribution of low-lying areas) and information on various obstacles (such as temporary garbage, operating equipment, and protective facilities), plans the optimal movement path that avoids unfavorable terrain and obstacles while ensuring both efficiency and safety. During path planning, it automatically avoids steep slopes, soft, waterlogged areas, and various obstructions to prevent the robot from tipping over, getting stuck, or colliding. The planned path is then converted into specific movement guidelines and sent to the ground robot in real time. Based on the robot's real-time location feedback, the movement route is flexibly adjusted to adapt to possible temporary changes on-site, ultimately guiding the robot precisely to the core delivery location in the key breeding areas to complete the quantitative delivery of probiotics. This not only avoids the safety hazards of manual entry into the high-pollution, high-risk areas of the landfill but also achieves targeted delivery of probiotics through precise path guidance, reducing resource waste caused by delivery location deviations and improving the overall efficiency and stability of the delivery operation.

[0054] For example, using a semi-decomposed kitchen waste dump in the southern suburbs of a city (dominant mosquito species is Culex pipiens, and its southwestern area is a high-risk breeding zone for mosquitoes) as the monitoring target, the suppression effect simulation unit first combined the distribution of breeding risks in the area with the action logic of different probiotic combinations (for example, one combination competes for the nutrient source of mosquito larvae, while another combination secretes substances that inhibit larval development), and tested the effects of different application conditions one by one: For example, when simulating "probiotic combination A + 50 grams per square meter + early morning application + once a week", the results showed that the mosquito density decreased by only 40%, and the suppression effect lasted only 3 days; when simulating "probiotic combination B + 60 grams per square meter + evening application (before the peak of mosquito activity) + twice a week", the results showed that the mosquito density in the area could decrease by 85%, and the suppression effect could be stably maintained for 12 days, without causing additional environmental burden due to excessive dosage. Finally, this unit selected this optimal combination of conditions, providing a precise basis for the formulation of subsequent application plans and avoiding insufficient effects or waste of resources caused by blind application.

[0055] In this embodiment of the application, the adaptive control module 500 includes, as follows: Figure 6 As shown, there are real-time monitoring unit and scheme adjustment unit.

[0056] The real-time monitoring unit continuously monitors changes in mosquito density, probiotic activity, and fluctuations in environmental parameters. The scheme adjustment unit analyzes the monitoring data, explores the correlation between changes in mosquito density and probiotic administration, and dynamically adjusts the administration scheme from three dimensions: administration dosage, administration frequency, and probiotic formula, to ensure a continuous and stable inhibitory effect.

[0057] It is understood that the embodiments of this application continuously track changes in mosquito density, probiotic activity, and fluctuations in environmental parameters through a real-time monitoring unit. The scheme adjustment unit deeply analyzes the monitoring data and explores the intrinsic correlation between mosquito density and probiotic administration. It dynamically optimizes the administration scheme from three dimensions: administration dosage, administration frequency, and probiotic formula. This achieves real-time response and precise adaptation to various variables in the mosquito control process, effectively avoiding the decline in the inhibitory effect caused by environmental changes, probiotic activity decay, or fluctuations in mosquito population dynamics. It not only ensures the continuous stability of the mosquito inhibitory effect, but also reduces unnecessary resource consumption through dynamic adjustment, thereby improving the flexibility, adaptability, and long-term effectiveness of the control.

[0058] It should be noted that the scheme adjustment unit first integrates multi-dimensional data collected by the real-time monitoring unit—including real-time trends in mosquito density (such as whether there is a rebound or whether the rate of decline slows down), the activity status of probiotics in the waste environment (such as colonization rate, reproductive capacity, and efficiency of antibacterial substance secretion), and dynamic fluctuations in environmental parameters (such as changes in temperature, humidity, and waste composting process)—to construct a complete data analysis dataset. Subsequently, data mining is used to deeply analyze the inherent correlations between the data: for example, identifying the positive correlation threshold between probiotic dosage and the rate of mosquito density decline (such as the faster rebound of mosquito density when the dosage is below a certain value), the correlation between the rate of probiotic activity decay and environmental humidity (such as the extended activity maintenance time in high humidity environments, allowing for a suitable reduction in dosage frequency), and the adaptability differences of different probiotic formulations under specific environmental conditions (such as the more stable antibacterial effect of a certain formulation in high-temperature environments), etc. Based on the identified correlations, the application plan is dynamically optimized and adjusted from three core dimensions: In terms of dosage, the dosage is adjusted according to the probiotic activity decay and mosquito density rebound trend (e.g., increasing the dosage appropriately when activity decays too quickly and mosquito density rises); in terms of application frequency, the application interval is adjusted based on the mosquito breeding cycle and the duration of the inhibitory effect (e.g., shortening the application cycle during peak mosquito breeding seasons or when the environment is conducive to mosquito breeding); and in terms of probiotic formulation, a more suitable strain combination is switched to address environmental parameter fluctuations (e.g., sudden temperature changes, transitions in waste composting stages) or changes in mosquito population structure (e.g., switching to a cold-resistant strain formulation in low-temperature environments). This data-driven dynamic adjustment model avoids the drawbacks of fixed application plans being unable to adapt to environmental changes and dynamic factors such as probiotic activity decay, ensuring that the mosquito inhibitory effect remains stable and achieving long-term control goals.

[0059] For example, taking mosquito control at a suburban landfill as an example, the program adjustment unit analyzed real-time monitoring data and found that after continuous rainfall, the ambient humidity increased by 15%, and the activity of the originally applied probiotics decreased by 30%, causing the mosquito density to rebound by 20% within 5 days. Furthermore, data showed that the antibacterial efficiency of the original strain combination was significantly reduced in high humidity environments, and when the dosage was less than 60 grams per square meter, the duration of the inhibitory effect shortened to 7 days. Based on this correlation, the unit dynamically adjusted the application plan: the probiotic formula was replaced with a moisture-resistant strain combination, the single application dosage was increased to 70 grams per square meter, and the application frequency was adjusted from once a week to once a week plus one supplementary application after rain. Within one week of the adjustment, the mosquito density returned to the target control range, and the inhibitory effect was maintained stably for 14 days, effectively avoiding control failure caused by environmental changes and the decline in probiotic activity, and ensuring the continuous and stable mosquito control effect.

[0060] This application proposes an image recognition-based probiotic precision delivery system for mosquito control in landfills. The system comprehensively collects landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data through an image acquisition and sensing module, providing multi-dimensional foundational data for mosquito analysis and risk assessment. A mosquito feature analysis center uses an improved image recognition algorithm to extract core mosquito features and combines them with environmental parameters to construct a multi-factor coupled reproductive risk assessment model, achieving precise mosquito feature analysis and overcoming the limitations of traditional analysis relying solely on single data sources. The reproductive risk assessment module integrates multi-source data to classify risk levels, locate key areas, and generate a breeding risk probability cloud map, improving the accuracy and intuitiveness of risk assessment and enhancing the ability to identify key areas. The probiotic delivery decision unit simulates and selects the optimal probiotic formula and delivery strategy based on the risk cloud map, driving a ground robot for targeted delivery, ensuring accuracy and efficiency. An adaptive control module monitors mosquito density and probiotic activity in real time and dynamically optimizes the delivery plan through a swarm intelligence algorithm, effectively improving the sustainability and stability of mosquito control in landfills and reducing the environmental impact of mosquito breeding. This solves the problems of high pollution costs and poor accuracy in existing technologies.

[0061] The following will illustrate a specific embodiment of a probiotic precision delivery system for mosquito control in landfills based on image recognition. Figure 7 As shown, it includes: The landfill in a large urban circular economy industrial park covers an area of ​​500,000 square meters and processes an average of 3,000 tons of domestic waste per day. Because the landfill mainly contains kitchen waste and food waste, and there are many low-lying waterlogged areas on the site, the mosquito density can reach as high as 30 mosquitoes per square meter in summer. Traditional chemical spraying poses a risk of pollution and has a short-term inhibitory effect. Therefore, the park has introduced a mosquito control system based on image recognition for precise delivery of probiotics to the landfill, achieving green and long-term mosquito control.

[0062] The image acquisition and sensing module, serving as the system's core data entry point, comprises a multi-source data acquisition matrix consisting of an image acquisition unit, an environmental parameter acquisition unit, a mosquito population characteristic acquisition unit, and a waste decomposition status acquisition unit, achieving full-scene data coverage. The image acquisition unit utilizes a high-definition infrared network camera (4K resolution, 25fps) and a low-altitude inspection drone (equipped with a 20-megapixel zoom camera) working in tandem: fixed cameras are deployed in 12 high-frequency mosquito activity areas, including the landfill perimeter fence, low-lying waterlogged areas, and waste leveling areas, capturing panoramic images every 30 minutes and automatically switching to infrared mode at night to capture mosquito activity trajectories; the drone conducts three daily inspections along a preset route, focusing on capturing close-up images of moving areas such as waste accumulation slopes and temporary dumping sites, ensuring no monitoring blind spots. The environmental parameter acquisition unit consists of 15 monitoring stations evenly distributed throughout the site. Each station integrates a temperature and humidity sensor (measurement range -40℃~85℃, 0~100%RH, accuracy ±0.5℃, ±2%RH), a light sensor (0~100000 lux, accuracy ±5%), a tipping bucket rain gauge (resolution 0.2mm), and a wind speed sensor (0~60m / s, accuracy ±0.3m / s). Data is collected every 10 minutes, and the collection time and latitude / longitude coordinates are recorded simultaneously. The mosquito population characteristic acquisition unit uses an intelligent trap (with built-in LED mosquito-attracting lamp and high-definition camera), operating from 18:00 to 6:00 the next day. Trapped mosquitoes are automatically released after image capture. Image recognition is used to obtain data on morphological characteristics (wing vein structure, body size), activity trajectory (staying time, flight range), and habitat preferences (such as preference for garbage accumulation surfaces or water edges). The waste composting status acquisition unit combines contact detection and image analysis: a handheld probe-type hygrometer (measurement range 0~100%, accuracy ±1%) and a pH meter (0~14pH, accuracy ±0.01pH) collect waste sample data daily from different accumulation areas; an odor sensor (detection range 0~500ppm, response time ≤2 seconds) monitors volatile gas concentrations in real time; simultaneously, image recognition of waste color (gradient change from black to brown) and bulkiness (change in bulk density) comprehensively determines the composting stage. All collected data is transmitted to edge computing nodes (using Huawei Atlas500 intelligent edge servers) via industrial Ethernet (fixed equipment) and 5G private network (drones, handheld devices). After preprocessing including Gaussian filtering for noise reduction, outlier removal (based on the 3σ criterion), and data format standardization (uniformly converted to JSON format), the data is uploaded to the Alibaba Cloud OSS database. The data transmission latency is ≤500ms, ensuring high-quality data support for subsequent analysis.

[0063] The mosquito feature analysis center, through the collaborative computation of the image recognition unit, feature extraction unit, and model building unit, completes mosquito feature analysis and risk model construction. The image recognition unit employs an improved GhostYOLO algorithm, optimizing the network structure to address the small size and varied flight postures of mosquitoes: a lightweight Ghost module is added to the backbone network, reducing feature map generation costs by 60%; an attention mechanism is introduced in the detection head section to strengthen the weights of key identification features such as wing veins and antennae. After training with 80,000 sets of labeled data (including six common landfill mosquitoes such as Culex, Aedes, and Anopheles, covering different morphologies of adults and larvae), the model achieves a mosquito species identification accuracy of 94%, an adult-to-larvae differentiation accuracy of 97%, and a single-frame image recognition speed of ≤0.3 seconds. The feature extraction unit performs multi-dimensional feature quantification based on the recognition results: mosquito density is calculated by counting target detection boxes and converting the area (e.g., if 200 mosquitoes are detected in a 100-square-meter area, the density is 2 mosquitoes / square meter); the aggregation area uses the DBSCAN clustering algorithm to divide areas with a distance of less than 5 meters and a density higher than 1.5 mosquitoes / square meter into aggregation areas, outputting the area boundary coordinates and core density values; the peak activity period is determined by statistically analyzing the number of mosquitoes detected in each time period within 24 hours, extracting the three most frequent periods as peak periods (e.g., 19:00-21:00, 4:00-6:00, and 12:00-14:00 for this landfill). The model construction unit uses the random forest algorithm to construct a multi-factor coupled reproductive risk assessment model, using the extracted mosquito features (density, aggregation degree, activity peak) and environmental parameters (temperature, humidity, precipitation) as input feature variables, and the mosquito larvae hatching rate from the field survey as the output label (quantified reproductive potential value). The model underwent 5-fold cross-validation, with a prediction error of ≤8%. When the output quantification value is ≥0.7, it is considered to have high reproductive potential, and when it is ≤0.3, it is considered to have low reproductive potential, providing a quantitative basis for subsequent risk assessment.

[0064] The mosquito breeding risk assessment module integrates model output with multi-source data, achieving precise risk quantification and visualization through risk level assessment units, key area positioning units, and cloud map generation units. The risk level assessment unit first integrates three types of core data: topographic data (obtained from the park's GIS system, including altitude, slope, and distribution of low-lying areas), garbage distribution data (generated by fusing GPS positioning of garbage trucks and drone inspection images, including the amount and duration of garbage accumulation in each area), and mosquito activity data (from the feature analysis center). The analytic hierarchy process (AHP) is used to determine the weights of each factor: with "mosquito breeding risk" as the target layer, topography (weight 0.3), garbage distribution (weight 0.4), and mosquito activity (weight 0.3) as the criterion layers, a 9×9 judgment matrix is ​​constructed. The weights are confirmed to be reasonable after a consistency test (CR=0.06<0.1). Subsequently, the data of each factor were standardized (normalized to the [0,1] interval using min-max normalization), and the comprehensive risk value was calculated by weighted summation. Risk levels were then classified according to the risk value: ≤0.3 for low risk, 0.3~0.5 for medium risk, 0.5~0.7 for high risk, and ≥0.7 for extremely high risk. For key area positioning units, risk value overlay analysis was used. The comprehensive risk value of each grid (10m×10m) was overlaid with the risk values ​​of the surrounding three grids. Grids with a risk value ≥0.8 after overlay were identified as key breeding areas. The area boundaries (e.g., 114°25′30″-114°25′45″ E, 23°12′10″-23°12′25″ N) and core points (deepest water accumulation point, highest garbage accumulation point) were marked using coordinate calibration. The cloud map generation unit uses a kernel density estimation algorithm with a bandwidth of 5 meters and a spatial dimension of 2. It maps the risk probability values ​​of each key area to a 3D geographic information model of the landfill built based on the OSG engine. It uses color gradients to represent the risk intensity (blue for low risk, green for medium risk, orange for high risk, and red for extremely high risk) and generates a probabilistic cloud map of breeding risk that includes risk level labels, key area coordinate pop-up windows, and risk diffusion trend arrows (based on wind speed and direction prediction). The update frequency is once a day.

[0065] Based on risk assessment results, the probiotic delivery decision-making unit achieves precise delivery through formula screening, effect simulation, scheme generation, and targeted drive. The formula screening module calls upon a pre-set strain database (containing 8 mosquito-inhibiting strains such as Bacillus subtilis and Bacillus licheniformis) and matches suitable formulas with monitoring data: targeting the dominant mosquito species Culex pipiens in the landfill and the semi-composted state of the waste (humidity 65%, pH 6.8), a strain combination of "Bacillus subtilis + Bacillus licheniformis = 3:2" was selected. This combination achieved an 85% colonization rate in the semi-composted waste environment and an inhibition rate of ≥90% against Culex pipiens larvae. The inhibition effect simulation unit employed a non-dominated sorting genetic algorithm (NSGA-Ⅲ) to construct a multi-objective optimization model. The objectives were to maximize the reduction in mosquito density, maximize the duration of inhibition, and minimize the deployment cost. Decision variables included formulation ratio, deployment dosage (50–100 g / m²), deployment time (6:00, 12:00, 18:00), and deployment frequency (1–3 times / week). Constraints included a strain colonization rate ≥70% and an environmental pH of 4–9. After 100 generations of iterative simulation, the Pareto optimal solution was obtained: deployment dosage 70 g / m², deployment time 18:00 (before the peak activity of Culex pipiens), and deployment frequency 1 time / week, corresponding to an 88% reduction in mosquito density, a duration of inhibition of 14 days, and a unit cost of 0.8 yuan / m². The disposal plan generation unit transforms the simulation results into a standardized plan, specifying the disposal range (within 20 meters of the core location), disposal volume (calculated by area, e.g., 70 kg for a 1000 square meter area), disposal time (every Wednesday at 18:00), and strain ratio (42 kg of Bacillus subtilis and 28 kg of Bacillus licheniformis) for each key area. The targeted disposal drive unit uses the A* path planning algorithm, inputting the coordinates of the core location in the key area, landfill topography data (slopes > 15° are impassable), and real-time obstacle information (such as the location of operating vehicles and temporary garbage piles), to plan the optimal path and avoid obstacles such as steep slopes and waterlogged areas. This drives three AGV ground robots (50 kg load, positioning accuracy ±3 cm) to travel along the path. Upon reaching the disposal point, they complete the quantitative disposal using a rotary feeding device (feeding accuracy ±50 g). The entire disposal process is uploaded to a cloud monitoring platform in real time, requiring no manual intervention.

[0066] The adaptive regulation module ensures the continuous and stable inhibition effect through real-time monitoring and dynamic adjustment. The real-time monitoring unit sets up 8 monitoring points in the key distribution areas: the mosquito density is monitored by the mosquito attractant lamp counting method (counted once every 24 hours, with an accuracy of ±1 mosquito / m²); the probiotic activity is detected once a week through a microbial detection test strip (containing a specific culture medium) (the activity is qualified when the colony count ≥ 10^6 CFU / g); the environmental parameter monitoring and image acquisition and perception module share data, and the changes in temperature, humidity, and the progress of garbage decomposition are tracked. The scheme adjustment unit uses the ant colony algorithm to conduct correlation analysis on the monitoring data and discovers the key rules: when the environmental humidity > 75%, the decay rate of probiotic activity accelerates by 30%; the rebound rate of mosquito density is negatively correlated with the dosage, and when the dosage < 60 g / m², the rebound cycle is shortened to 7 days; when the garbage decomposition enters the fully decomposed stage (pH value 7.5, humidity 55%), the inhibition effect of the original formula drops by 20%. During a certain monitoring, it was found that after 3 consecutive days of rainfall, the humidity rose to 80%, the probiotic activity dropped to 60%, and the mosquito density rebounded from 0.5 mosquito / m² to 1.2 mosquitoes / m². Based on this rule, the unit immediately adjusted the scheme dynamically: the formula was changed to a high-humidity-resistant combination of "Bacillus subtilis + Bacillus amyloliquefaciens = 2:3", the single dosage was increased to 80 g / m², and the dosing frequency was adjusted to once a week + a supplementary dose within 48 hours after rain. The monitoring on the 5th day after the adjustment showed that the mosquito density dropped to 0.4 mosquito / m² and the probiotic activity recovered to 82%; on the 14th day, the density still remained at 0.6 mosquito / m², and the stability of the effect was improved by 40% compared with the original fixed scheme, achieving the long-term mosquito control goal.

[0067] In summary, in the embodiment of the present application, the image acquisition and perception module collects multi-dimensional accurate data, providing high-quality input for subsequent analysis. The mosquito feature analysis center accurately analyzes mosquito features and constructs a risk assessment model with an improved algorithm, laying a scientific foundation for prevention and control decisions; the breeding risk assessment module realizes the classification of mosquito breeding risks, the accurate positioning and visual presentation of key areas, ensuring the targeting of prevention and control; the probiotic dosing decision unit avoids the pollution risk of traditional chemical agents and improves the mosquito inhibition efficiency through the selection of suitable formulations, effect simulation, and robot-targeted dosing; the adaptive regulation module dynamically adjusts the dosing scheme based on real-time monitoring, maintaining the continuous and stable inhibition effect and reducing resource waste; overall, it avoids the subjectivity and inefficiency of manual prevention and control, realizes the green, accurate, and long-term prevention and control of mosquitoes in the landfill, reduces the comprehensive prevention and control cost, and improves the ecological environment quality of the park.

[0068] Secondly, a method for inhibiting mosquitoes in a landfill by accurately dosing probiotics based on image recognition proposed according to an embodiment of the present application is described with reference to the accompanying drawings.

[0069] As Figure 8 shown, the method for inhibiting mosquitoes in a landfill by accurately dosing probiotics based on image recognition includes the following steps: In step S101, the landfill image data, environmental parameters, mosquito population characteristic data, and landfill decomposition status data are acquired.

[0070] It is understood that the embodiments of this application, by acquiring landfill image data, environmental parameters, mosquito population characteristic data, and landfill decomposition status data, enable mosquito characteristic analysis, risk assessment, and deployment decisions to accurately correlate mosquito population characteristics, environmental impact, and landfill decomposition degree. This allows for real-time capture of dominant mosquito species types, density, and activity patterns, the impact of environmental factors on mosquito reproduction, and the suitability of the landfill decomposition stage for probiotic colonization. This provides high-quality data for the subsequent mosquito characteristic analysis center to build a precise risk assessment model, the reproduction risk assessment module to achieve risk classification and key area positioning, and the probiotic deployment decision-making unit to formulate suitable deployment plans. This precise perception and targeted focus on key factors influencing mosquito reproduction avoids the blindness of traditional control methods, improves the targeting and effectiveness of mosquito suppression, and reduces resource waste and environmental risks.

[0071] In step S102, based on landfill image data and mosquito population characteristic data, the GhostYOLO algorithm is used to extract mosquito species, density, and aggregation area characteristics. Combined with environmental parameters and landfill decomposition status data, a mosquito breeding risk assessment model is constructed.

[0072] Among them, the mosquito breeding risk assessment model is a quantitative analysis model that integrates multi-dimensional data such as mosquito population characteristics, environmental parameters, topography and landforms and garbage distribution, and is constructed through correlation analysis to assess the breeding potential of mosquitoes and classify the breeding risk level.

[0073] It is understood that the embodiments of this application utilize a mosquito breeding risk assessment model, integrating population characteristics such as mosquito species, density, and aggregation areas extracted by the GhostYOLO algorithm, as well as multi-dimensional data such as environmental parameters, garbage decomposition status, topography, and garbage distribution for quantitative analysis. This accurately assesses the mosquito breeding potential and classifies the breeding risk level, avoiding the blindness of relying solely on experience in traditional mosquito control. It provides scientific data support for identifying key breeding areas and developing appropriate probiotic delivery plans, improving the targeting and effectiveness of mosquito suppression, and reducing the waste of control resources and environmental risks.

[0074] It should be noted that the formula for the mosquito breeding risk assessment model is as follows: ; ; ; ; Where R is the mosquito breeding risk index; The weights of mosquito characteristic factors; These are characteristic factors of mosquitoes; The weights of environmental adaptation factors; Environmental adaptation factors; The weights of the waste decomposition state factors; The factor representing the state of waste decomposition; For scene correction items; Dn is the species risk coefficient; Dn is the density normalized value; A is the clustering index; For temperature adaptability; For humidity adaptation; For rainfall adaptation; M represents wind speed adaptability; M represents the degree of decay index. The percentage of organic waste; This is the normalized value for the stacking height.

[0075] For example, a 200㎡ suburban mixed landfill mainly consists of kitchen waste (75%), with a pile height of 1.8m. On a summer afternoon, staff used the GhostYOLO algorithm to analyze infrared monitoring images of the site: identifying Aedes mosquitoes (60%) and Culex mosquitoes (40%), with a density of 18 mosquitoes / ㎡ in a single area, and the clustering area occupying 30% of the site. Based on this, the mosquito characteristic factor Fm=0.81 was calculated. Combining this with real-time environmental data (average daily temperature 29℃, humidity 82%, daily rainfall 45mm, wind speed 1.2m / s), the environmental suitability factor Fe=0.92 was calculated. Through sampling and testing of the landfill, with a decomposition degree of 0.8, an organic waste ratio of 75%, and a normalized value for pile height of 0.9, the waste decomposition state factor Fᵥ=0.83 was calculated. With weights of α=0.45, β=0.3, and γ=0.25, and a summer scenario correction term of ε=0.04, the final risk index R≈0.82 was calculated, which was determined to be "extremely high risk". Two rounds of efficient disinfection were immediately initiated on-site, the pile was cleared on the same day, ventilation fans were deployed for dehumidification, and the monitoring frequency was increased to once every 2 hours.

[0076] In step S103, based on the mosquito breeding risk assessment model, the topography, garbage distribution and mosquito activity data are integrated to assess the mosquito breeding risk level and key breeding areas, and generate a breeding risk probability cloud map.

[0077] Among them, the breeding risk probability cloud map refers to a two-dimensional image that visualizes the distribution of mosquito breeding risk probability in different areas of a landfill using color or grayscale gradients, based on mosquito population characteristics, environmental parameters, and data on the state of garbage decomposition. This is achieved through a mosquito breeding risk assessment model combined with spatial interpolation technology.

[0078] It is understood that the embodiments of this application intuitively present the risk level and probability distribution of mosquito breeding in different areas of the landfill through visual color or grayscale gradient, accurately pinpointing key breeding areas formed by the superposition of topography, garbage distribution and mosquito activity. This not only avoids the blindness of traditional prevention and control, but also provides data support for the optimized allocation of prevention and control resources. It helps staff to focus on high-risk areas to carry out targeted disinfection, precise garbage removal and environmental improvement, greatly improving the efficiency and effectiveness of prevention and control, reducing manpower and material costs, and providing a scientific basis for subsequent dynamic monitoring of risk changes, prediction of breeding trends and formulation of long-term prevention and control plans.

[0079] In step S104, based on the probability cloud map of breeding risk, the inhibitory effect of different probiotic formulations and delivery strategies is simulated to generate a delivery plan, which drives the ground robot to carry out targeted delivery. At the same time, the mosquito density changes and probiotic activity are monitored in real time, and the delivery plan is dynamically adjusted.

[0080] It is understood that the embodiments of this application use the probability cloud map of breeding risk as the basis for precise positioning. By simulating the inhibitory effects of different probiotic formulas and delivery strategies, a scientifically adapted targeted delivery plan is generated, driving the ground robot to focus on high-risk areas for precise delivery. This avoids the waste and blind delivery of probiotics and maximizes the inhibitory effect of probiotics on mosquito breeding. At the same time, by monitoring changes in mosquito density and probiotic activity in real time and dynamically adjusting delivery parameters, not only is the mosquito suppression efficiency significantly improved and the control cost reduced, but the environmental impact of chemical disinfection is also reduced, promoting the transformation of mosquito control in landfills towards intelligence, precision, greenness, and long-term effectiveness.

[0081] This application proposes a method for precise probiotic delivery to suppress mosquitoes in landfills based on image recognition. The method comprehensively collects landfill image data, environmental parameters, mosquito population characteristics, and landfill decomposition status data through an image acquisition and sensing module, providing multi-dimensional foundational data for mosquito analysis and risk assessment. A mosquito feature analysis center uses an improved image recognition algorithm to extract core mosquito features and combines them with environmental parameters to construct a multi-factor coupled reproductive risk assessment model, achieving precise mosquito feature analysis and overcoming the limitations of traditional analysis relying solely on single data sources. The reproductive risk assessment module integrates multi-source data to classify risk levels, locate key areas, and generate a breeding risk probability cloud map, improving the accuracy and intuitiveness of risk assessment and enhancing the ability to identify key areas. A probiotic delivery decision-making unit simulates and selects the optimal probiotic formula and delivery strategy based on the risk cloud map, driving a ground robot for targeted delivery to ensure accuracy and efficiency. An adaptive control module monitors mosquito density and probiotic activity in real time and dynamically optimizes the delivery plan through a swarm intelligence algorithm, effectively improving the sustainability and stability of mosquito suppression in landfills and reducing the environmental impact of mosquito breeding. This solves the problems of high pollution costs and poor accuracy in existing technologies.

[0082] The following specific embodiment will illustrate a method for precise delivery of probiotics to suppress mosquitoes in landfills based on image recognition. Figure 9 As shown, it includes: A medium-sized suburban municipal solid waste treatment plant (covering 5000㎡, with a daily processing capacity of 80 tons, divided into five functional areas: unloading, sorting, composting, temporary storage, and collection; primarily consisting of 45% kitchen waste, 20% recyclables, and 35% other waste) was selected as the implementation site to construct a comprehensive mosquito control system. The primary task was to complete the full-scale collection and deployment of multi-source data. Image data acquisition for the waste site utilized a Hikvision DS-2TD2636-6 / V1 infrared thermal imager, with one unit deployed at the highest point in each of the five functional areas. The imager has a resolution of 640×512, a frame rate of 25fps, and a temperature range of -20°C to 150°C, capable of penetrating nighttime environments to capture the thermal radiation characteristics of mosquito aggregation. Simultaneously, a Hikvision DS-2CD3T46DWD-I5 high-definition camera (1080P resolution, 15fps) was used to capture mosquito morphology during the day for species identification. Environmental parameter monitoring utilizes a Huace HC-600 automatic weather station located at the center of the site. Monitoring parameters include air temperature (-40-85℃, accuracy ±0.1℃), relative humidity (0-100%RH, accuracy ±2%RH), rainfall (0-200mm / h, accuracy ±0.2mm), wind speed (0-60m / s, accuracy ±0.1m / s), and wind direction (0-360°, accuracy ±1°), with a sampling frequency of 1Hz. An additional set of temperature and humidity sensors (SHT30) is deployed in each functional area to focus on local microenvironmental differences. Mosquito population characteristic data employs a dual acquisition mode of "trapping + image recognition": two Maxwell MW-001 mosquito-attracting lamps with a wavelength of 365nm and a trapping radius of 10m are deployed at the boundary of each functional area, equipped with detachable mosquito collection boxes. Mosquito samples are collected daily in the early morning for species identification. A miniature camera is installed next to the mosquito-attracting lamps to capture real-time images of the number of mosquitoes in the collection boxes, assisting in density statistics. Data on the composting status of waste was collected daily using an Aokoma AKM-03 soil sampler. Three representative sampling points (0-30cm from the surface layer of the waste pile) were selected from each functional area. Samples were sent to the on-site laboratory: the organic matter degradation rate was measured using a Shimadzu UV-1800 ultraviolet spectrophotometer (accuracy ±0.5%), the moisture content was measured using a Mettler HL200 moisture analyzer (accuracy ±0.1%), and the carbon-nitrogen ratio was measured using an elemental analyzer (accuracy ±0.1%). Sampling was conducted once daily. All data were aggregated via a Huawei AR502H industrial Ethernet gateway and transmitted to a local server via a 5G module (latency ≤100ms). InfluxDB time-series database was used to store infrared images and high-frequency environmental data (retained for 30 days), while MySQL was used to store mosquito population statistics and waste composting data. A complete traceability system was constructed by associating data tags with functional areas and collection times.

[0083] Raw data requires systematic processing to support model construction, encompassing three main stages: cleaning, feature extraction, and standardization. Data cleaning employs differentiated processing for different signal types: infrared images are subject to interference from rain, snow, and equipment shadows, so Gaussian filtering (kernel size 5×5) is used for noise reduction, combined with morphological opening operations to remove small noise points, improving the signal-to-noise ratio to over 30dB; mosquito density data uses the 3σ criterion to identify outliers (such as zero values ​​or sudden increases caused by mosquito-attracting lamp malfunctions), and values ​​exceeding the range of [μ-3σ, μ+3σ] are corrected by a weighted average of the three nearest sampling points (the more recent the time, the higher the weight); occasional missing data (<5%) of garbage decomposition data is filled using linear interpolation, and missing data from three consecutive sampling points is marked as invalid and triggers a sampling equipment malfunction warning. Feature extraction takes into account both data characteristics and model requirements: At the mosquito feature level, the GhostYOLO algorithm (an improvement on YOLOv5, with the introduction of the Ghost module reducing the number of parameters by 30%) is used to process infrared and high-resolution images, outputting four core features: mosquito species (Aedes, Culex, Anopheles, etc.), density (number of mosquitoes / m²), proportion of the aggregation area to the functional area, and coordinates of the aggregation center, generating a 10-dimensional feature vector (including the proportion of each mosquito species); for environmental features, five time-domain features are extracted from meteorological data: daily average temperature, daily maximum temperature, daily average humidity, cumulative rainfall, and daily average wind speed. The frequency domain energy proportion of rainfall over 24 hours and wind speed over 12 hours is obtained through FFT transformation, constructing an 8-dimensional feature vector; for waste composting features, four key indicators are extracted: organic matter degradation rate, moisture content, carbon-nitrogen ratio, and pile temperature. The composting progress is calculated in combination with the pile stacking time, generating a 5-dimensional feature vector. The standardization process uses the Z-score formula to eliminate dimensional differences, mapping all features to the [-1,1] interval. A structured dataset containing timestamps, functional area numbers, 31-dimensional feature vectors (10+8+5+8 derived features), and risk level labels (subsequently manually labeled) is constructed. The dataset is divided into a training set (12,000 records) and a test set (3,000 records) in an 8:2 ratio. The dataset is compressed in Parquet format (40% compression rate) and a real-time data pipeline is built using Apache Kafka to ensure the timeliness of the model input data.

[0084] The mosquito breeding risk assessment model is constructed with multi-source feature fusion as its core, combined with spatial data to generate a visual cloud map. The core of the model employs a weighted fusion algorithm, with the basic formula being: The weight parameters were determined using the Analytic Hierarchy Process (AHP): Five environmental sanitation experts and three pest control engineers were invited to construct a judgment matrix, calculating the weights of the mosquito characteristic factor (α=0.45), environmental adaptability factor (β=0.3), and garbage decomposition factor (γ=0.25), satisfying α+β+γ=1. The scenario correction term ε was adjusted according to season and region: +0.05 in summer (June-August), -0.05 in winter (December-February), and an additional +0.03 during the rainy season. The calculation of each sub-factor was refined to adapt to the landfill scenario: mosquito characteristic factor... Among them, the risk coefficient of each type The density was calculated using a weighted average of Aedes mosquitoes (0.9), Culex mosquitoes (0.7), and Anopheles mosquitoes (0.6) (the weighted sum of the proportions of mixed populations was used), and the normalized density value was calculated. (D is the measured density, 20 units / m² is the critical value), aggregation index ( The area of ​​the cluster is The area of ​​the functional zone, (The aggregation intensity coefficient is corrected by Moran's I index); environmental adaptability factor. Temperature adaptability Using Gaussian function ( , Humidity adaptability Linear normalization ( , Rainfall suitability Wind speed adaptability Waste composting factors The maturity index M is set at 0 (fresh) - 1 (fully decomposed), and the organic content is... These are measured values, and the normalized stack height is the actual value. (H is the stack height, and 3m is the critical value). The model was optimized using gradient descent during training, and the test set accuracy reached 96.2% after 50 iterations. To achieve spatial assessment, topographic and distribution data were integrated: a 1:500 digital elevation model (DEM) with an accuracy of ±0.1m was generated by aerial photography of the site using a DJI Phantom 4 RTK drone, identifying low-lying areas prone to water accumulation; data on the amount of waste accumulated and turnover cycle in each functional area were obtained by connecting to the landfill MES system, combined with monitoring data on mosquito activity periods (18:00-22:00), and a breeding risk probability cloud map was generated using the Kriging interpolation method in ArcGIS 10.8, divided into 5 levels according to the R value: red (R≥0.8, very high), orange (0.6≤R<0.8, high), yellow (0.4≤R<0.6, medium), green (0.2≤R<0.4, low), and blue (R<0.2, very low), marking key breeding areas in each functional area (such as the low-lying area in the southwest corner of the composting area, R=0.85).

[0085] Based on risk cloud mapping, targeted delivery and dynamic optimization of probiotics are achieved, constructing a closed loop of "simulation-delivery-monitoring-adjustment". The probiotics use a compound formula system, with Bacillus subtilis CGMCC1.1086 and Bacillus licheniformis CGMCC1.1598 as the core components. A total of 12 schemes were set up with 3 ratios (1:1, 2:1, 1:2), 2 dosages (50g / ㎡, 80g / ㎡), and 2 frequencies (once daily, once every 2 days). A simulation model of the suppression effect was constructed using MATLAB R2023a. Inputs included the risk index of each functional area, waste decomposition degree, and environmental temperature and humidity data. The 72-hour mosquito density reduction rate was used as the evaluation index to simulate the optimal solution: For the composting area (extremely high risk), a 1:1 ratio of waste mix, 80g / m², once daily; for the temporary storage area (high risk), a 2:1 ratio of waste mix, 50g / m², once daily; for the unloading and sorting areas (medium risk), a 1:2 ratio of waste mix, 50g / m², once every two days; and for the collection area (low risk), a 1:2 ratio of waste mix, 50g / m², once every three days. The delivery was executed using a modified Ecovacs DEEBOTX2Pro ground robot, equipped with a GPS+UWB dual positioning module (positioning accuracy ±5cm), a 5kg payload, a 4-hour battery life, and a spiral feeder (delivery accuracy ±2g). The robot received path instructions from cloud mapping via a 4G module and delivered waste according to the priority of the functional area's risk level. A real-time monitoring system is deployed synchronously: Three CK-007 mosquito density sensors (detection range 0-100 mosquitoes / m², accuracy ±5%) are deployed in each functional area, with a sampling frequency of once every 2 hours; BTK-880 probiotic activity detectors (detection range 0-1000 CFU / g, accuracy ±10 CFU / g) are deployed in the distribution area, with activity measured once daily. Monitoring data is transmitted back to the server in real time. When the mosquito density decrease rate in a certain area is <30% or the probiotic activity is <500 CFU / g, the system automatically triggers an adjustment: if the density decrease rate in the composting area is only 25%, the distribution amount is increased to 100g / m², the ratio is adjusted to 1:0.8, and the distribution frequency is increased to twice daily; if the activity is insufficient, the probiotic storage time is extended, and the batch is replaced.

[0086] The 3D visualization monitoring and interaction system is built on Unity3D 2022.1 to achieve full-scene control of the prevention and control process. Geometric modeling adopts a dual-source fusion of "CAD + laser scanning": importing the original 3D models of the site terrain, buildings, and equipment (including 42 core components) from SolidWorks, performing topology optimization through Blender, deleting non-critical features such as bolts and signs with diameters <10mm, and simplifying the surface using the Catmull-Clark subdivision algorithm, reducing the number of polygons from 8 million to 2 million; using a FaroFocus S150 laser scanner to perform a 360° panoramic scan of the site to obtain point cloud data (point density 50 points / mm²), registering it with the CAD model using GeomagicWrap software (ICP algorithm, registration error <0.05mm), correcting the deviation between the actual shape of the compost pile and the model (e.g., the actual slope of the compost pile is 35°, while the original model was 30°), and finally constructing a 1:1 scale geometric twin in Unity3D, supporting LOD level of detail display (switching to a simplified model with 50,000 polygons at distances >10m). The rendering utilizes the URP rendering pipeline, configured with real-time global illumination (baking precision 512 texels / m, baking time 3 hours), achieving a stable frame rate of 60fps. A PBR material system is employed, with the pile material set to a metallicity of 0.1 and a roughness of 0.8 to simulate the texture of waste, and the robot set to a metallicity of 0.7 and a roughness of 0.3 to simulate a metal shell. Dynamic rendering focuses on core information: the risk cloud map colors are directly mapped to the 3D scene, extremely high-risk areas flash red (frequency 2Hz, duty cycle 50%), the robot's movement trajectory is updated in real-time with a solid blue line, and the delivery point is marked with a green dot displaying the delivery time and dosage. In abnormal states (such as a sudden increase in density), the corresponding area flashes red and a pop-up window displays details (parameter name, current value, threshold, and historical 3-day trend). The interactive features support multi-device operation: mouse interaction is achieved through ray detection; left-clicking an area brings up a parameter panel (including mosquito density, probiotic activity, and release records); right-clicking allows dragging to rotate the view (speed 20° / s); and scrolling with the scroll wheel zooms (coefficient 1.2). VR is compatible with the Pico4 headset, using ray detection via controllers and supporting gesture recognition (pinch to zoom the scene, wave to display data from the previous 2 hours), with a gesture recognition accuracy of 92%. The timeline control supports backtracking of nearly 72 hours of history, with data queried from InfluxDB by time range and smoothed transitions through interpolation algorithms. It integrates the ChartJS plugin to generate density-time and activity-time curves and supports exporting PDF reports (including screenshots and data tables).

[0087] The maintenance and control strategy constructs a multi-dimensional decision-making system to achieve precise prevention and control and efficient operation and maintenance. Based on a three-dimensional decision matrix of "risk level - production priority - environmental change," the basic strategy is as follows: For extremely high risk (R≥0.8), emergency prevention and control is implemented, with robots deploying fertilizer every 4 hours, and manual assistance in spraying 0.3% high-efficiency cypermethrin (diluted at 1:1000). The collection truck will increase the daily collection of fertilizer from the composting area to ensure the pile height is ≤2.5m. For high risk (0.6≤R<0.8), enhanced prevention and control is implemented, with robots deploying fertilizer twice daily, daily checks of mosquito traps and sensor calibration status, and weekly cleaning of surface water in the pile. For medium risk (0.4≤R<0.6), routine prevention and control is implemented, with deployment according to the standard plan and equipment inspection every 3 days. For low risk (R<0.4), simplified prevention and control is implemented, with deployment every 3 days and weekly inspection. The strategy optimization combines waste collection and environmental changes: When the disposal plan conflicts with the collection operation in the composting area, the system calculates the risk value of delayed disposal (risk value = mosquito density growth rate × delay time). If the risk value is <0.2 (e.g., the risk value is 0.15 for a 1-hour delay), the disposal time will be automatically postponed until the collection is completed. During this period, the monitoring frequency will be increased from once every 2 hours to once every 30 minutes. If the risk value is ≥0.2, the backup robot will be dispatched to complete the disposal in the high-risk area first.

[0088] A dual-terminal collaborative monitoring system of "local + remote" is constructed to improve the coverage of control. Local monitoring is deployed in the central control room, using a 32-inch 4K monitor (resolution 3840×2160) with split-screen display: the main window on the left displays a 3D scene, the middle displays a risk cloud map and parameter dashboard, and the right displays a fault warning window, supporting touch operation (response time ≤100ms); operators wear Pico4 headsets (resolution 2160×2160 / eye) to immerse themselves in the observation of mosquito gathering inside the reactor through 6DOF positioning (accuracy ±1mm), and use LeapMotion controllers to realize gesture interaction (such as virtually disassembling the reactor shell to observe the internal environment). Remote monitoring adopts a lightweight WebGL solution, converting Unity3D models to glTF format (compressing file size to 1 / 6 of the original model), and rendering it in a browser using the Three.js engine, supporting access on PC (Chrome / Firefox) and WeChat mini-programs; managers can view real-time risk levels, equipment operating status, and deployment statistics, and use annotation tools (rectangles, arrows) to mark key inspection areas in the scene (such as abnormal data from a certain sensor), and the annotation information is synchronized to the local system in real time.

[0089] Adaptive control enables full-process optimization, encompassing lifespan prediction and dynamic adjustment. Lifespan prediction for key equipment employs a Weibull distribution model, based on historical failure data from 30 sensors and 20 robot components. The parameters are fitted using maximum likelihood estimation, yielding β=2.5 and η=6000 hours. Inputting the current sensor operating temperature (35℃) and vibration value (0.05g), the remaining lifespan L10 is calculated to be 150 hours (90% confidence reliable lifespan), displayed as a countdown progress bar in a 3D scene. A spare parts replacement warning is triggered when the lifespan is less than 48 hours. The dynamic adjustment mechanism addresses risk anomalies: when the mosquito density in the composting area increases from 30 mosquitoes / m² to 40 mosquitoes / m² (exceeding the 35 mosquitoes / m² threshold), the system initiates twin simulation to simulate the suppression effect under different ratios and dosages, determining the optimal adjustment direction as "increasing the proportion of Bacillus subtilis + increasing the dosage"; the PLC controller (Siemens S7-1500) sends instructions to the robot via the Profinet protocol to adjust the ratio to 1:0.8, increase the dosage to 100g / m², and optimize the path to "center-of-aggregation priority"; continuous monitoring is performed 3 times (24-hour intervals), and if the density drops below 25 mosquitoes / m², the parameters are fixed; otherwise, the simulation-adjustment process is repeated. In case of an emergency (such as a sudden increase in density to 60 animals / m² due to water accumulation in the temporary storage area after a rainstorm), a three-level response will be triggered: Level 1 (0-200ms): The 3D scene flashes red across the entire screen, and the buzzer in the central control room sounds a 90dB alarm; Level 2 (200-500ms): Two backup robots are dispatched to the area to carry out temporary waste disposal (100g / m², 1:1 ratio); Level 3 (500ms-2s): An alert is pushed to the management personnel's mobile APP via the Alibaba Cloud IoT platform (including the location of the risk area and suggested measures), manual drainage is arranged and emergency disinfectant is sprayed, and the MES system is simultaneously marked to suspend waste storage in the area until the risk level drops below the medium level.

[0090] In summary, this application's embodiments, through comprehensive multi-source data collection and standardized processing, combined with GhostYOLO algorithm feature extraction and weighted fusion risk assessment model construction, along with the visualization of breeding risk probability cloud maps, can accurately identify mosquito breeding risks and pinpoint key breeding areas, ensuring the scientific rigor and relevance of risk assessment. Targeted probiotic delivery and dynamic optimization, 3D visualization monitoring, and multi-dimensional maintenance strategies work together to reduce waste of control resources, lower manual control costs, and significantly improve mosquito suppression effects. Dual-end collaborative monitoring and adaptive control mechanisms accelerate the response speed to risk anomalies and enhance emergency response capabilities in sudden scenarios; equipment lifespan prediction and graded response mechanisms enable early warning of potential faults and rapid risk management, comprehensively improving the accuracy, efficiency, environmental friendliness, and long-term effectiveness of mosquito control in landfills.

[0091] 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.

[0092] When the processor 1002 executes the program, it implements the method for precise delivery of probiotics to suppress mosquitoes in landfills based on image recognition, as provided in the above embodiments.

[0093] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0094] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0095] The memory 1001 may include high-speed RAM (Random Access Memory) and may also include non-volatile memory, such as at least one disk storage.

[0096] 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.

[0097] 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.

[0098] 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.

[0099] 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 precise delivery of probiotics to suppress mosquitoes in landfills based on image recognition.

[0100] Furthermore, this application also provides a computer program product, including a computer program or instructions, which, when executed, implement the above-mentioned method for precise delivery of probiotics to suppress mosquitoes in landfills based on image recognition.

[0101] 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.

[0102] 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.

[0103] 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.

[0104] 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.

[0105] Those skilled in the art will understand that all or part of the steps of the methods described 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.

[0106] 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 probiotic precision delivery system for mosquito control in landfills based on image recognition, characterized in that, include: The system includes an image acquisition and sensing module, a mosquito feature analysis center, a reproductive risk assessment module, a probiotic delivery decision unit, and an adaptive control module; among which, The image acquisition and sensing module is used to collect image data of the landfill, environmental parameters, mosquito population characteristics data, and landfill decomposition status data. The mosquito characteristic analysis center is used to extract mosquito species, density, and aggregation area characteristics, and to construct a mosquito breeding risk assessment model in combination with environmental parameters. The breeding risk assessment module is used to assess the mosquito breeding risk level and key breeding areas based on the mosquito breeding risk assessment model, integrating topographic data, garbage distribution data and mosquito activity data, and generating a breeding risk probability cloud map. The probiotic delivery decision unit is used to simulate the inhibitory effects of different probiotic formulations and delivery strategies based on the growth risk probability cloud map, generate a delivery plan, and drive the ground robot to perform targeted delivery. The adaptive control module is used to monitor changes in mosquito density and probiotic activity in real time, and dynamically adjusts the release plan through a swarm intelligence algorithm.

2. The probiotic precision delivery system for suppressing mosquitoes in landfills based on image recognition as described in claim 1, characterized in that, The image acquisition and sensing module includes an image acquisition unit, an environmental parameter acquisition unit, a mosquito population characteristic acquisition unit, and a garbage decomposition status acquisition unit. The image acquisition unit acquires panoramic and close-up images of the landfill at different times, covering areas with high mosquito activity. The environmental parameter acquisition unit collects data on temperature, humidity, light intensity, precipitation, and wind speed, simultaneously recording the acquisition time and location information. The mosquito population characteristic acquisition unit collects data on mosquito morphology, activity patterns, and habitat preferences. The garbage decomposition status acquisition unit collects data on the humidity, pH, odor level, and decomposition stage of the garbage, combining this data with the appearance of the garbage in the images to comprehensively determine the degree of decomposition.

3. The probiotic precision delivery system for suppressing mosquitoes in landfills based on image recognition as described in claim 1, characterized in that, The mosquito feature analysis center includes an image recognition unit, a feature extraction unit, and a model building unit. The image recognition unit uses the GhostYOLO algorithm to accurately identify mosquito species and distinguish between adults and larvae. The feature extraction unit extracts core features such as mosquito density, spatial distribution of aggregation areas, and peak activity periods. The model building unit correlates the extracted mosquito features with environmental parameters to construct a multi-factor coupled mosquito reproductive risk assessment model and outputs a quantitative value of reproductive potential.

4. The probiotic precision delivery system for suppressing mosquitoes in landfills based on image recognition as described in claim 1, characterized in that, The breeding risk assessment module includes a risk level assessment unit, a key area positioning unit, and a cloud map generation unit. The risk level assessment unit integrates topographic data, garbage distribution data, and mosquito activity data, and uses the analytic hierarchy process (AHP) to determine the weights of each factor, classifying mosquito breeding risk into four levels: low, medium, high, and extremely high. The key area positioning unit uses risk value overlay analysis to identify key breeding areas and marks area boundaries and core locations. The cloud map generation unit uses a kernel density estimation algorithm to map risk probability values ​​to a three-dimensional geographic information model of the landfill, and uses color gradients to represent the risk intensity distribution, generating a breeding risk probability cloud map that includes risk level, key area coordinates, and risk diffusion trends.

5. The probiotic precision delivery system for suppressing mosquitoes in landfills based on image recognition as described in claim 1, characterized in that, The probiotic delivery decision unit includes a formula screening module, an inhibition effect simulation unit, a delivery plan generation unit, and a targeted delivery drive unit. The formula screening module selects suitable probiotic strain combinations based on mosquito species and the state of waste decomposition. The inhibition effect simulation unit, based on the mosquito inhibition mechanisms of different probiotic formulas and combined with a breeding risk probability cloud map, uses a multi-objective optimization algorithm to simulate the decrease in mosquito density under different formulas, dosages, delivery times, and delivery frequencies, evaluating the inhibition effect and duration. The delivery plan generation unit, based on the simulation results, determines the specific probiotic formula, single-dose dosage, delivery frequency, precise delivery area, and optimal delivery time, forming a standardized delivery plan. The targeted delivery drive unit plans the delivery path of the ground robot, generating a target movement path based on the coordinates of key breeding areas and combined with landfill terrain and obstacle information, driving the robot to the delivery area to complete the targeted delivery operation.

6. The probiotic precision delivery system for suppressing mosquitoes in landfills based on image recognition as described in claim 1, characterized in that, The adaptive control module includes a real-time monitoring unit and a scheme adjustment unit. The real-time monitoring unit is used to continuously monitor changes in mosquito density, probiotic activity, and fluctuations in environmental parameters. The scheme adjustment unit analyzes the monitoring data, explores the correlation between changes in mosquito density and probiotic administration, and dynamically adjusts the administration scheme from three dimensions: administration dosage, administration frequency, and probiotic formula, to ensure a continuous and stable inhibitory effect.

7. A method for applying an image recognition-based probiotic precise delivery system for mosquito control in landfills, as described in any one of claims 1-6, characterized in that, The method includes: Acquire image data of landfills, environmental parameters, mosquito population characteristics, and landfill decomposition status data; Based on the landfill image data and mosquito population characteristic data, the GhostYOLO algorithm is used to extract mosquito species, density, and aggregation area characteristics. Combined with the environmental parameters and the landfill decomposition status data, a mosquito breeding risk assessment model is constructed. Based on the mosquito breeding risk assessment model, the topography, garbage distribution and mosquito activity data are integrated to assess the mosquito breeding risk level and key breeding areas, and generate a breeding risk probability cloud map. Based on the aforementioned breeding risk probability cloud map, the inhibitory effects of different probiotic formulations and delivery strategies are simulated to generate a delivery plan, which drives a ground robot to perform targeted delivery. At the same time, the changes in mosquito density and probiotic activity are monitored in real time, and the delivery plan is dynamically adjusted accordingly.

8. An electronic device, characterized in that, The invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the method of image recognition-based precise delivery of probiotics to suppress mosquitoes in landfills, as described in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When a computer program or instruction is executed, it implements the method of image recognition-based precise delivery of probiotics to suppress mosquitoes in landfills, as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When a computer program or instruction is executed, it implements the method of image recognition-based precise delivery of probiotics to suppress mosquitoes in landfills, as described in claim 7.

Images