Collection and transportation system optimization method based on rural household garbage spatial distribution characteristics

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a collection and transportation system optimization method based on spatial distribution characteristics of rural household garbage. The method comprises the following steps: (1) remote sensing image identification: identifying and extracting typical land images of a rural area by adopting a D-LinkNet model; (2) resident area partitioning: according to the resident space aggregation degree, dividing the resident area into three space aggregation forms of high aggregation, medium aggregation and dispersion by adopting an optimized hotspot analysis method; (3) collecting point site selection: constructing a relation function between the resident area and the garbage yield, clustering the resident points in the three spatial aggregation forms by adopting a multiple K-means clustering algorithm, setting a buffer distance, and optimizing and determining the position of the garbage collecting point; and (4) carrying out collection and transportation path optimization: based on the collection and transportation requirements of the three space aggregation form areas, taking the shortest collection and transportation path as a target, and adopting a genetic algorithm to solve an optimal garbage collection and transportation path scheme, so that scientific planning of a rural household garbage collection and transportation system is realized, and collection and transportation cost and environmental pollution are reduced.

Description

manual

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a collection and transportation system optimization method based on spatial distribution characteristics of rural domestic waste. [Background technology]

[0002] The domestic waste collection and transportation system is a key link in the whole process management system of rural domestic waste. Its cost accounts for 60% to 80% of the total cost of the waste treatment system. Inefficient collection of waste and unreasonable planning of transportation routes will significantly increase the collection and transportation costs, making the waste collection and transportation system unable to operate in the long term, resulting in a large amount of waste not being cleared and transported in time and piling up, causing serious domestic waste pollution. In addition, due to the complex terrain conditions in rural areas, uneven population distribution, large differences in area, and relatively scattered residence of farmers, the generation of rural domestic waste is unevenly distributed, the garbage collection density is small and widely distributed, making it difficult to collect and transport garbage, and the cost is high, making it difficult to establish a sound garbage collection and transportation system. In recent years, there have been many studies on rural domestic waste collection and transportation systems at home and abroad. In developed countries, the research on rural domestic waste collection and transportation systems started early, the facilities are relatively complete, and the collection and transportation system is mature. However, my country's research on garbage collection and transportation systems started late, and there are currently few studies on rural domestic waste collection and transportation systems. Most of the research focuses on exploring the composition, properties, and collection and transportation modes of rural domestic waste. In the exploration of path planning, previous studies have only made qualitative judgments based on internal requirements such as convenient transportation and residents' acceptance, lacking the use of quantitative methods to analyze the impact of these factors on facility site selection, and failing to fully consider the characteristics of uneven distribution of rural population and the spatial distribution characteristics of different aggregation forms of rural domestic waste, resulting in the poor applicability of rural domestic waste collection and transportation modes. Therefore, according to the spatial distribution characteristics of rural domestic waste and considering the garbage collection needs in areas with different household concentrations, a domestic waste collection point site selection model and collection and transportation route planning model suitable for rural areas are constructed, which is of great significance for the scientific design and efficient operation of the rural domestic waste collection and transportation system. [Summary of the invention]

[0003] The purpose of the present invention is to provide a method for optimizing a rural domestic waste collection and transportation system by combining the characteristics of rural waste generation with comprehensive consideration of scientificity, economy and environmental performance. The technical solution of the present invention is: 1. A collection and transportation system optimization method based on the spatial distribution characteristics of rural domestic waste, the steps are: (1) Remote sensing image recognition: Using the D-LinkNet model, the pixel areas corresponding to specific target elements in remote sensing images are automatically divided to identify and extract basic rural environmental information; (2) Zoning of residential areas: Based on the degree of spatial aggregation of residents, the optimized hot spot analysis method is used to divide the residential areas into three spatial aggregation forms: high aggregation, medium aggregation and dispersed areas; (3) Collection point location selection: Use the multiple K-means clustering algorithm to cluster the three types of spatial clustering forms of residents; set buffer distances based on the principle of environmental friendliness; conduct buffer analysis to optimize the location of garbage collection points; (4) Optimization of collection and transportation routes: Based on the collection and transportation needs of the three types of spatial aggregation areas, a garbage collection and transportation route optimization model is constructed with the goal of minimizing the collection and transportation route. A genetic algorithm is used to solve the optimal garbage collection and transportation route plans for high-aggregation areas, medium-aggregation areas, and dispersed areas respectively. The regional garbage collection and transportation plan is optimized based on the balance of collection and transportation workload. 2. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: in the step (1), the remote sensing image recognition method is: (2-1) Based on the latest national land survey results and high-resolution satellite remote sensing images, select sample blocks of no less than 1000*1000 pixels, including the outlines of house roofs, road boundaries, water boundaries, and different land types such as cultivated land and forest land, as well as sensitive target boundaries; (2-2) The selected sample blocks are used for D-LinkNet network model training. Based on LinkNet, a hole convolution layer is added to its core for road extraction tasks. The receptive field of the convolution kernel is expanded by calculating the linear combination of a kernel parameter and a finite number of pixels within its range. A dilation block consisting of four dilated convolution processes with different dilation degrees is introduced between the encoder and the decoder. Secondly, network training and prediction are performed. All semantic segmentation network models are implemented using the PyTorch framework. The model weights pre-trained with ImageNet are used to initialize D-LinkNet. After D-LinkNet enters the network, epochs are trained on the dataset. Data augmentation strategies including random cropping, rotation, translation, and horizontal flipping operations are applied to expand the dataset and reduce overfitting caused by insufficient data. (2-3) The trained model is used to identify typical rural land types, and to identify and extract vector data of different land types such as residential points, roads, water bodies, cultivated land, forest land, and other sensitive targets. The D-LinkNet network model is used to extract the building outline, calculate the center point of the outline, and then use the outline area as the attribute value of the residential point; extract the center line of the road recognition result and construct a road network; identify the cultivated land boundary (ridge) and construct a complete cultivated land plot. 3. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: in the step (2), the method for zoning the household areas is: (3-1) The topological analysis method is used to determine the spatial relationship between each unit, and the projection coordinates of the household surface elements in the study area are set. Based on the recognition of the household remote sensing images, the household surface elements can be obtained and converted into household point elements. (3-2) Spatial statistical analysis: Explore the spatial distribution relationship of feature classes by quantifying spatial relationships and conduct statistical analysis on the attribute information of spatial objects, including quantitative calculation of spatial data (area calculation) and spatial statistical analysis (aggregated, dispersed or random spatial patterns). (3-3) Optimized Hot Spot Analysis (OHSA): By calculating the high / low clustering (Getis-Ord Gi*) statistics of each data point in the region, it is determined whether the point belongs to the same category as the adjacent points. The model is as follows: As shown in formula 3-1. In the formula, x j is the attribute value of household j, w i,j is the spatial weight of the spatial adjacency between household point i and its neighboring household point j, which is defined by using the chord distance between the two points and the spatial distance threshold. n is the total number of features. The statistic is the z-score. If the z-score is positive and the z-value is higher, it indicates that the spatial clustering is tighter (Hotspot), which is the household point clustering area; if the z-score is close to zero, it means that there is no obvious spatial clustering (NS), which is the household point medium clustering area; if the z-score is negative and the z-value is lower, it indicates that the spatial clustering is more dispersed (Coldspot), which is the household point dispersed area. (3-4) Combined with ArcMap and other hot spot analysis tools, the rural households were divided into three different residential areas with different concentration levels: high concentration area, medium concentration area and dispersed area. Based on the zoning results, ArcMap and other geoprocessing clipping tools were used to take the original extracted household surface data as input elements and the point data of the three different spatial concentration degree household partitions as clipping elements to clip the household surface data of the three areas. 4. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: in the step (3), the site selection method for establishing the waste collection point is: (4-1) Calculate the number of garbage collection points and determine the initial number of clusters; calculate the service radius of the garbage collection points and determine the clustering threshold; input the initial number of clusters, the clustering threshold and the extracted household point vector data into the multiple K-means algorithm to perform multiple K-means clustering to obtain highly clustered, medium clustered and dispersed household point cluster sets respectively; (4-2) According to the clustering of household points, set up household buffer zones, where the buffer distance for residents in high-density areas shall be no less than 20m, for residents in medium-density areas shall be no less than 30m, and for residents in dispersed areas shall be no less than 50m; set up sensitive target buffer zones, where the buffer distance for water bodies shall be no less than 50m, the buffer distance for roads shall be no less than 5m, and the buffer distance for farmland shall be no less than 50m; (4-3) Using the buffer analysis model of GIS and other software, we conduct buffer analysis of households, water bodies, roads, farmlands and other sensitive targets, and obtain the buffer distance vector data set of households, water bodies, roads and farmlands. We intersect and erase them to obtain the vector data set of candidate areas suitable for collection points. Based on the principle of giving priority to the location of the cluster center of the household points, the intersection of roads, and the roadside accessible to the collection vehicles, we obtain the vector data set of the garbage collection point locations. 5. The method for selecting a location for a garbage collection point according to claim 4, characterized in that: in said step (4-1), (5-1) The number of garbage collection points is calculated as follows: First, extract the household point vector dataset and the household point area dataset, construct the per capita daily garbage production dataset and the per capita housing area dataset, and establish the relationship function between household area and garbage production: In the formula, q represents the daily garbage output of each household (kg / d), and a represents the area occupied by each household (m 2 ), θ represents the per capita housing area (m 2 / person), c represents the daily per capita garbage output (kg / person / d), A1 is the daily garbage discharge weight unevenness coefficient, and A2 is the resident population variation coefficient. Secondly, the summary tool in GIS software was used to obtain the total daily garbage production data set of the three areas of high concentration area, medium concentration area and dispersed area; according to the garbage collection point quantity calculation formula in the Technical Code for Collection and Transportation of Domestic Waste (CJJ 205), the collection and transportation frequency was 1d / time in high concentration area, 2-3d / time in medium concentration area, and 4-7d / time in dispersed area, the garbage collection point quantity data set of the three areas was calculated respectively. (5-2) The calculation method of the service radius of the garbage collection point is: First, according to the central place theory, the relationship function between the number of garbage collection points, regional area and service radius of garbage collection points is established as follows: Where A is the total area of ​​the region (m 2 );N max S is the number of garbage points required to be set up during the garbage peak period; cis the hexagonal area of ​​the collection point service range (m 2 ); r is the service radius of the garbage collection point (m). Secondly, based on the household point vector dataset, the GIS software was used to create surface feature tools to obtain the regional total area datasets of high-density areas, medium-density areas and dispersed areas. According to formula (5-2), the service radius of the garbage collection points in high-density areas, medium-density areas and dispersed areas was calculated. (5-3) The multiple K-means clustering method is: According to the number of garbage collection points and the service radius of garbage collection points, the household vector data of three spatial clustering forms, namely high clustering, medium clustering and dispersed areas, are input respectively. The maximum distance (MD) between each point in each cluster and the size of the service radius (SD) of the garbage collection point are compared. If MD is less than SD, the cluster is assigned as a completed cluster. Otherwise, clustering is continued until all cluster centers no longer change, and the household clustering analysis results of the three areas are obtained. 6. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: in the step (4), the collection and transportation path optimization method is: (6-1) The objective function with the shortest collection and transportation path as the goal is: Where V is the collection point set, where 0 represents the garbage transfer station, 1, 2, 3, ..., n represents the garbage collection point; K is the required garbage collection vehicle set; d ij represents the actual distance between garbage collection points or between garbage collection points and garbage transfer stations, that is, the distance from node i to node j. ijk Defined using 0-1 integer decision variables: (6-2) Set the full coverage constraint to: Constraint 1: All collection vehicles must leave a collection point after arriving at the collection point: Constraint 2: Eliminate sub-loop constraints: In the formula, Q is a proper subset of the node set V and does not include the nodes of the transfer station; Constraint 3: All collection vehicles are required to start from the transfer station and return to the transfer station after completing the collection task: (6-3) Set the load constraint condition as: In the formula, qi is the amount of garbage collected at point i; Q k is the maximum load of the vehicle; (6-4) Set the maximum driving distance constraint as: Where L is the maximum restricted driving distance of the vehicle; (6-5) Input the garbage collection and transportation route optimization model established in steps (6-1), (6-2), (6-3), and (6-4); (6-6) A daily garbage production dataset of three regions with high concentration, medium concentration and dispersion was used; the newly created OD cost matrix tool in GIS software was used to obtain the distance set between each garbage collection point and the distance set between each garbage collection point and the garbage compression transfer station; the collection vehicle load and driving distance limit were set; and the genetic algorithm was used to solve the problem, and the optimal collection and transportation path set for the high concentration area, medium concentration area and dispersion area was obtained respectively. (6-7) Based on the assumption that the daily collection distance of each garbage collection truck does not exceed 20 km, different optimization paths in high-density areas, medium-density areas, and dispersed areas are combined and optimized to obtain the regional garbage collection and transportation optimization plan. The present invention comprehensively considers the technical, economic and environmental aspects, combines the spatial distribution characteristics of rural domestic waste, obtains the optimization plan of the rural domestic waste collection and transportation system through cluster analysis and establishment of a path optimization model, and effectively improves the efficiency of rural waste collection and transportation. It is of great significance to reduce the cost of domestic waste treatment, improve the level of waste management, and improve rural ecological environment pollution.

Brief Description of the Drawings

[0004] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them: Figure 1 Optimize the flow chart for rural domestic waste collection and transportation; Figure 2 This is a schematic diagram of the overall architecture of D-LinkNet; Figure 3 Schematic diagram of typical example image patches with corresponding labels; Figure 4 Schematic diagram of data preprocessing process; Figure 5 Schematic diagram of the hotspot analysis aggregation process; Figure 6 Schematic diagram of the multiple K-means clustering algorithm process; Figure 7It is a schematic diagram of the multiple K-means clustering process; Figure 8 This is a schematic diagram of the regional partitioning results provided by the present invention in an embodiment; Fig. 9 The following is a schematic diagram of the regional clustering results provided by the present invention in an embodiment; Fig.10 The embodiment is a schematic diagram of the environmental constraint buffering result provided by the present invention; Fig.11 This is a schematic diagram of the collection point site selection results provided by the present invention in an embodiment; Fig.12 This is a schematic diagram of the optimization result of the garbage collection and transportation path provided by the present invention in an embodiment; [Specific implementation method]

[0005] The present invention is further described below in conjunction with specific embodiments to help understand the content of the present invention. 1. Residential area division 1) Methods First, for the above-mentioned embodiment steps, the regional remote sensing image was downloaded from the BIGEMAP high-definition map resource website, and then the basic data of rural household points was extracted using the D-LinkNet model. After the basic data is ready, start the ArcMap 10.6 software, load the basic data into the software interface, open the ArcToolbox toolbox → spatial statistics tool → cluster distribution mapping → optimized hot spot analysis to start the regional partition function, and the specific operations are as follows: The basic data of household points are added to the input features, and the event data aggregation strategy chooses to capture nearby events to create weighted points, that is, spatial point aggregation is used for regional zoning. 2) Result analysis Hot spot analysis in ArcMap 10.6 software divides the area into three categories: high concentration area, medium concentration area and scattered area according to the spatial concentration of household points. The software gives the number of households in each category (see Table 1) and the distribution of household points (see Figure 8. The households in the high-density area are mainly divided into clusters and strips. The clusters are mainly located in the central area of ​​the study area (Dongcheng Community and Yongfu Community) and the town-town combination area within two kilometers (Yanjing Village). The strips are mainly located in two villages with relatively dense populations in the town (Fangbei Village and Hepingqiao Village). The distance between households does not exceed 50m. The households in the medium-density area mainly present a relatively sparse network and strip structure in space. Most of them are composed of 20 to 30 households gathered together to form a unit. The distance between each aggregation unit does not exceed 300m, and the distance between households in the aggregation unit does not exceed 100m. The scattered area has the characteristics of self-contained scattered settlement. Most of them are single households and surrounding bamboo and wood forests, courtyards, etc. to form a residential unit. There are also 3 to 5 households forming a larger courtyard residential unit. The distance between each household is about 200m. Table 1 Regional division results 2. Location of collection points According to the relationship between the household area and the garbage output, the daily garbage output of the area was calculated, and then the location of the collection point was determined by using the multiple K-means clustering algorithm and environmental buffer analysis. 1) Clustering process First, the geoprocessing clipping tool of ArcMap 10.6 software was used to convert the divided household points into household areas, and the area of ​​each household area in the three regions was obtained. Then, according to the above steps (5-1) and (5-2), the number of garbage collection points and service radius of each area were calculated. The calculation parameter settings are shown in Table 2. Table 2 Calculation parameters of the number of garbage collection points and service radius The calculated number of garbage collection points and service radius in each area, that is, the number of clusters and clustering threshold parameters of the multiple clustering algorithm are shown in Table 3. Table 3 Clustering parameters Open the VS Code software, load the multiple K-means clustering algorithm into the operation interface, add the household point vector files of the high-density area, medium-density area and dispersed area obtained in steps (4-7) respectively, and set the parameters according to the results in Table 3. Click the Iteration button to enter the iteration dialog box until all clusters meet the maximum service distance requirement. Stop the iteration, export the clustering result vector file, and load the file into ArcMap 10.6 software for visualization. The results are shown in Fig. 9 . 2) Clustering results According to the results of multiple K-means cluster analysis, cluster analysis was performed on three types of clustering areas. In high-density areas, the households were divided into 26 categories, with an average of 180 households in each category. In medium-density areas, the households were divided into 54 categories, with an average of 97 households in each category. In dispersed areas, the households were divided into 25 categories, with an average of 52 households in each category. Based on the calculation of three people per household, if the high-density area collects garbage once a day, one collection point can serve 540 people, if the medium-density area collects garbage once every two days, one collection point can serve 582 people, and if the dispersed area collects garbage once every three days, one collection point can serve 468 people. 3) Buffer operation Combined with the environmental factors that affect site selection, four environmental factors were considered, including the distance to water bodies, the distance to residential points, and the distance to roads. The buffer distance settings are shown in Table 4. Table 4 Buffer distance Open the ArcMap 10.6 software, select the Analysis Tool → Domain Analysis → Buffer Tool in the ArcToolbox toolbox, and use the high-density household point area, medium-density household point area, and dispersed household point area as input elements, set the buffer distance to 20m, 30m, and 50m respectively, and output the minimum protection area of ​​the household points in the three areas, that is, it is not suitable to set up garbage collection points in this area. In the same way, input the household points of the three areas as input elements, set the collection and transportation radius to 462m, 630m, and 534m respectively, and output the maximum delivery area of ​​the household points in the three areas, that is, garbage collection points should be selected within this range to meet the requirements of garbage collection services. Since the area within the delivery distance includes the minimum protection area, and the collection point cannot be set within the minimum protection area, the erase tool in ArcMap 10.6 software is used to remove the minimum protection area to obtain the household buffer area, see Fig.10 . In addition, water bodies and roads are also used as input elements, and the buffer radius is set to 50m and 5m respectively, and then the water body buffer zone and road buffer zone are output respectively. Since garbage collection points should be set on the roadside and away from water bodies, that is, inside the road buffer zone and outside the water body buffer zone. Therefore, the household point buffer zone and the road buffer zone are intersected, and then the water body is removed using the erase tool of ArcMap 10.6, and finally the suitable site selection area for garbage collection points is obtained. Finally, combined with the environmental constraint buffer results, according to the collection point setting method of the "Environmental Sanitation Facilities Setting Standard" (CJJ 27-2012), the locations of garbage collection points are determined in the alternative areas of each district. See Figure 2. Fig.11 . 4) Buffering results There are two main types of garbage collection point candidate areas. The first type is that the collection point can be selected on the road connecting two villages. The entrance or the end of the village are the best locations for the collection point. The second type can be selected on the main road next to the village. To facilitate collection and delivery, the collection point can be set at the intersection of the road or at a location that is moderately far from the two villages. 3. Path Optimization 1) Path optimization model parameters For the above implementation steps, the garbage collection vehicles are planned to be self-unloading garbage collection and transportation tricycles with a load of 1.5t and a speed of 20km / h. The maximum driving distance of each vehicle is set to 40km. The total daily output of domestic garbage in the three areas is collected once a day in high-density areas, once every two days in medium-density areas, and once every three days in dispersed areas. According to the total amount of garbage in each area, the number of garbage collection points, the collection and transportation frequency of each area, and the load capacity of each vehicle, the daily garbage output of each garbage collection point and the number of vehicles required in the area are calculated, see Table 5. Table 5. Parameters related to solving the vehicle routing optimization model In areas with high household concentration, 6.43 tons of garbage needs to be collected in one cycle if the garbage is collected every day. According to the rated load of the garbage collection and transportation tricycle, at least 6 garbage collection routes need to be arranged every day, and each route can collect up to 4 garbage collection points to ensure that the garbage collection work in the area is completed under the condition of daily production and daily clearance; in areas with medium concentration, 13.54 tons of garbage needs to be collected in one cycle if the garbage is collected every two days, and at least 12 garbage collection and transportation routes need to be arranged in total, with an average of 5 garbage collection points collected on each route. Combined with the actual work arrangements, 6 collection and transportation routes need to be evenly arranged every day to ensure that the garbage collection work in the area is evenly completed at a collection frequency of every two days; in areas with dispersed areas, 6.30 tons of garbage needs to be collected in one cycle if the garbage is collected every three days, and at least 6 garbage collection and transportation routes need to be arranged in total, with a maximum of 4 garbage collection points collected on each route. Combined with the actual work arrangements, 2 collection and transportation routes need to be evenly arranged every day to ensure that the garbage collection work in the area is evenly completed at a collection frequency of every three days. Open ArcMap 10.6 software and use the new OD cost matrix tool in the software to measure the driving distance between the locations of each garbage collection point set in steps (4-7), as well as the driving distance between each garbage collection point and the location of the garbage compression transfer station in the area. The above data and the number of vehicles, vehicle rated load, and vehicle maximum driving distance limit are brought into the path optimization model, and the model is solved using Matlab software. 2) Path optimization results The serial numbers and travel distances of garbage collection and transportation routes in high-density residential areas are shown in Table 6. The total distance traveled by vehicles is 43.03 km, and the number of vehicles deployed is 1. Table 6 The order of garbage collection and transportation in high-density areas The order of garbage collection and transportation in high-density areas The serial numbers and travel distances of garbage collection and transportation routes in the residential cluster area are shown in Table 7. The total travel distance of the vehicles is 139.44 km, the number of vehicles deployed per day is 2, and the total travel distance of each collection vehicle does not differ by more than 2 km. Table 7 Garbage collection and transportation sequence in the cluster area The serial number and travel distance of garbage collection and transportation routes in the residential scattered area are shown in Table 8. The total distance traveled by the vehicle is 82.98 km, and the number of vehicles deployed per day is 1. The collection and transportation distance of this collection vehicle does not differ by more than 2 km every day. Table 8 Sequence of garbage collection and transportation in scattered areas In summary, the high-density area collects and transports garbage on 6 routes every day, with a total route of 43.03 km and 1 vehicle. The medium-density area also collects and transports garbage on 6 routes every day, with a total route of 69.72 km and 2 vehicles per day. The average distance traveled by each collection vehicle is 34.86 km. The dispersed area collects and transports garbage on 2 routes every day, with a total route of 27.66 km and 1 vehicle per day. The difference in the distance traveled by this collection vehicle per day is no more than 2 km. The garbage collection and transportation routes of the three areas can be seen in Figure 2. Fig.12 Among the three areas, the medium-aggregate area has the largest number of garbage collection points and covers the largest area. Therefore, the total collection and transportation routes every day are as follows: medium-aggregate area > high-aggregate area > dispersed area.

Claims

1. A collection and transportation system optimization method based on the spatial distribution characteristics of rural domestic waste, the steps are: (1) Remote sensing image recognition: Using the D-LinkNet model, the pixel areas corresponding to specific target elements in remote sensing images are automatically divided to identify and extract basic rural environmental information; (2) Zoning of residential areas: Based on the degree of spatial aggregation of residents, the optimized hot spot analysis method is used to divide the residential areas into three spatial aggregation forms: high aggregation, medium aggregation and dispersed areas; (3) Collection point location selection: Use the multiple K-means clustering algorithm to cluster the three types of spatial clustering forms of residents; set buffer distances based on the principle of environmental friendliness; conduct buffer analysis to optimize the location of garbage collection points; (4) Optimization of collection and transportation routes: Based on the collection and transportation needs of the three types of spatial aggregation areas, a garbage collection and transportation route optimization model is constructed with the goal of minimizing the collection and transportation route. A genetic algorithm is used to solve the optimal garbage collection and transportation route plans for high-aggregation areas, medium-aggregation areas, and dispersed areas respectively. The regional garbage collection and transportation plan is optimized based on the balance of collection and transportation workload.

2. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: In step (1), the remote sensing image recognition method is: (2-1) Based on the latest national land survey results and high-resolution satellite remote sensing images, select sample blocks of no less than 1000*1000 pixels, including the outlines of house roofs, road boundaries, water boundaries, and different land types such as cultivated land and forest land, as well as sensitive target boundaries; (2-2) The selected sample blocks are used for D-LinkNet network model training. Based on LinkNet, a hole convolution layer is added to its core for road extraction tasks. The receptive field of the convolution kernel is expanded by calculating the linear combination of a kernel parameter and a finite number of pixels within its range. A dilation block consisting of four dilated convolution processes with different dilation degrees is introduced between the encoder and the decoder. Secondly, network training and prediction are performed. All semantic segmentation network models are implemented using the PyTorch framework. The model weights pre-trained with ImageNet are used to initialize D-LinkNet. After D-LinkNet enters the network, epochs are trained on the dataset. Data augmentation strategies including random cropping, rotation, translation, and horizontal flipping operations are applied to expand the dataset and reduce overfitting caused by insufficient data. (2-3) The trained model is used to identify typical rural land types, and to identify and extract vector data of different land types such as residential points, roads, water bodies, cultivated land, forest land, and other sensitive targets. The D-LinkNet network model is used to extract the building outline, calculate the center point of the outline, and then use the outline area as the attribute value of the residential point; extract the center line of the road recognition result and construct a road network; identify the cultivated land boundary (ridge) and construct a complete cultivated land plot.

3. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: In step (2), the method of zoning the household areas is: (3-1) The topological analysis method is used to determine the spatial relationship between each unit, and the projection coordinates of the household surface elements in the study area are set. Based on the recognition of the household remote sensing images, the household surface elements can be obtained and converted into household point elements. (3-2) Spatial statistical analysis: Explore the spatial distribution relationship of feature classes by quantifying spatial relationships and conduct statistical analysis on the attribute information of spatial objects, including quantitative calculation of spatial data (area calculation) and spatial statistical analysis (aggregated, dispersed or random spatial patterns). (3-3) Optimized Hot Spot Analysis (OHSA): By calculating the high / low clustering (Getis-OrdGi*) statistic of each data point in the area, it is determined whether the point belongs to the same category as the adjacent points. The model is shown in Formula 3-1. In the formula, x j is the attribute value of household j, w i,j is the spatial weight of the spatial adjacency relationship between household point i and its neighboring household point j, which is defined by using the chord distance between the two points and the spatial distance threshold, n is the total number of elements, G i * The statistic is the z-score. If the z-score is positive and the z-value is higher, it indicates that the spatial clustering is tighter (Hotspot), which is the household point clustering area; if the z-score is close to zero, it means that there is no obvious spatial clustering (NS), which is the household point medium clustering area; if the z-score is negative and the z-value is lower, it indicates that the spatial clustering is more dispersed (Coldspot), which is the household point dispersed area. (3-4) Combined with ArcMap and other hot spot analysis tools, the rural households were divided into three different residential areas with different concentration levels: high concentration area, medium concentration area and dispersed area. Based on the zoning results, ArcMap and other geoprocessing clipping tools were used to take the original extracted household surface data as input elements and the point data of the three different spatial concentration degree household partitions as clipping elements to clip the household surface data of the three areas.

4. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: In step (3), the site selection method for establishing the garbage collection point is: (4-1) Calculate the number of garbage collection points and determine the initial number of clusters; calculate the service radius of the garbage collection points and determine the clustering threshold; The initial cluster number, cluster threshold and extracted household point vector data are input into the multiple K-means algorithm to perform multiple K-means clustering to obtain highly clustered, medium clustered and dispersed household point cluster sets respectively; (4-2) According to the clustering of household points, set up household buffer zones, where the buffer distance for residents in high-density areas shall be no less than 20m, for residents in medium-density areas shall be no less than 30m, and for residents in dispersed areas shall be no less than 50m; set up sensitive target buffer zones, where the buffer distance for water bodies shall be no less than 50m, the buffer distance for roads shall be no less than 5m, and the buffer distance for farmland shall be no less than 50m; (4-3) Using the buffer analysis model of GIS and other software, we conduct buffer analysis of households, water bodies, roads, farmlands and other sensitive targets, and obtain the buffer distance vector data set of households, water bodies, roads and farmlands. We intersect and erase them to obtain the vector data set of candidate areas suitable for collection points. Based on the principle of giving priority to the location of the cluster center of the household points, the intersection of roads, and the roadside accessible to the collection vehicles, we obtain the vector data set of the garbage collection point locations.

5. The method for selecting a location for a garbage collection point according to claim 4, characterized in that: In the step (4-1), (5-1) The number of garbage collection points is calculated as follows: First, extract the household point vector dataset and the household point area dataset, construct the per capita daily garbage production dataset and the per capita housing area dataset, and establish the relationship function between household area and garbage production: In the formula, q represents the daily garbage output of each household (kg / d), and a represents the area occupied by each household (m 2 ), θ represents the per capita housing area (m 2 / person), c represents the daily per capita garbage output (kg / person / d), A1 is the daily garbage discharge weight unevenness coefficient, and A2 is the resident population variation coefficient. Secondly, the summary tools in GIS and other software were used to obtain the total daily garbage production data set of the three areas of high concentration area, medium concentration area and dispersed area; according to the garbage collection point quantity calculation formula in the Technical Code for Collection and Transportation of Domestic Waste (CJJ 205), the collection and transportation frequency was 1d / time in high concentration area, 2-3d / time in medium concentration area, and 4-7d / time in dispersed area, the garbage collection point quantity data set of the three areas was calculated respectively. (5-2) The calculation method of the service radius of the garbage collection point is: First, according to the central place theory, the relationship function between the number of garbage collection points, regional area and service radius of garbage collection points is established as follows: Where A is the total area of ​​the region (m 2 );N max S is the number of garbage points required to be set up during the garbage peak period; c is the hexagonal area of ​​the collection point service range (m 2 ); r is the service radius of the garbage collection point (m). Secondly, based on the household point vector dataset, the GIS software was used to create surface feature tools to obtain the regional total area datasets of high-density areas, medium-density areas and dispersed areas. According to formula (5-2), the service radius of the garbage collection points in high-density areas, medium-density areas and dispersed areas was calculated. (5-3) The multiple K-means clustering method is: According to the number of garbage collection points and the service radius of garbage collection points, the household vector data of three spatial clustering forms, namely high clustering, medium clustering and dispersed areas, are input respectively. The maximum distance (MD) between each point in each cluster and the size of the service radius (SD) of the garbage collection point are compared. If MD is less than SD, the cluster is assigned as a completed cluster. Otherwise, clustering is continued until all cluster centers no longer change, and the household clustering analysis results of the three areas are obtained.

6. The method for optimizing the collection and transportation system based on the spatial distribution characteristics of rural domestic waste according to claim 1, characterized in that: In step (4), the collection and transportation path optimization method is: (6-1) The objective function with the shortest collection and transportation path as the goal is: Where V is the collection point set, where 0 represents the garbage transfer station, 1, 2, 3, ..., n represents the garbage collection point; K is the required garbage collection vehicle set; d ij represents the actual distance between garbage collection points or between garbage collection points and garbage transfer stations, that is, the distance from node i to node j. ijk Defined using 0-1 integer decision variables: (6-2) Set the full coverage constraint to: Constraint 1: All collection vehicles must leave a collection point after arriving at the collection point: Constraint 2: Eliminate sub-loop constraints: In the formula, Q is a proper subset of the node set V and does not include the nodes of the transfer station; Constraint 3: All collection vehicles are required to start from the transfer station and return to the transfer station after completing the collection task: (6-3) Set the load constraint condition as: In the formula, q i is the amount of garbage collected at point i; Q k is the maximum load of the vehicle; (6-4) Set the maximum driving distance constraint as: Where L is the maximum restricted driving distance of the vehicle; (6-5) Input the garbage collection and transportation route optimization model established in steps (6-1), (6-2), (6-3), and (6-4); (6-6) Using daily garbage production data sets from three regions: high concentration, medium concentration, and dispersion; The newly created OD cost matrix tool in GIS software was used to obtain the distance set between each garbage collection point and the distance set between each garbage collection point and the garbage compression transfer station; Set the load capacity of collection vehicles and limit the driving distance; use genetic algorithm to solve and obtain the optimal collection and transportation path sets in high-concentration areas, medium-concentration areas and dispersed areas respectively. (6-7) Based on the assumption that the daily collection distance of each garbage collection truck does not exceed 20 km, different optimization paths in high-density areas, medium-density areas, and dispersed areas are combined and optimized to obtain the regional garbage collection and transportation optimization plan.

Images