A method for constructing a garbage classification model based on data analysis and machine learning

Abstract

The present application relates to the technical field of model construction, and particularly relates to a method for constructing a garbage classification model based on data analysis and machine learning. The method comprises the following steps: determining a garbage discharge area through a GIS visualization system, and discretely dividing the garbage discharge area into three-dimensional discharge units with height, area and adjacency relationship; then, for each three-dimensional discharge unit, analyzing the discharge concentration characteristics of garbage in the horizontal direction and the accumulation and diffusion characteristics of garbage in the vertical direction to determine the garbage discharge composition state; on this basis, establishing a corresponding relationship between the physical composition of garbage and the garbage classification ratio, and constructing a garbage classification prediction model; finally, using the model to classify and predict the garbage discharge area, and visually displaying the prediction results through the GIS system. The present application describes the garbage discharge and migration characteristics through three-dimensional GIS and time continuous analysis, and realizes the accurate prediction and visual management of garbage source classification by combining machine learning.

Description

Technical Field

[0001] This invention relates to the field of model building technology, and in particular to a method for building a garbage classification model based on data analysis and machine learning. Background Technology

[0002] Current waste sorting management mainly relies on manual inspections, fixed-point statistics, or data analysis based on two-dimensional images, which has limited ability to depict the spatial structure and dynamic changes of waste disposal areas. In actual disposal processes, waste usually exhibits obvious spatial accumulation characteristics and cross-regional migration phenomena. It is difficult to accurately reflect the true state of waste disposal based solely on planar information or data from a single time segment, which can easily lead to lags or errors in classification judgments.

[0003] Meanwhile, existing technologies for waste sorting modeling often rely on empirical rules or simple statistical features, lacking comprehensive analysis of key factors such as lateral diffusion, vertical accumulation, and migration constraints during waste disposal. This results in insufficient adaptability of the models to complex disposal scenarios. Furthermore, traditional waste sorting results are often presented in tabular or discrete data formats, lacking effective integration with spatial information and failing to provide intuitive support for refined management and decision-making. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for constructing a waste classification model based on data analysis and machine learning to solve at least one of the aforementioned technical problems.

[0005] To achieve the above objectives, a method for constructing a waste classification model based on data analysis and machine learning is provided, the method comprising the following steps:

[0006] Step S1: Confirm the waste disposal area through the preset GIS visualization system, and spatially discretize the waste disposal area, dividing the disposal area into several three-dimensional disposal units with spatial height, area and adjacency relationship;

[0007] Step S2: For each three-dimensional emission unit, analyze the lateral emission concentration characteristic data and vertical accumulation and diffusion characteristic data of the waste emission area to confirm the emission composition status of the waste emission area;

[0008] Step S3: Based on the emission composition status of the waste emission area, confirm the correspondence between the physical composition of waste and the waste classification ratio, and construct a waste classification prediction model;

[0009] Step S4: Use the waste classification prediction model to classify and predict the waste discharge area, output the classification prediction results of the waste source, and visualize them through the GIS visualization system.

[0010] The beneficial effects of this invention lie in its ability to integrate spatial and temporal characterization of the waste discharge process by incorporating 3D GIS visualization and continuous temporal modeling mechanisms into the waste discharge area. This enables refined perception and structured expression of waste discharge status, migration behavior, and accumulation characteristics. By constructing 3D discharge units with height, area, and adjacency relationships, and combining them with lateral discharge concentration characteristics, vertical accumulation and diffusion characteristics, and restricted discharge migration characteristics, the invention comprehensively reflects the spatial composition and dynamic evolution of waste during actual discharge, effectively avoiding the difficulty of traditional two-dimensional or static statistical methods in characterizing three-dimensional accumulation and cross-regional migration. Furthermore, through the analysis of internal and external spatial change trajectories within continuous time slices, the invention accurately identifies whether waste is in a state of continuous discharge and restricted migration, providing highly reliable feature inputs for modeling the relationship between waste physical composition and classification ratio. Finally, by combining machine learning to construct a waste classification prediction model, the invention achieves intelligent prediction and GIS visualization of waste source classification results, significantly improving the accuracy, stability, and interpretability of waste classification judgments. This provides reliable data support and technical assurance for refined waste management, source classification decisions, and regulatory enforcement. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the steps involved in constructing a waste sorting model based on data analysis and machine learning.

[0012] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2.

[0013] Figure 3 This is a time-period prediction trend chart for a waste classification model construction method based on data analysis and machine learning proposed in this application;

[0014] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0015] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0016] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0017] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0018] To achieve the above objectives, please refer to Figures 1 to 3 A method for constructing a waste classification model based on data analysis and machine learning, the method comprising the following steps:

[0019] Step S1: Confirm the waste disposal area through the preset GIS visualization system, and spatially discretize the waste disposal area, dividing the disposal area into several three-dimensional disposal units with spatial height, area and adjacency relationship;

[0020] In one embodiment, a 3D GIS visualization management system used for urban construction waste management is employed to load and display the target area as a whole. After the user logs in, the system enters a panoramic data overview interface, which overlays waste disposal point information onto a 3D GIS map based on building models, surface morphology, and road networks, thereby intuitively confirming the spatial distribution range of various waste disposal areas.

[0021] After confirming the waste disposal area, the spatial analysis and grid division functions built into the GIS system are used to perform spatial discretization processing on the waste disposal area: based on the geographic coordinates, the disposal area is divided in the horizontal direction according to the preset grid scale, and in the vertical direction, it is layered by combining building height, stacking height or terrain undulation information, so as to form a three-dimensional division result with spatial height parameters.

[0022] Furthermore, the corresponding floor area, height range, and adjacency relationship with adjacent spatial units are calculated for each divided spatial unit, and the above spatial attributes are uniformly encapsulated to generate several three-dimensional emission units with clear spatial height, area attributes, and adjacency topology relationships.

[0023] In one embodiment, the grid scale is determined based on the spatial size of the waste disposal area. First, the total area of ​​the waste disposal area is calculated, and then a corresponding horizontal division scale is selected according to different area ranges. For example:

[0024] When the area of ​​the emission zone is less than 5,000 square meters, the horizontal grid is divided into units of 5 meters by 5 meters.

[0025] When the area of ​​the emission zone is between 5,000 and 50,000 square meters, the horizontal grid is divided into units of 10 meters by 10 meters.

[0026] When the emission area is larger than 50,000 square meters, the horizontal grid is divided into units of 20 meters by 20 meters.

[0027] By using the above method, a finer grid is used in smaller emission areas, while a relatively larger grid scale is used in larger emission areas, in order to balance spatial accuracy and computational efficiency.

[0028] In the vertical direction, the stratification threshold is determined based on the height of the waste pile or the building height. First, the maximum pile height of the target area is obtained, and then stratification is performed according to the following rules:

[0029] When the maximum stacking height does not exceed three meters, divide it into layers of one meter each;

[0030] When the maximum stacking height is between three and ten meters, divide it into layers every two meters.

[0031] When the maximum stacking height exceeds ten meters, divide it into layers of five meters each.

[0032] For example, when the maximum stacking height of a construction waste dumping area is six meters, the vertical space is divided into three height levels: zero to two meters, two meters to four meters, and four meters to six meters, thus forming a three-dimensional discharge unit with height parameters.

[0033] To ensure consistency and comparability of spatial partitioning results generated across different regions and at different times, a unified urban basic grid system is established. Specifically, using a unified urban geographic coordinate benchmark as a spatial reference, a standard grid framework with a fixed starting point is pre-constructed throughout the entire city and expanded outwards according to a unified grid scale. When spatially discretizing all waste disposal areas, this unified grid framework is directly used for overlay and partitioning, without regenerating independent grids.

[0034] Meanwhile, each three-dimensional emission unit is assigned a unique grid number, which consists of a planar location number and a height level number. When the same location is updated at different times, it still corresponds to the same grid number, thus enabling direct comparison of waste accumulation changes over different time periods and achieving spatial alignment and dynamic change analysis.

[0035] In another embodiment, the footprint is the horizontal projected area of ​​the spatial unit on the ground plane. In actual calculations, the area is calculated based on the boundary range of the spatial unit in the geographic coordinate plane. For example, when the horizontal grid is divided into 10-meter by 10-meter sections, the footprint corresponding to each spatial unit is 100 square meters; when the spatial unit is located at the boundary of the emission area, its effective area is calculated based on the portion of its projection that actually falls within the emission area.

[0036] The height range is determined by the vertical layering results. Specifically, after completing the vertical layering, each layer corresponds to a fixed height range, such as zero to two meters, two meters to four meters, or four meters to six meters. The height range of a spatial unit is the height range corresponding to the layer in which that spatial unit is located. The lower boundary of this height range is the starting height of that layer, and the upper boundary is the ending height of that layer, thus forming a spatial height range with clearly defined upper and lower boundaries.

[0037] The adjacency relationship is determined by the spatial contact relationship between spatial units and is determined according to unified rules. Specifically, when two spatial units share a common boundary in the horizontal direction, they are determined to be horizontally adjacent; when two spatial units overlap vertically and their planar positions coincide, they are determined to be vertically adjacent; when two spatial units only contact each other at a corner point and do not form a boundary contact, they are not considered to be adjacent.

[0038] For example, at the same height level, a spatial unit can typically form horizontal adjacency relationships with a maximum of four spatial units to its east, west, south, and north; at different height levels, two spatial units located vertically on the same plane form vertical adjacency relationships. Complete spatial adjacency topologies can be automatically generated using these rules.

[0039] Step S2: For each three-dimensional emission unit, analyze the lateral emission concentration characteristic data and vertical accumulation and diffusion characteristic data of the waste emission area to confirm the emission composition status of the waste emission area;

[0040] In one embodiment, information on waste discharge points, vehicle parking locations, and historical discharge records can be recorded in the system database and displayed in a spatial layer overlay on the map interface. Statistical analysis of the spatial distribution density of waste within the same height layer can be performed to extract the distribution range, density, and continuity characteristics of waste in the horizontal direction, thereby forming corresponding lateral discharge concentration characteristic data to characterize the aggregation state of waste in a planar area.

[0041] Specifically, all three-dimensional emission units within the same height layer are divided into several statistical units. For each statistical unit, the number of waste emissions or the volume of waste emitted within a certain time period are recorded, and this data is divided by the area occupied by that statistical unit to obtain the emission intensity per unit area. When the emission intensity per unit area exceeds ten emission records per 100 square meters or five cubic meters of waste volume per 100 square meters, the spatial unit is determined to be a high-density emission area; when the emission intensity per unit area is between half and the above values, it is determined to be a medium-density emission area; and when the emission intensity per unit area is less than half of the above values, it is determined to be a low-density emission area. This forms data on the density of waste in the horizontal direction.

[0042] In terms of continuity analysis, the determination is based on the emission density distribution of adjacent three-dimensional emission units. When a spatial unit and at least two of its adjacent spatial units are simultaneously identified as high-density emission areas, the area is considered to form a continuous emission zone; when only a single spatial unit is a high-density emission area and the surrounding adjacent units are medium-density or low-density areas, the area is considered to be a local concentrated emission point.

[0043] In vertical accumulation and diffusion analysis, the proportion of waste occupying the same plane location at different height layers is statistically analyzed. Specifically, the actual volume occupied by waste in each height layer is compared with the total volume of the corresponding spatial unit. When the proportion of waste occupying the upper layer exceeds one-third of the proportion occupied by the lower layer, a clear vertical accumulation trend is identified. When multiple adjacent height layers have a waste occupancy rate exceeding half, it is identified as a high accumulation state. When the occupancy rate of the upper layer gradually decreases while the occupancy rate of the lower layers in adjacent spatial units gradually increases, it is identified as a vertical diffusion state.

[0044] After obtaining data on the lateral concentration characteristics and the vertical accumulation and diffusion characteristics, the emission composition of the waste discharge area is determined based on the combination relationship between the two types of characteristics. For example, when the lateral density is high and the vertical occupancy is mainly concentrated in the lowest height layer, it is determined to be a concentrated flat type; when the lateral density is moderate but the vertical occupancy of multiple height layers continues to increase, it is determined to be a local high-pile type; when the lateral density gradually decreases and adjacent spatial units show continuous diffusion characteristics, it is determined to be a diffusion and spread type.

[0045] Simultaneously, by combining the height parameters of the three-dimensional emission unit and the information on the changes in the height of the waste pile at different times, the accumulation and diffusion of waste in the vertical direction are analyzed. By comparing the proportion of waste occupying different height layers in the same three-dimensional emission unit, the accumulation trend of waste from low to high layers and the diffusion characteristics from local high piles to surrounding height layers are identified, thereby generating vertical accumulation and diffusion characteristic data to reflect the evolution state of waste in the vertical direction.

[0046] Based on this, the lateral emission concentration characteristic data and the vertical accumulation and diffusion characteristic data are comprehensively correlated and analyzed. According to the coupling relationship between the two in the same three-dimensional emission unit, the emission composition state of the corresponding waste emission area is confirmed. The emission composition state is used to characterize whether the waste in the area is mainly concentrated and flat, locally piled up, or diffuse and spread.

[0047] In one embodiment, regarding lateral distribution, the degree of waste concentration within a planar area is determined by calculating the waste emission intensity per unit area and the number of consecutive high-density spatial units. For example, when the emission intensity per unit area exceeds fifteen emission records per 100 square meters and there are at least three adjacent consecutive high-density spatial units, it is determined to be a laterally highly concentrated state; when the emission intensity per unit area is between five and fifteen times per 100 square meters and the number of consecutive high-density spatial units is less than three, it is determined to be a moderately concentrated state; and when the emission intensity per unit area is less than five times per 100 square meters, it is determined to be a low-concentration state.

[0048] Regarding vertical stacking, the height of the waste stack is obtained through three-dimensional measurement data, and the average stacking height and maximum stacking height are calculated. When the maximum stacking height does not exceed two meters and most of the waste is distributed in the lowest stacking layer, it is determined to be a low stacking state; when the maximum stacking height is between two and five meters and multiple stacking layers have significant waste occupying volume, it is determined to be a medium stacking state; when the maximum stacking height exceeds five meters and the volume of waste in the upper layer accounts for more than one-third of the total waste volume, it is determined to be a high stacking state.

[0049] Regarding the degree of diffusion, it is determined by the changes in emission density between the central three-dimensional emission unit and its surrounding adjacent units over a continuous time period. When the emission density of the central unit continuously decreases while the emission density of at least three adjacent units continuously increases, and the surrounding units reach a medium-density level or above, it is determined that there is a significant diffusion trend; when waste emissions are mainly concentrated in a single spatial unit and the density changes of the surrounding units are small, it is determined that there is no significant diffusion.

[0050] After obtaining the horizontal concentration state, vertical accumulation state, and diffusion trend, the composition state of waste discharge is confirmed through a combination judgment logic, with the specific rules as follows:

[0051] When the horizontal concentration is high, the vertical stacking height is low, and the diffusion is weak, it is judged as a concentrated flat type;

[0052] When the lateral concentration is moderate or high, and the vertical stacking height reaches a moderate or high level, it is judged as a local high stacking type.

[0053] When the degree of horizontal concentration gradually decreases and the surrounding units show a clear trend of diffusion, it is determined to be a diffusion and spread type.

[0054] For example, in a certain waste disposal area, when the emission intensity per unit area is 20 emission records per 100 square meters, there are four consecutive high-density spatial units and the maximum stacking height is only one meter, the area is judged as a concentrated flat type; when the emission intensity per unit area is 12 emission records per 100 square meters and the maximum stacking height reaches six meters, it is judged as a local high-stacking type; when the emission density in the central area gradually decreases while the emission density in multiple surrounding units continues to increase, it is judged as a diffusion and spread type.

[0055] In another embodiment, for example, within a statistical period, if a three-dimensional emission unit records twenty waste emissions and the unit occupies an area of ​​one hundred square meters, then its emission intensity per unit area is twenty emission records per one hundred square meters. When the emission intensity per unit area exceeds fifteen emission records per one hundred square meters, the area is determined to be a high-density emission area; when the emission intensity per unit area is between five and fifteen times per one hundred square meters, it is determined to be a medium-density emission area; when the emission intensity per unit area is less than five times per one hundred square meters, it is determined to be a low-density emission area, thereby forming lateral emission concentration characteristic data.

[0056] The vertical accumulation characteristics are calculated using stacking height measurement data obtained from a 3D GIS system. Specifically, the stacking height data can originate from on-site 3D laser scanning, UAV oblique photogrammetry, or a 3D point cloud model generated by a fixed monitoring camera device. By extracting the surface height of the waste pile and comparing it with the ground reference height, the waste accumulation height of each 3D discharge unit is obtained. When waste volume is detected in multiple consecutive height layers at the same planar location, the degree of accumulation is calculated based on the proportion of volume occupied by each height layer. For example, when the volume of the upper layer of waste reaches more than half of the volume of the lower layer of waste, it is determined that there is significant vertical accumulation in that area.

[0057] The degree of diffusion is calculated by the changes in waste distribution between adjacent three-dimensional emission units. Specifically, the waste emission density of a certain three-dimensional emission unit and its adjacent units is compared over multiple consecutive time periods. When the emission density of the central unit gradually decreases while the emission density of the surrounding adjacent units continues to increase, and the emission density of at least three adjacent units exceeds the medium density threshold, it is determined that there is a diffusion and spread trend in the area; when the high density is maintained only in a single spatial unit and the emission density of the surrounding units is low, it is determined to be a localized concentrated emission state.

[0058] After obtaining data on the lateral emission concentration characteristics, vertical accumulation characteristics, and diffusion degree, the emission composition of the waste discharge area is confirmed through comprehensive analysis of these three types of data. For example, when the lateral density is high and the accumulation height is low, it is identified as a concentrated flat type; when the lateral density is medium but the accumulation height continues to increase, it is identified as a locally high-pile type; and when the lateral density expands to surrounding units and the diffusion degree continues to rise, it is identified as a diffusion and spread type.

[0059] Step S3: Based on the emission composition status of the waste emission area, confirm the correspondence between the physical composition of waste and the waste classification ratio, and construct a waste classification prediction model;

[0060] In one embodiment, based on the differences in the horizontal concentration characteristics and vertical accumulation and diffusion characteristics of waste under different emission composition states, and combined with on-site sampling records, historical disposal data and image recognition results, physical composition characteristic data of waste presented in the three-dimensional emission unit are extracted. The physical composition characteristics include, but are not limited to, the proportion of blocky waste, the proportion of loose particulate waste, the proportion of recyclable materials, and the proportion of mixed impurities.

[0061] Subsequently, the physical composition characteristics of the waste were correlated with the waste classification statistics within the corresponding time period to establish a correspondence between the physical composition of waste and the waste classification ratio. Specifically, by comparing the changes in the proportion of various physical composition characteristics in the classification results under different emission composition states, the influence pattern of emission composition states on the waste classification ratio was identified, thereby forming a standardized feature-ratio mapping relationship dataset.

[0062] Based on this, a waste sorting prediction model is constructed using the emission composition status and physical composition characteristics of waste as input features, and the waste sorting ratio as the output target. This prediction model is used to predict the sorting ratio of various types of waste within a corresponding waste emission area, given the known waste emission composition status.

[0063] In another embodiment, the construction process of the waste sorting prediction model includes stages such as feature construction, sample generation, model training, and model validation, as follows: First, based on the waste emission composition state obtained in step S2, feature vectorization processing is performed on each three-dimensional emission unit. Specifically, the lateral emission concentration feature data is quantized into waste density per unit area, lateral expansion radius, and concentration index; the vertical accumulation and diffusion feature data is quantized into maximum accumulation height, average accumulation height, and height change gradient; and the quantization results are encoded with the corresponding emission composition state and combined to form the input feature vector of the three-dimensional emission unit.

[0064] This involves counting the number of waste discharges or the volume of waste within a three-dimensional emission unit during a specific statistical period, and then converting the data based on the unit's floor area. For example, if twenty waste discharges are recorded within a 100-square-meter area, the waste density of that spatial unit is 20 discharge records per 100 square meters. To facilitate model training, the values ​​can be further quantified according to density ranges: a value of one is assigned when the density is below five discharge records per 100 square meters, a value of two is assigned when the density is between five and fifteen discharge records per 100 square meters, and a value of three is assigned when the density is above fifteen discharge records per 100 square meters.

[0065] Secondly, the lateral expansion radius is calculated. First, a set of spatial units with continuous waste discharge records within the same height layer is identified. The unit with the highest discharge density is taken as the central unit, and the maximum horizontal distance between this central unit and the boundary of the outermost continuous discharge unit is calculated. This distance is the lateral expansion range. For example, when the continuous waste distribution range covers an area with a diameter of approximately thirty meters, the lateral expansion radius is approximately fifteen meters. To facilitate input into the model, the expansion range can be quantified and graded according to its size; for example, less than ten meters is classified as Level 1, ten to thirty meters as Level 2, and greater than thirty meters as Level 3.

[0066] Next, the concentration index is calculated. The number of consecutive high-density spatial units is counted, and this number is compared with the total number of spatial units to reflect the degree of concentration of waste distribution. For example, if there are ten spatial units in an area, and six of them are high-density spatial units, then the concentration level is 60%. Further quantification is performed according to the concentration ratio; for example, below 30% is classified as Level 1, 30% to 60% as Level 2, and above 60% as Level 3.

[0067] Regarding vertical features, the maximum accumulation height is calculated first. The surface height of the waste pile is obtained through 3D point cloud measurement or image reconstruction and compared with a ground reference height to determine the maximum accumulation height. For example, if the highest measured surface height of the waste pile is four meters above the ground, the maximum accumulation height is four meters. The average accumulation height is then calculated by averaging the height values ​​of all spatial units containing waste.

[0068] Furthermore, the height change gradient is calculated using the rate of change in waste volume between adjacent height layers. Specifically, the ratio of the volume difference between two adjacent height layers to the volume of waste in the lower layer is used to reflect the vertical trend of waste change. For example, when the volume of waste in the second height layer is half that of the first height layer, the change ratio is 50%. Based on the magnitude of the change ratio, it can be divided into three levels: slow change, moderate change, and rapid change.

[0069] Based on historical waste disposal records and classified weighing data, waste classification ratio data corresponding to the three-dimensional emission unit in time and space is obtained. The waste classification ratio data includes the proportion information of recyclable waste, recyclable waste and other waste. The waste classification ratio data is used as a supervision label and together with the corresponding input feature vector constitutes the model training sample set.

[0070] Then, during the model training phase, a modeling approach combining multivariate regression and classification prediction is employed to fit the mapping relationship between the input feature vector and the waste classification ratio. Through multiple rounds of iterative learning on the training sample set, the model parameters are gradually adjusted to minimize the error between the model's output waste classification ratio and the actual classification ratio, thereby obtaining the initial waste classification prediction model.

[0071] Finally, in the model validation stage, historical three-dimensional emission unit data that was not used in training is input into the waste classification prediction model, and the model output results are compared and analyzed with the actual waste classification ratio. When the prediction error is lower than the preset threshold, the model construction is confirmed to be complete. When the prediction error is higher than the preset threshold, the feature weights or sample division method are readjusted, and the above training process is repeated to obtain a stable and reliable waste classification prediction model.

[0072] Step S4: Use the waste classification prediction model to classify and predict the waste discharge area, output the classification prediction results of the waste source, and visualize them through the GIS visualization system.

[0073] In one embodiment, each three-dimensional emission unit obtained in step S1 is used as the prediction object. The horizontal emission concentration feature data, vertical accumulation and diffusion feature data, and emission composition status information of the corresponding three-dimensional emission unit are collected in real time or periodically, and the above data are transmitted into the waste classification prediction model according to the model input format.

[0074] The waste sorting prediction model, based on the input three-dimensional emission unit features, outputs the predicted waste source sorting results for the corresponding waste emission area within the current time period. These prediction results include the predicted sorting ratios of various types of waste at the source emission stage and the dominant waste type. The model output is further bound to the spatial identifiers of the three-dimensional emission units, forming waste sorting prediction data with clear spatial orientation.

[0075] Subsequently, the predicted waste sorting results are transmitted back to the GIS visualization system and displayed visually through layer overlay and attribute mapping. Specifically, different waste categories or sorting ratio ranges are presented in the 3D GIS scene using different colors, transparency, or height indicators, making the source sorting status of each waste discharge area spatially identifiable. At the same time, users can interactively view the predicted sorting ratio, historical trends, and comparative analysis results for any 3D discharge unit.

[0076] Specifically, the waste sorting prediction model is constructed using a multi-layer prediction structure. The model consists of an input feature layer, a feature mapping layer, a proportion prediction layer, and a result determination layer. The input feature layer receives feature data from three-dimensional emission units, including waste density level, lateral expansion range level, concentration level, maximum accumulation height level, average accumulation height level, height variation level, and emission composition status code. These features are presented as a feature sequence in numerical form and used as model input.

[0077] In the feature mapping stage, the input feature sequences are weighted and combined. Multiple sets of weight coefficients are used to adjust the importance of different features, enabling the model to highlight the feature factors that have a significant impact on the waste distribution structure. Specifically, the model first performs numerical normalization on each input feature to ensure that data of different dimensions are within a uniform numerical range. Then, through multiple rounds of weighted calculations, a comprehensive feature value is formed to characterize the overall emission structure characteristics of the three-dimensional emission unit.

[0078] In the proportion prediction phase, the model calculates the predicted proportions of various types of waste at the source emission stage based on comprehensive feature values. Specifically, the model calculates the predicted proportions of recyclable waste, resource-recoverable waste, and other waste separately, and performs proportion normalization to ensure that the sum of the proportions of the three types of waste is equal to one, thus obtaining a complete waste classification prediction result. At the same time, the dominant waste type in the area is determined based on the maximum value among the predicted proportions of the three types of waste.

[0079] During the model training phase, an iterative training approach based on error minimization is employed for parameter optimization. The training samples consist of historical 3D emission unit feature data and corresponding actual waste classification ratio data. In each training iteration, the model calculates the predicted ratio based on the input features and compares it with the actual classification ratio, calculating the difference between the two. Subsequently, the ratio differences for all samples are squared and averaged to obtain the overall prediction error. The model then gradually adjusts the feature weights based on this error value, causing the predicted ratio to progressively approach the true classification ratio.

[0080] During model iteration, when the overall prediction error decreases below a preset range after multiple training rounds, the model is considered to have reached a stable state and training is considered complete. To ensure model reliability, independent historical data is used for further verification. Data from three-dimensional emission units not used in training are input into the model, and the average error between the predicted and actual proportions is calculated. When the average error is below a set allowable range, the waste sorting prediction model training is confirmed to be complete.

[0081] In another embodiment, reference Figure 3 Using the changes in emission characteristics of waste discharge areas at different time periods as predictive input, and combining the time-dimensional statistical results shown in the figure, the waste emission data at 08:00, 10:00, 12:00, 14:00, 16:00, 18:00, and 20:00 are modeled and analyzed. Specifically, the waste classification prediction model jointly discriminates the emission intensity, growth trend, and peak changes of concrete, brick and ceramic, metal, wood, and mixed waste at each time node, thereby outputting the source classification prediction results of the corresponding waste discharge areas at different time periods. The prediction results are further integrated into a GIS visualization system. Through time-space linkage, the distribution status and changing trends of various types of waste in the waste discharge area are visualized in the form of curves or layers, enabling managers to intuitively grasp the emission patterns of different types of waste throughout the day, providing a basis for subsequent classification, collection, and scheduling.

[0082] As an example of the present invention, reference is made to Figure 2 As shown, in this example, step S2 includes:

[0083] Step S21: For each three-dimensional emission unit, analyze the waste distribution pattern in the same height layer of the waste emission area to obtain the lateral emission concentration characteristic data of the horizontal expansion range and aggregation degree of waste;

[0084] Step S22: Based on the three-dimensional emission unit along the three-dimensional GIS spatial height direction, perform layered analysis on the changes in waste accumulation in different height layers of the waste emission area, extract the accumulation and diffusion characteristics of waste in the vertical direction, and generate vertical accumulation and diffusion feature data;

[0085] Step S23: During the continuous emission process, analyze the changes in the migration path of the three-dimensional emission unit in the horizontal and vertical directions, identify the migration constraints of the waste accumulation process, and obtain emission migration restriction characteristic data;

[0086] Step S24: Confirm the composition status of waste emissions based on the horizontal emission concentration characteristic data, vertical accumulation and diffusion characteristic data, and emission migration restriction characteristic data.

[0087] In one embodiment, for each three-dimensional emission unit, the same spatial height layer is selected in the three-dimensional GIS visualization system, and the spatial distribution pattern of waste within that height layer is analyzed. By statistically analyzing the coverage area, distribution density, and continuous distribution area of ​​waste within that height layer, the horizontal expansion boundary and aggregation degree of waste are calculated, thereby obtaining lateral emission concentration characteristic data characterizing the planar diffusion and concentration level of waste. This lateral emission concentration characteristic data reflects whether waste is concentrated or dispersed within the same height layer.

[0088] Based on the height division results of the three-dimensional emission units in the three-dimensional GIS space, the waste accumulation situation in different height layers of the waste emission area is analyzed in a hierarchical manner along the vertical direction. By comparing the proportion of waste and the changes in accumulation height in adjacent height layers, the accumulation trend of waste from low to high layers and the diffusion characteristics from local high piles to surrounding height layers are identified, thereby extracting the vertical accumulation and diffusion characteristics of waste and generating corresponding vertical accumulation and diffusion feature data.

[0089] During continuous waste disposal, the migration paths of waste in the horizontal and vertical directions are compared and analyzed by combining the state changes of three-dimensional disposal units at different times. By identifying the displacement direction, displacement amplitude, and restricted area of ​​the waste accumulation center of gravity, it is determined whether the waste is constrained by terrain boundaries, building structures, or spatial capacity during the diffusion process, thereby identifying the migration constraint characteristics during the waste accumulation process and obtaining emission migration constraint characteristic data.

[0090] After obtaining data on the lateral emission concentration characteristics, vertical accumulation and diffusion characteristics, and emission migration restriction characteristics, a comprehensive analysis of these multidimensional characteristics is conducted. Based on the combination relationships between these characteristics, the emission composition status of the waste emission area is confirmed. This emission composition status characterizes whether waste emission is mainly of planar concentration, vertical accumulation, restricted diffusion, or mixed evolution type, providing a basic basis for subsequent waste classification prediction and management strategy formulation.

[0091] Preferred methods for confirming continuous emission processes include:

[0092] The waste discharge area is divided into three-dimensional spaces at preset time intervals to ensure that the same spatial location corresponds to the same three-dimensional discharge unit in different time slices, thereby constructing a three-dimensional discharge unit sequence with temporal continuity.

[0093] For the three-dimensional emission unit sequence, the change trajectories of the three-dimensional emission units in the horizontal coverage, vertical stacking height and spatial morphological boundary of each time slice are extracted to confirm the internal spatial change trajectory of the three-dimensional emission units.

[0094] Based on the spatial adjacency relationship between adjacent three-dimensional emission units, we analyze whether waste migrates across units within a continuous time slice, and determine the migration direction and migration range to confirm the external spatial change trajectory of the three-dimensional emission units.

[0095] By observing the continuity of the internal and external spatial change trajectories, it can be determined whether the three-dimensional emission unit is in a continuous emission process.

[0096] In one embodiment, the waste disposal area is periodically divided into three-dimensional spaces at preset time intervals. The time interval can be set according to the waste disposal frequency, for example, updated hourly or daily. Within each time slice, the same spatial grid division rule is used to perform three-dimensional discretization of the waste disposal area, thereby ensuring that the same geographic location always corresponds to the same three-dimensional disposal unit in different time slices, thus constructing a three-dimensional disposal unit sequence with temporal continuity.

[0097] Subsequently, for the three-dimensional emission unit sequence, the spatial state changes of a single three-dimensional emission unit within each time slice are extracted. Specifically, the changes in the horizontal waste coverage area, the vertical waste accumulation height, and the expansion or contraction of the waste spatial boundary of the three-dimensional emission unit within each time slice are statistically analyzed. These changes are recorded in chronological order to form an internal spatial change trajectory reflecting the evolution of waste within a single three-dimensional emission unit.

[0098] Based on this, and combined with the pre-established spatial adjacency relationships between three-dimensional emission units, the state changes of adjacent three-dimensional emission units within continuous time slices are compared and analyzed. By identifying whether waste diffuses or transfers from the current three-dimensional emission unit to adjacent three-dimensional emission units, and quantifying the direction and magnitude of migration, an external spatial change trajectory describing the diffusion behavior of waste between different three-dimensional emission units is formed.

[0099] Finally, the internal space change trajectory and the external space change trajectory are jointly analyzed. When the two show continuous, stable and consistent change characteristics in the time series, the corresponding three-dimensional emission unit is determined to be in a continuous emission process; when the change trajectory shows discontinuous or random fluctuation characteristics, it is determined to be a non-continuous emission process.

[0100] Preferably, based on the spatial adjacency relationship between adjacent three-dimensional emission units, the analysis of whether waste migrates across units within a continuous time slice includes:

[0101] Based on the three-dimensional spatial partitioning results, spatial adjacency relationships between each three-dimensional emission unit are established. These spatial adjacency relationships include at least horizontal, vertical, and diagonal adjacency relationships, forming a corresponding set of three-dimensional emission unit adjacency relationships.

[0102] Within a continuous time slice, the waste occupancy status of each three-dimensional emission unit in the same waste emission area is compared hourly to identify the characteristics of the waste occupancy status shifting from the three-dimensional emission unit to the spatially adjacent three-dimensional emission unit.

[0103] When the waste occupancy state is weakened by the three-dimensional emission unit in adjacent time slices and synchronously strengthened in at least one spatially adjacent three-dimensional emission unit, waste migration behavior across three-dimensional emission units is confirmed.

[0104] In one embodiment, based on the three-dimensional spatial division of the waste disposal area, the spatial positional relationships between each three-dimensional disposal unit are analyzed, and a corresponding spatial adjacency model is established. Specifically, according to the three-dimensional spatial topology, three-dimensional disposal units sharing the same height layer and having adjacent boundaries are defined as horizontally adjacent; three-dimensional disposal units located within the same plane projection range and having adjacent height layers are defined as vertically adjacent; and three-dimensional disposal units that simultaneously satisfy partial horizontal overlap and adjacent height layers in space are defined as obliquely adjacent. Based on the above classification, a set of three-dimensional disposal unit adjacency relationships covering all three-dimensional disposal units is formed for subsequent migration path analysis.

[0105] Subsequently, within consecutive time slices, the waste occupancy status of each three-dimensional emission unit in the same waste emission area is compared and analyzed hourly. Specifically, the changes in waste occupancy ratio, accumulation volume, or occupancy intensity of each three-dimensional emission unit within each time slice are statistically analyzed. Combined with the adjacency relationship set, it is identified whether the waste occupancy status has shifted from a certain three-dimensional emission unit to its spatially adjacent three-dimensional emission unit, thereby extracting the temporal characteristics of waste changes across units.

[0106] Furthermore, when a significant weakening trend in the waste occupancy state of a certain three-dimensional emission unit is detected within adjacent time slices, and a simultaneous increase in waste occupancy state occurs in at least one spatially adjacent three-dimensional emission unit, waste migration behavior across three-dimensional emission units is confirmed. This migration behavior can be further determined by combining the type of adjacency relationship, classifying it as horizontal migration, vertical migration, or diagonal migration.

[0107] Preferably, the hourly comparison of waste occupancy in each three-dimensional emission unit within the same waste emission area also includes:

[0108] In adjacent time slices, the spatial coverage, height occupied, and boundary changes of waste within the same three-dimensional emission unit are compared accordingly.

[0109] Simultaneously, a comparative analysis was conducted on the changes in waste occupancy of adjacent three-dimensional emission units that have a spatial adjacency with the three-dimensional emission unit;

[0110] Based on the correspondence between the changes in waste occupancy in three-dimensional emission units and the changes in waste occupancy in spatially adjacent units within the time slices, the transfer characteristics of waste occupancy status between different three-dimensional emission units are identified.

[0111] In one embodiment, within adjacent time slices, for the same three-dimensional emission unit, waste spatial state parameters are extracted from the preceding and following time slices and compared accordingly. These spatial state parameters include changes in the spatial coverage of waste within the three-dimensional emission unit, changes in the height occupied by waste accumulation, and the expansion or contraction of the waste spatial boundary. By comparing these parameters item by item, intra-unit variation characteristics reflecting the changing trend of waste occupancy within the three-dimensional emission unit are formed.

[0112] Secondly, adjacent three-dimensional emission units that are spatially adjacent to the aforementioned three-dimensional emission unit are selected simultaneously, and their waste occupancy changes within adjacent time slices are compared and analyzed. Specifically, the waste occupancy ratio, accumulation height change, and spatial distribution range change of each adjacent three-dimensional emission unit within the preceding and following time slices are statistically analyzed, and the results are time-aligned with the intra-unit change characteristics of the target three-dimensional emission unit.

[0113] Finally, based on the correspondence between the changes in waste occupancy of the target 3D emission unit and the changes in waste occupancy of adjacent units within the preceding and following time slices, it is determined whether the waste occupancy state has shifted from the target 3D emission unit to an adjacent 3D emission unit. When the waste occupancy of the target 3D emission unit shows a decreasing trend and the waste occupancy of at least one adjacent 3D emission unit shows an increasing trend, it is determined that there are characteristics of waste occupancy state shifting between different 3D emission units, thus providing a basis for confirming waste cross-unit migration behavior.

[0114] Preferably, step S23 includes the following steps:

[0115] Step S231: Within a continuous time slice, establish a time series relationship for the three-dimensional emission units corresponding to the same waste emission area;

[0116] Step S232: Based on the three-dimensional emission unit sequence, extract the cross-unit migration path of waste in the horizontal direction and the accumulation migration path in the vertical direction;

[0117] Step S233: Identify whether the waste migration is concentrated in a preset spatial direction or within a limited spatial range based on the horizontal and vertical migration paths;

[0118] Step S234: By using the directional stability features or range-limited features of the waste migration path within a continuous time slice, confirm that there are migration constraints in the waste accumulation process, and generate corresponding emission migration-limited feature data accordingly.

[0119] In one embodiment, for the same waste disposal area, the corresponding three-dimensional disposal unit status information is acquired at preset time intervals, and the spatial identifier of the three-dimensional disposal unit is used as an index to establish a time sequence relationship between the three-dimensional disposal units in each time slice. In this way, the waste occupancy status of the same spatial location in different time slices can be correlated in chronological order to form a three-dimensional disposal unit time sequence reflecting the waste disposal evolution process.

[0120] Based on the time series of the three-dimensional emission units, the migration paths of waste in different directions are extracted. Specifically, by analyzing the change process of waste occupancy from one three-dimensional emission unit to its horizontally adjacent three-dimensional emission unit within adjacent time slices, the cross-unit migration path of waste in the horizontal direction is extracted; simultaneously, by analyzing the change process of waste occupancy between three-dimensional emission units at the same plane position but different height layers, the accumulation and migration path of waste in the vertical direction is extracted, thus forming horizontal migration path data and vertical migration path data respectively.

[0121] After obtaining the horizontal and vertical migration paths, a comprehensive analysis of the two types of migration paths is conducted to determine whether waste migration is concentrated in a predetermined spatial direction or within a limited spatial range. Specifically, by statistically analyzing the dominant directional distribution and path coverage of the migration paths, it is possible to identify whether waste migration is consistently concentrated in a fixed direction, a combination of a few directions, or a limited set of adjacent units, thus reflecting whether there are obvious directional or spatial limitations in the waste migration process.

[0122] Based on the directional stability or range-limited characteristics exhibited by the migration path within continuous time slices, migration constraints are confirmed in the waste accumulation process. These migration constraints are then quantified to generate corresponding emission migration-limited feature data. This emission migration-limited feature data is used to characterize the migration restriction state of waste during the emission process caused by topographic conditions, spatial structure, or external environmental factors.

[0123] Preferably, identifying whether waste migration is concentrated in a preset spatial direction or within a limited spatial range based on the horizontal and vertical migration paths includes:

[0124] Based on the spatial coordinate relationship of the three-dimensional emission unit, the migration path of waste in the horizontal direction is classified to determine the main horizontal spatial direction of waste migration in a continuous time slice.

[0125] Along the height direction of the three-dimensional emission unit, the migration path of waste in the vertical direction is statistically analyzed to determine the main height range that changes during the waste accumulation process;

[0126] By jointly analyzing the directional classification results of horizontal migration paths and the height range results of vertical migration paths, it can be identified whether waste migration is mainly concentrated in a preset combination of spatial directions or a limited spatial height range.

[0127] When waste migration is continuously concentrated in a preset combination of spatial directions or a limited spatial height range within a continuous time slice, it is confirmed that waste migration has spatially restricted characteristics.

[0128] In one embodiment, the horizontal migration paths of waste across units are categorized based on the coordinate relationships of each three-dimensional emission unit in a three-dimensional GIS space. Specifically, according to the relative positional relationship between the starting three-dimensional emission unit and the target three-dimensional emission unit in planar coordinates, the horizontal migration paths are divided into multiple preset spatial direction categories, such as along the main road, along the site boundary, or along a specific axial direction. The frequency of occurrence of each direction category within a continuous time slice is then counted to determine the main horizontal spatial direction of waste migration.

[0129] Secondly, a hierarchical statistical analysis of the vertical migration path of waste was conducted along the height direction of the three-dimensional emission unit. By statistically analyzing the frequency and magnitude of changes in waste occupancy between three-dimensional emission units at different height levels, the height range where changes are most concentrated during waste accumulation was identified, thereby determining the main height range in which waste migrates vertically.

[0130] Based on this, the directional classification results of horizontal migration paths and the height range results of vertical migration paths are jointly analyzed. Specifically, by constructing a directional-height joint distribution relationship, it is possible to identify whether waste migration is mainly concentrated within a preset spatial directional combination or a limited spatial height range, in order to determine the spatial concentration and restriction characteristics of waste migration.

[0131] Finally, when it is detected that waste migration is continuously concentrated in the preset spatial direction combination or the limited spatial height range within a continuous time slice, it is confirmed that waste migration has spatially restricted characteristics.

[0132] Preferably, the directional classification of waste migration paths in the horizontal direction specifically involves:

[0133] Based on the spatial relative position of the three-dimensional emission units in the horizontal plane, the migration direction of waste from the three-dimensional emission units to the spatially adjacent three-dimensional emission units within a continuous time slice is marked.

[0134] The migration direction is divided into several horizontal migration direction categories according to its orientation relative to the three-dimensional emission unit;

[0135] By statistically analyzing the migration occurrence of each horizontal migration direction category within a continuous time slice, the main horizontal migration direction of waste can be determined.

[0136] In one embodiment, based on the spatial coordinate relationship of the three-dimensional emission units in the horizontal plane, the migration behavior of waste from the target three-dimensional emission unit to its spatially adjacent three-dimensional emission units within a continuous time slice is directionally marked. Specifically, according to the relative orientation relationship between the migration initiation unit and the migration target unit in the planar coordinates, a corresponding directional identifier is assigned to each migration behavior to reflect the direction of movement of waste in horizontal space.

[0137] Secondly, the migration directions are categorized according to their orientation relative to the three-dimensional emission unit. Specifically, with the center point of the three-dimensional emission unit as a reference, the horizontal migration directions are divided into several preset direction categories, such as forward, backward, left, right, or combinations thereof, thereby forming a standardized set of horizontal migration direction categories to uniformly describe the directional characteristics of different migration behaviors.

[0138] Finally, within consecutive time slices, statistical analysis was performed on the migration occurrences corresponding to each horizontal migration direction category. By comparing the migration frequency, duration, or cumulative migration magnitude of different direction categories, the main migration directions in which waste migration is most concentrated in the horizontal direction were identified, providing a basis for subsequent identification of waste migration space-constrained features and migration constraint analysis.

[0139] Preferably, based on the correspondence between the changes in waste occupancy of three-dimensional emission units and the changes in waste occupancy of spatially adjacent units within different time slices, the characteristics of waste occupancy transfer between different three-dimensional emission units include:

[0140] Within adjacent time slices, the changes in the waste occupancy state in the three-dimensional emission unit are analyzed to determine the direction of change in the degree of waste occupancy.

[0141] Simultaneously analyze the changes in the waste occupancy state in at least one adjacent three-dimensional emission unit that has a spatial adjacency relationship with the three-dimensional emission unit, and determine the corresponding direction of change;

[0142] The direction of change in the occupancy of waste in a three-dimensional emission unit is compared with that of adjacent three-dimensional emission units to identify whether there is a corresponding relationship where the occupancy of a three-dimensional emission unit weakens and the occupancy of adjacent three-dimensional emission units strengthens.

[0143] When the corresponding change relationship is repeated within a continuous time slice, it is confirmed that the waste occupancy state forms a stable transfer characteristic between different three-dimensional emission units.

[0144] In one embodiment, within adjacent time slices, the changes in the waste occupancy state of the target three-dimensional emission unit are statistically analyzed. The waste occupancy state can be characterized by parameters such as occupancy volume, occupancy ratio, or accumulation height. By comparing the previous time slice with the next time slice, it is determined whether the degree of waste occupancy shows an increasing trend, a decreasing trend, or is basically stable, thereby clarifying the direction of change in the degree of waste occupancy.

[0145] Secondly, at least one adjacent three-dimensional emission unit with spatial adjacency to the target three-dimensional emission unit is simultaneously selected, and its waste occupancy status changes within adjacent time slices are analyzed. By adopting the same occupancy status representation method as the target three-dimensional emission unit, the waste occupancy change direction of each adjacent three-dimensional emission unit is determined, and corresponding adjacent unit change direction data is generated.

[0146] Based on this, the direction of waste occupancy change of the target three-dimensional emission unit is compared with the direction of waste occupancy change of each adjacent three-dimensional emission unit. When a weakening trend of waste occupancy of the target three-dimensional emission unit is detected, and an increasing trend of waste occupancy of at least one adjacent three-dimensional emission unit is detected simultaneously within the same time slice, it is identified as a corresponding change relationship where the occupancy of the three-dimensional emission unit weakens and the occupancy of the adjacent three-dimensional emission unit increases.

[0147] Finally, when the corresponding change relationship repeats itself across multiple consecutive time slices, it is confirmed that the waste occupancy state forms a stable transfer characteristic between different three-dimensional emission units. This stable transfer characteristic is used to characterize the continuous spatial migration direction and path of the waste.

[0148] Preferably, when the corresponding change relationship repeats within a continuous time slice, specifically:

[0149] Within a continuous period of no less than two and no more than ten time slices, the volume occupied by waste in the three-dimensional emission unit shows a continuous decrease of 5% to 40%, while the volume occupied by waste in at least one spatially adjacent three-dimensional emission unit shows a corresponding increase of 5% to 40% within the same time slice, wherein the volume occupied by waste is measured in cubic meters.

[0150] Furthermore, when the direction of volume change remains consistent across consecutive time slices, the corresponding change relationship is determined to be stable and repetitive, thus confirming that the waste occupancy state forms an effective transfer characteristic between different three-dimensional emission units.

[0151] In one embodiment, to avoid occasional fluctuations interfering with the determination of waste migration, a clear quantitative threshold and time constraint are set for the changes in waste occupancy between three-dimensional emission units to confirm whether the waste occupancy state forms a stable and effective transfer characteristic, as follows:

[0152] During continuous monitoring, waste-occupied volume data of each three-dimensional emission unit within the same waste emission area is acquired at preset time intervals, with the waste-occupied volume measured in cubic meters. Statistical analysis is performed on the changes in waste-occupied volume of the target three-dimensional emission unit over at least two and no more than ten consecutive time slices. When a continuous decreasing trend in waste-occupied volume is detected within consecutive time slices, and the decrease within a single time slice or cumulative time slice is within the range of 5% to 40%, the three-dimensional emission unit is marked as an occupied volume reduction unit.

[0153] Simultaneously, within the same time slice, at least one adjacent three-dimensional emission unit that has a spatial adjacency relationship with the target three-dimensional emission unit is analyzed synchronously; when it is detected that the waste-occupied volume of the adjacent three-dimensional emission unit shows a continuous upward trend within the corresponding time slice, and its increase is also within the range of 5% to 40%, the adjacent three-dimensional emission unit is marked as an occupancy enhancement unit.

[0154] Furthermore, the consistency of the volume change direction of the aforementioned occupation reduction unit and occupation enhancement unit within a continuous time slice is determined; when the waste occupation volume of the target three-dimensional emission unit continues to decrease and the waste occupation volume of the adjacent three-dimensional emission unit continues to increase, and the change direction of the two remains consistent within a continuous time slice, the occupation change relationship is determined to be a stable and repetitive occurrence.

[0155] Based on the above judgment results, it is confirmed that the waste occupancy status forms an effective transfer feature between different three-dimensional emission units. This transfer feature is used to characterize the actual spatial migration behavior of waste, rather than short-term disturbances or random stacking changes, providing a reliable basis for subsequent waste migration constraint identification and emission composition status analysis.

[0156] Specifically, three-dimensional vision acquisition devices or laser scanning devices are deployed within the waste disposal area to periodically scan the waste accumulation area and acquire three-dimensional point cloud data of the waste pile. Subsequently, a spatial surface model of the waste pile is generated through three-dimensional reconstruction processing. Combined with the spatial boundaries of the corresponding three-dimensional disposal unit, the model is volume-trimmed to calculate the actual spatial volume occupied by the waste within that three-dimensional disposal unit. The volume data is recorded in cubic meters and stored as continuous time-series data.

[0157] In another implementation, the volume occupied by waste can be obtained through height measurement and area conversion. Specifically, a height measuring device is installed within the three-dimensional discharge unit to obtain the average height of the waste pile surface, and this is combined with the bottom area of ​​the unit for volume conversion. For example, if the bottom area of ​​a three-dimensional discharge unit is 25 square meters and the average stacking height is 2 meters, then the volume occupied by waste in that unit is 50 cubic meters. By continuously recording the height changes at each time point, continuous data on the change in the volume occupied by waste can be obtained.

[0158] To ensure data comparability across different three-dimensional emission units, the waste-occupied volume is further normalized. Specifically, the waste-occupied volume of each three-dimensional emission unit in the current time slice is proportionally converted to the maximum occupied volume within the unit's historical monitoring period, thus transforming the occupancy level of all units into a uniform proportional value. Through this processing, three-dimensional emission units with different areas or spatial structures can all undergo trend analysis at the same scale.

[0159] In the process of determining volume change, the volume change ratio between consecutive time slices is first calculated. When the change ratio between the current time slice volume and the previous time slice volume is between 5% and 40%, the change is considered to be within the effective range; when the change ratio is below 5%, it is considered to be random fluctuation; when the change ratio exceeds 40%, it is determined to be a sudden accumulation or concentrated clearing behavior, and it is not included in the judgment of stable transfer characteristics.

[0160] After completing the volume measurement and normalization process described above, the volume change trend within the continuous time slice is then assessed. When the target three-dimensional emission unit continuously exhibits an effective downward change within the continuous time slice, while adjacent three-dimensional emission units exhibit a corresponding effective upward change within the same time slice, and the two directions of change remain consistent within the continuous time slice, it is confirmed that the waste has formed a stable spatial transfer characteristic between different three-dimensional emission units.

[0161] Of particular importance is confirming the correspondence between the physical composition of waste and the proportion of waste sorting based on the emission composition status of waste emission areas, including:

[0162] Based on the emission composition state of the waste emission area, the spatial distribution characteristics of waste in the three-dimensional emission unit under the emission composition state are obtained;

[0163] Based on the actual waste sorting results under the corresponding emission composition state in the historical waste disposal records, extract the proportion information of each waste category under the emission composition state to determine the waste sorting ratio;

[0164] A correlation analysis was conducted on the physical composition characteristics of waste and the waste classification ratio to establish a mapping relationship between different emission composition states and corresponding waste classification ratios. This mapping relationship was then used as the correspondence between the physical composition of waste emission areas and the waste classification ratios.

[0165] In one embodiment, based on the emission composition state of the waste emission area, the spatial distribution characteristics of the waste corresponding to the emission composition state are obtained. Specifically, based on three-dimensional emission units, spatial distribution characteristics such as the lateral distribution density, vertical stacking hierarchy structure, and cross-unit migration characteristics of waste under the emission composition state are extracted to characterize the overall spatial manifestation of waste under the emission composition state, thereby providing spatial constraints for physical composition analysis.

[0166] Secondly, based on historical waste disposal records, historical sample data consistent with the current emission composition are selected. These historical waste disposal records include actual sorting, weighing, or volume statistics. Through statistical analysis of various types of waste in the historical samples, the actual proportion of different waste categories under this emission composition is extracted to determine the corresponding waste sorting ratio. This waste sorting ratio reflects the relative composition of various types of waste in source emissions under the same emission composition.

[0167] Finally, a correlation analysis was conducted between the physical composition characteristics of the waste and the waste sorting ratio. Specifically, the physical composition characteristics of the waste (including the degree of blockiness, looseness, recyclability, and mixing) were compared with historical waste sorting ratios to establish a mapping relationship between different emission composition states and corresponding waste sorting ratios. This mapping relationship was then stored as a correspondence between the physical composition of waste emission areas and waste sorting ratios.

[0168] Of particular importance is the correlation analysis between the physical composition characteristics of waste and the proportion of waste sorting, including:

[0169] Waste discharge areas are grouped according to the confirmed emission composition status, and waste discharge areas with the same or similar emission composition status are grouped into the same composition status category.

[0170] For each constituent status category, the physical composition characteristics of waste and the actual waste sorting ratio data of the corresponding waste discharge area are summarized to form a data set corresponding to the status;

[0171] Within the dataset corresponding to the state, the physical composition characteristics of waste and the waste classification ratio data are compared and analyzed to identify the proportional distribution pattern of different waste categories under the same emission composition state.

[0172] Based on the proportional distribution pattern, a mapping relationship is established between the emission composition status and the corresponding waste classification ratio.

[0173] In one embodiment, based on the waste emission composition status confirmed in the aforementioned steps, each waste emission area is grouped according to its status. Specifically, waste emission areas with the same emission composition status or similar spatial distribution characteristics and migration restriction characteristics are uniformly classified into the same composition status category to eliminate interference factors between different emission forms and form several representative emission composition status categories.

[0174] Secondly, for each waste status category, the physical composition characteristics data and actual waste sorting ratio data corresponding to each waste discharge area belonging to that category are summarized. The physical composition characteristics data include indicators such as the degree of waste blockiness, looseness, mixing degree, and the significance of recyclables; the actual waste sorting ratio data comes from historical sorting and weighing or statistical records. Through summarization and processing, a one-to-one corresponding data set of statuses is formed.

[0175] Subsequently, within the dataset corresponding to the stated state, the data on the physical composition characteristics of waste and the data on waste classification ratios are compared and analyzed. By comparing the differences in waste classification ratios corresponding to different combinations of physical composition characteristics under the same emission state, the proportional distribution patterns of various types of waste under this emission state are identified, thereby extracting stable features—proportional relationships.

[0176] Finally, based on the aforementioned proportional distribution pattern, a mapping relationship between emission composition states and corresponding waste classification ratios is established, and this mapping relationship is stored as a rule model or parameter model to describe the typical distribution of waste classification ratios under specific emission composition states, providing basic support for the construction and application of subsequent waste classification prediction models.

[0177] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0178] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for constructing a waste classification model based on data analysis and machine learning, characterized in that, Includes the following steps: Step S1: Confirm the waste disposal area through the preset GIS visualization system, and spatially discretize the waste disposal area, dividing the disposal area into several three-dimensional disposal units with spatial height, area and adjacency relationship; Step S2: For each three-dimensional emission unit, analyze the lateral emission concentration characteristic data and vertical accumulation and diffusion characteristic data of the waste emission area to confirm the emission composition status of the waste emission area; wherein, step S2 includes the following steps: Step S21: For each three-dimensional emission unit, analyze the waste distribution pattern in the same height layer of the waste emission area to obtain the lateral emission concentration characteristic data of the horizontal expansion range and aggregation degree of waste; Step S22: Based on the three-dimensional emission unit along the three-dimensional GIS spatial height direction, perform layered analysis on the changes in waste accumulation in different height layers of the waste emission area, extract the accumulation and diffusion characteristics of waste in the vertical direction, and generate vertical accumulation and diffusion feature data; Step S23: During the continuous emission process, analyze the changes in the migration path of the three-dimensional emission unit in the horizontal and vertical directions, identify the migration constraints of the waste accumulation process, and obtain emission migration restriction characteristic data; Step S24: Confirm the composition of waste emissions based on horizontal emission concentration characteristic data, vertical accumulation and diffusion characteristic data, and emission migration restriction characteristic data; Step S3: Based on the emission composition status of the waste emission area, confirm the correspondence between the physical composition of waste and the waste classification ratio, and construct a waste classification prediction model; Step S4: Use the waste classification prediction model to classify and predict the waste discharge area, output the classification prediction results of the waste source, and visualize them through the GIS visualization system.

2. The method for constructing a waste classification model based on data analysis and machine learning according to claim 1, characterized in that, Methods for confirming continuous emission processes include: The waste discharge area is divided into three-dimensional spaces at preset time intervals to ensure that the same spatial location corresponds to the same three-dimensional discharge unit in different time slices, thereby constructing a three-dimensional discharge unit sequence with temporal continuity. For the three-dimensional emission unit sequence, the change trajectories of the three-dimensional emission units in the horizontal coverage, vertical stacking height and spatial morphological boundary of each time slice are extracted to confirm the internal spatial change trajectory of the three-dimensional emission units. Based on the spatial adjacency relationship between adjacent three-dimensional emission units, we analyze whether waste migrates across units within a continuous time slice, and determine the migration direction and migration range to confirm the external spatial change trajectory of the three-dimensional emission units. By observing the continuity of the internal and external spatial change trajectories, it can be determined whether the three-dimensional emission unit is in a continuous emission process.

3. The method for constructing a waste classification model based on data analysis and machine learning according to claim 2, characterized in that, Based on the spatial adjacency between adjacent three-dimensional emission units, the analysis of whether waste migration occurs across units within consecutive time slices includes: Based on the three-dimensional spatial partitioning results, spatial adjacency relationships between each three-dimensional emission unit are established. These spatial adjacency relationships include at least horizontal, vertical, and diagonal adjacency relationships, forming a corresponding set of three-dimensional emission unit adjacency relationships. Within a continuous time slice, the waste occupancy status of each three-dimensional emission unit in the same waste emission area is compared hourly to identify the characteristics of the waste occupancy status shifting from the three-dimensional emission unit to the spatially adjacent three-dimensional emission unit. When the waste occupancy state is weakened by the three-dimensional emission unit in adjacent time slices and synchronously strengthened in at least one spatially adjacent three-dimensional emission unit, waste migration behavior across three-dimensional emission units is confirmed.

4. The method for constructing a waste classification model based on data analysis and machine learning according to claim 3, characterized in that... The hourly comparison of waste occupancy in each three-dimensional emission unit within the same waste emission area also includes: In adjacent time slices, the spatial coverage, height occupied, and boundary changes of waste within the same three-dimensional emission unit are compared accordingly. Simultaneously, a comparative analysis was conducted on the changes in waste occupancy of adjacent three-dimensional emission units that have a spatial adjacency with the three-dimensional emission unit; Based on the correspondence between the changes in waste occupancy in three-dimensional emission units and the changes in waste occupancy in spatially adjacent units within the time slices, the transfer characteristics of waste occupancy status between different three-dimensional emission units are identified.

5. The method for constructing a waste classification model based on data analysis and machine learning according to claim 1, characterized in that, Step S23 includes the following steps: Step S231: Within a continuous time slice, establish a time series relationship for the three-dimensional emission units corresponding to the same waste emission area; Step S232: Based on the three-dimensional emission unit sequence, extract the cross-unit migration path of waste in the horizontal direction and the accumulation migration path in the vertical direction; Step S233: Identify whether the waste migration is concentrated in a preset spatial direction or within a limited spatial range based on the horizontal and vertical migration paths; Step S234: By using the directional stability features or range-limited features of the waste migration path within a continuous time slice, confirm that there are migration constraints in the waste accumulation process, and generate corresponding emission migration-limited feature data accordingly.

6. The method for constructing a waste classification model based on data analysis and machine learning according to claim 5, characterized in that, Identifying whether waste migration is concentrated in a preset spatial direction or within a limited spatial range based on horizontal and vertical migration paths includes: Based on the spatial coordinate relationship of the three-dimensional emission unit, the migration path of waste in the horizontal direction is classified to determine the main horizontal spatial direction of waste migration in a continuous time slice. Along the height direction of the three-dimensional emission unit, the migration path of waste in the vertical direction is statistically analyzed to determine the main height range that changes during the waste accumulation process; By jointly analyzing the directional classification results of horizontal migration paths and the height range results of vertical migration paths, it can be identified whether waste migration is mainly concentrated in a preset combination of spatial directions or a limited spatial height range. When waste migration is continuously concentrated in a preset combination of spatial directions or a limited spatial height range within a continuous time slice, it is confirmed that waste migration has spatially restricted characteristics.

7. The method for constructing a waste classification model based on data analysis and machine learning according to claim 6, characterized in that, The directional classification of waste migration paths in the horizontal direction is as follows: Based on the spatial relative position of the three-dimensional emission units in the horizontal plane, the migration direction of waste from the three-dimensional emission units to the spatially adjacent three-dimensional emission units within a continuous time slice is marked. The migration direction is divided into several horizontal migration direction categories according to its orientation relative to the three-dimensional emission unit; By statistically analyzing the migration occurrence of each horizontal migration direction category within a continuous time slice, the main horizontal migration direction of waste can be determined.

8. The method for constructing a waste classification model based on data analysis and machine learning according to claim 4, characterized in that, Based on the correspondence between the changes in waste occupancy in three-dimensional emission units and the changes in waste occupancy in spatially adjacent units within different time slices, the transfer characteristics of waste occupancy status between different three-dimensional emission units are identified, including: Within adjacent time slices, the changes in the waste occupancy state in the three-dimensional emission unit are analyzed to determine the direction of change in the degree of waste occupancy. Simultaneously analyze the changes in the waste occupancy state in at least one adjacent three-dimensional emission unit that has a spatial adjacency relationship with the three-dimensional emission unit, and determine the corresponding direction of change; The direction of change in the occupancy of waste in a three-dimensional emission unit is compared with that of adjacent three-dimensional emission units to identify whether there is a corresponding relationship where the occupancy of a three-dimensional emission unit weakens and the occupancy of adjacent three-dimensional emission units strengthens. When the corresponding change relationship is repeated within a continuous time slice, it is confirmed that the waste occupancy state forms a stable transfer characteristic between different three-dimensional emission units.

9. The method for constructing a waste classification model based on data analysis and machine learning according to claim 8, characterized in that, When the corresponding change relationship repeats within a continuous time slice, specifically: Within a continuous period of no less than two and no more than ten time slices, the volume occupied by waste in the three-dimensional emission unit shows a continuous decrease of 5% to 40%, while the volume occupied by waste in at least one spatially adjacent three-dimensional emission unit shows a corresponding increase of 5% to 40% within the same time slice, wherein the volume occupied by waste is measured in cubic meters. Furthermore, when the direction of volume change remains consistent across consecutive time slices, the corresponding change relationship is determined to be stable and repetitive, thus confirming that the waste occupancy state forms an effective transfer characteristic between different three-dimensional emission units.

Images