Garbage sorting method based on artificial intelligence

Abstract

The invention relates to the technical field of intelligent garbage sorting, and discloses a garbage sorting method based on artificial intelligence. According to the method, heterogeneous data fusion is performed on a multi-source garbage feature data set through a semantic alignment module, feature dimensions are unified, redundant items are eliminated, and a standardized feature set is output. And extracting a spatial position identifier and a physical state identifier of the junk entity, and dynamically dividing time slices by adopting a transfer learning model to generate a periodic junk flow data set. The coordinates of the sorting starting point and the sorting ending point are converted into a three-dimensional space grid to construct a migration track topological graph, flow data and physical state identifiers are combined, sorting paths are clustered and analyzed according to material categories, the material flow intensity in a specific period of a target area is quantified, and a dynamic index matrix is generated. According to the method, space-time correlation analysis of multi-modal junk data is realized, and the adaptability of sorting path planning in a complex environment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent garbage sorting, and in particular to a garbage sorting method based on artificial intelligence. Background Art

[0002] With the acceleration of urbanization, the amount of garbage generated continues to rise. As a prerequisite for resource recycling, garbage sorting's efficiency and accuracy directly affect the operation of the subsequent processing chain. Currently, garbage sorting mainly relies on a combination of manual sorting and mechanical sorting. Manual sorting is limited by labor costs and subjective judgment, making it difficult to adapt to the processing needs of large-scale, multi-category garbage. Mechanical sorting is mostly based on a single physical characteristic (such as density and magnetism). For garbage with complex composition and similar morphology, misclassification or omission is often the case.

[0003] Intelligent technologies are increasingly being applied to waste sorting, and the trend is towards multi-source data collection, including image data, weight data, and material detection data. However, these data come from dispersed sources, in heterogeneous formats, and with significantly different feature dimensions. For example, image data is presented as a pixel matrix, weight data as time-series values, and material data as spectral curves. This makes it difficult to directly integrate and analyze data from these different sources. Furthermore, duplicate samples are prone to occur during data collection, carrying redundant information that not only increases the data processing load but also potentially interferes with the accuracy of feature extraction.

[0004] In existing technologies, the description of garbage entities mostly stays at the level of static features, lacking the tracking of dynamic changes in spatial position, making it difficult to reflect the complete migration path of garbage from generation to sorting end point. Even if some technologies attempt to perform trajectory analysis, the lack of a unified three-dimensional spatial grid system leads to confusion in the mapping of starting and end point coordinates, and the inability to form a clear topological relationship. In addition, the physical state of garbage (such as humidity and integrity) changes over time, and traditional methods often ignore such dynamic characteristics, making it difficult to judge the flow intensity of materials.

[0005] When analyzing waste flow patterns, existing technologies often use fixed time intervals to partition data, failing to account for the cyclical nature of waste generation and sorting. This results in an inadequate depiction of material flow variations within different time periods. Furthermore, when analyzing material sorting paths, a lack of multi-dimensional data-based clustering methods makes it difficult to accurately distinguish the sorting patterns of different materials, significantly reducing the effectiveness of targeted resource recovery. These issues collectively hinder the development of efficient and precise waste sorting. Summary of the Invention

[0006] The purpose of the present invention is to provide a garbage sorting method based on artificial intelligence to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides a garbage sorting method based on artificial intelligence, the method comprising:

[0008] Perform heterogeneous data fusion operations on multi-source garbage feature datasets, unify the feature dimensions of data from different sources through the semantic alignment module, remove dimensional redundant items, eliminate duplicate samples based on time series identification, and output a standardized feature set;

[0009] Continuously extract the spatial location identifier and physical state identifier of garbage entities based on a standardized feature set;

[0010] A transfer learning model is used to implement dynamic time slicing on the standardized feature set to generate a garbage flow dataset with periodic attributes.

[0011] Combining the garbage flow dataset with spatial location identifiers, the coordinates of the starting and ending points of garbage sorting are converted into a three-dimensional spatial grid system to construct a topological map of garbage migration trajectories.

[0012] The garbage migration trajectory topology map, garbage flow dataset and physical state identifier are associated to perform cluster analysis on garbage sorting paths by material category, quantify the flow intensity of different materials in a specific period in the target sorting area, and generate a dynamic indicator matrix of material sorting behavior.

[0013] Preferably, the heterogeneous data fusion operation specifically includes:

[0014] Analyze the semantic relevance of all feature labels in the multi-source garbage feature dataset, map heterogeneous feature names to a unified concept space through the domain ontology library, and output the initial feature table;

[0015] Detect feature dimensions with value conflicts in the initial feature table, retain feature values ​​of high-priority data sources based on data source credibility weights, remove low-priority conflicting items, and generate an intermediate feature table;

[0016] Based on the time series identifier and entity code of each sample in the intermediate feature table, the entity samples that appear repeatedly in the time window are identified, and the data records of the latest time series version are retained to form a standardized feature set.

[0017] Preferably, the operation of constructing the garbage migration trajectory topology map includes:

[0018] Analyze the initial location coordinates and sorting target location coordinates of garbage entities from the standardized feature set, and convert the unstructured location description into spatial coordinate processing results;

[0019] By using the entity codes in the garbage flow dataset, the garbage flow records in each time slice are associated and mapped with the spatial coordinate processing results to generate a coordinate trajectory dataset with spatiotemporal attributes.

[0020] Convert all spatial coordinate pairs in the coordinate trajectory dataset into a three-dimensional spatial grid system, and construct garbage migration paths with flow direction vectors based on the conversion results to form a spatial trajectory data set;

[0021] A weighted directed network is constructed based on the spatial trajectory data set, in which network nodes represent the sorting area codes, and network edges represent the migration direction and frequency, and finally a garbage migration trajectory topology map is generated.

[0022] Preferably, the operation of generating a dynamic indicator matrix further includes:

[0023] The waste flow dataset is split into multiple material category subsets based on physical state identifiers, and the sorting starting and ending areas in the waste migration trajectory topology map are associated to form a set of migration paths divided by material category and area.

[0024] The input and output of each type of material in the target area in the migration path set are counted to construct a three-dimensional tensor of material category region period. Based on this tensor, the flow intensity function is used to calculate the flow intensity value of different materials in each area and each period, and the flow intensity sequence is output;

[0025] The cross-period variation characteristics of the flow intensity sequence are analyzed, the flow intensity change rate is calculated by the time series fluctuation algorithm, and the dynamic indicator matrix of the material sorting behavior is generated by combining the standard deviation normalization process.

[0026] Preferably, the operation process of the semantic alignment module is implemented by the following operations:

[0027] The names of all feature dimensions in the multi-source garbage feature dataset are collected to build a global feature name set pool; through the predefined industry terminology mapping table, each name in the global feature name set pool is semantically parsed to identify feature items with the same meaning but different names between different data sources, and a standardized naming rule library is determined; the identified synonymous feature names are replaced with word meanings according to the standardized naming rule library, so that all data sources adopt unified feature naming; finally, the feature set with completed name replacement is dimensionally aligned according to the dimension order defined by the preset feature structure template to generate an initial feature table.

[0028] Preferably, the conflict item handling process includes the following steps:

[0029] Mark the original data source identifier corresponding to each feature value in the initial feature table, and assign priority weight values ​​to different sources based on the preset trust scoring system; scan all feature dimensions with the same name in the initial feature table, compare the numerical differences from different data sources, and when numerical conflicts are detected, only retain the feature value of the data source with the highest priority weight value; perform a clearing operation on feature values ​​with low priority weight values ​​or inconsistent values; reintegrate all feature data that have been conflict-cleared by dimension to generate an intermediate feature table.

[0030] Preferably, the specific execution process of the operation of eliminating duplicate samples is:

[0031] Extract the time series identifier field and the junk entity code field from the intermediate feature table and combine them to form a composite primary key index. Set a time difference tolerance threshold interval. When the timestamp difference between multiple records with the same composite primary key index is within this tolerance interval, it is determined to be a duplicate sample set. Only the data record with the latest timestamp is retained in the duplicate sample set, and the remaining records are physically deleted. After executing this deduplication process on the entire table, a standardized feature set is output.

[0032] Preferably, the spatial coordinate processing result conversion process includes:

[0033] Convert the original coordinate coefficient values ​​in the coordinate trajectory data set to the standardized coordinates of the three-dimensional space grid system;

[0034] Apply spatial grid coding rules to perform regional discretization coding on standardized coordinates;

[0035] By integrating entity coding and time slice identification, the garbage migration path is represented as a vector line segment from the starting grid code to the end grid code, and the material category and flow frequency attributes are added;

[0036] All vector segments are integrated at the time slice level to form a spatial trajectory data set.

[0037] Preferably, the construction logic of the weighted directed network is:

[0038] Map the starting grid code and the ending grid code to the standard sorting area nodes;

[0039] Aggregate the migration paths with the same starting and ending areas in the same time slice, and accumulate their flow frequencies as edge weights;

[0040] A weighted directed network is constructed with the sorting area nodes as vertices, the aggregation migration direction as the directed edge, and the flow frequency accumulation value as the edge weight.

[0041] Visualize the topology of garbage migration trajectories output by a weighted directed network.

[0042] Preferably, the flow intensity function calculation process includes:

[0043] Extract the input and output of each type of material to the target area at each time slice from the three-dimensional tensor of the material category area period;

[0044] Obtaining the processing capacity parameter of the sorting equipment in the target area as a normalization coefficient;

[0045] Divide the input and output by the normalization coefficient of the corresponding area to obtain the material flow intensity value;

[0046] Arrange the flow intensity values ​​in time slice order to form a flow intensity sequence.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This method fuses heterogeneous data from multiple sources of garbage feature datasets, unifying the feature dimensions of these different sources with the help of a semantic alignment module. This method eliminates analytical barriers caused by differences in data format and feature scale, allowing previously dispersed data types such as image, weight, and material to be integrated into a collaboratively analyzed whole. Furthermore, removing redundant dimensional items and eliminating duplicate samples based on time series identifiers reduces the impact of invalid data on subsequent processing, allowing the standardized feature set to better reflect the true characteristics of garbage entities and providing a more reliable data foundation for subsequent analysis.

[0049] Extracting spatial location identifiers and physical state identifiers for garbage entities based on standardized feature sets transcends the limitations of traditional static feature descriptions and characterizes garbage entities from both spatial dynamics and physical attributes. Spatial location identifiers accurately record the location of garbage at different nodes, while physical state identifiers capture changes in the garbage's state during its circulation. The combination of these two provides a more comprehensive description of garbage entities and helps to more precisely understand the evolving characteristics of garbage during the sorting process.

[0050] A transfer learning model is used to dynamically slice the standardized feature set into time slices, generating a garbage flow dataset with cyclical properties that adapts to the cyclical patterns of garbage generation and sorting. Compared to partitioning based on fixed time intervals, this dynamic partitioning allows the length and interval of time slices to be adjusted according to the actual flow rhythm, making the garbage flow data more consistent with actual flow patterns and facilitating the capture of differences in flow characteristics within different cycles.

[0051] Combining the waste flow dataset with spatial location identifiers, the coordinates of the starting and ending points are converted into a three-dimensional grid system, and a topological map of the waste migration trajectory is constructed. This clearly shows the complete migration path of waste from generation to sorting. The introduction of the three-dimensional grid system makes the coordinate mapping more unified, and the trajectory topology map intuitively displays the relationships between each node, making the waste migration process more concrete and helping to clarify the logic of waste flow between different regions.

[0052] By correlating migration trajectory topology maps, waste flow datasets, and physical state identifiers, we perform cluster analysis of waste sorting paths by material type. This allows us to group waste of the same material with similar sorting paths together and refine the sorting behavior characteristics of different materials. Based on this, we quantify the flow intensity in target areas within a specific cycle and generate a dynamic indicator matrix. This matrix intuitively reflects the flow patterns of different materials during the sorting process, providing multi-dimensional reference information for understanding material sorting patterns and optimizing sorting processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a working principle diagram of the artificial intelligence-based garbage sorting method of the present invention;

[0054] Figure 2 It is a flowchart of heterogeneous data fusion operation;

[0055] Figure 3 A diagram showing the working principle for generating a dynamic indicator matrix;

[0056] Figure 4 A flowchart of the conflict handling process. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] See also Figure 1 The present invention provides a garbage sorting method based on artificial intelligence, the method comprising:

[0059] Perform heterogeneous data fusion operations on multi-source garbage feature datasets, unify the feature dimensions of data from different sources through the semantic alignment module, remove dimensional redundant items, eliminate duplicate samples based on time series identification, and output a standardized feature set;

[0060] The multi-source garbage feature dataset is derived from sensor data at garbage collection points, manually entered garbage information, and GPS track data from garbage transport vehicles. These data vary in feature names, formats, and dimensions. The semantic alignment module processes these feature names, unifying features with different names but the same meaning. For example, "garbage weight" and "waste mass" are unified into "garbage quality." After unifying the feature dimensions, duplicate or unnecessary dimensions are removed. Then, based on the time series identifier of each data item, duplicate samples are identified and the most recent ones are retained, ultimately resulting in a standardized feature set.

[0061] Continuously extract the spatial location identifier and physical state identifier of garbage entities based on a standardized feature set;

[0062] The standardized feature set contains various types of information about garbage entities. The spatial location identifier can be extracted from location-related fields in the data, such as the latitude and longitude of the garbage, area code, etc.; the physical state identifier involves the physical properties of the garbage, such as hardness, humidity, compressibility, etc., which are obtained by parsing and extracting the corresponding data in the standardized feature set.

[0063] A transfer learning model is used to implement dynamic time slicing on the standardized feature set to generate a garbage flow dataset with periodic attributes.

[0064] The transfer learning model leverages existing experience in garbage data processing and, based on the patterns and characteristics of garbage flow, divides the standardized feature set into different time slices in chronological order. These time slices have specific periodic attributes, such as hourly or daily. The data within each time slice forms information related to garbage flow during that time period, which is then integrated to form a garbage flow dataset.

[0065] Combining the garbage flow dataset with spatial location identifiers, the coordinates of the starting and ending points of garbage sorting are converted into a three-dimensional spatial grid system to construct a topological map of garbage migration trajectories.

[0066] We extract garbage flow records from the garbage flow dataset over time, combine them with the starting and ending coordinates identified by spatial location identifiers, and transform these coordinates into a three-dimensional grid. By processing these coordinates and flow records, we construct a topological map that reflects the path, direction, and frequency of garbage migration—the garbage migration trajectory topology map.

[0067] The garbage migration trajectory topology map, garbage flow dataset and physical state identifier are associated to perform cluster analysis on garbage sorting paths by material category, quantify the flow intensity of different materials in a specific period in the target sorting area, and generate a dynamic indicator matrix of material sorting behavior.

[0068] By correlating the migration paths reflected in the garbage migration trajectory topology map, the flow information in the garbage flow dataset, and the physical properties of the garbage reflected in the physical state identifier, the sorting paths are clustered according to the material category of the garbage, such as plastic, metal, and paper. The flow of different materials in the target sorting area within a specific period is analyzed, and their flow intensity is calculated. Finally, this information is organized into a dynamic indicator matrix that dynamically reflects the sorting behavior characteristics of different materials.

[0069] Example 1: See Figure 2 The heterogeneous data fusion operation begins by analyzing the semantic relevance of all feature labels in the multi-source garbage feature dataset. The multi-source garbage feature dataset may include image recognition data from garbage generation points, weighing data from smart garbage bins, loading records from transport vehicles, preliminary classification data from sorting centers, etc. These data come from different collection terminals and systems, and feature labels have significant differences in naming methods and expression habits. The domain ontology library contains a standardized concept system for the garbage disposal field, covering multiple categories such as material properties, spatial location, time dimension, physical state, etc. By mapping heterogeneous feature names to this unified concept space, it is possible to eliminate misunderstandings caused by different naming habits. For example, "object material" in image recognition data, "garbage type" in weighing data, and "material type" in sorting records all correspond to the core concept of "material category" in the domain ontology library. After mapping, they are uniformly classified into the same feature dimension, and the final output is an initial feature table containing unified feature names.

[0070] The semantic alignment module plays a key role in this process. First, the names of all feature dimensions in the multi-source garbage feature dataset are collected. These names may include "collection time", "weight value", "community", "moisture content", "recyclable", etc. These names are aggregated to build a global feature name collection pool. The predefined industry term mapping table covers synonyms, antonyms and nicknames commonly used in the garbage disposal industry, such as the correspondence between "pickup time" and "collection time", "dryness and humidity" and "moisture content", "community number" and "area code". Based on this mapping table, each name in the global feature name collection pool is semantically parsed, and the core meaning of the feature names is compared one by one to identify feature items with the same meaning but different names in different data sources. Then a standardized naming rule base is determined. The rule base is based on common terminology in the field and combines the convenience of data processing to determine the standard name of each feature. For example, "weight value", "garbage quality" and "material weight" are standardized as "garbage weight".

[0071] After completing the semantic analysis, the synonymous feature names identified will be replaced with their meanings according to the standardized naming rule library. The replacement process needs to traverse all feature fields of each data source to ensure that features with the same meaning use exactly the same names in different data sources. For example, all features involving the moisture content of garbage will be uniformly changed to "moisture content", and features involving the location of garbage production will be uniformly changed to "production area". After the semantic replacement, the feature set will be arranged in the order of dimensions defined by the preset feature structure template. The feature structure template may set the dimensions in the order of time attributes, spatial attributes, physical attributes, and material attributes, and arrange features such as "collection time", "production area", "moisture content", and "material category" in this order to form an initial feature table with a unified structure.

[0072] After the initial feature table is generated, the process begins by detecting conflicting feature dimensions. Due to varying data collection methods and accuracy across multiple sources, inconsistent values ​​may occur within the same feature dimension. For example, the "waste weight" of the same batch of garbage may be 50 kg in the transport vehicle's records, but 52 kg when weighed at the sorting center. This constitutes a conflict. These conflicts are addressed based on the data source's credibility weight, which is determined by factors such as the accuracy of the collection equipment, the degree of automation in the collection process, and the robustness of the data verification mechanism. For example, readings from intelligent weighing equipment, which are automatically collected and regularly calibrated, carry a higher credibility weight than manual records. Professional inspection data from the sorting center, which undergoes multiple verifications, carries a higher credibility weight than estimated data collected during transportation. Feature values ​​from high-priority data sources are retained according to their weighting, while low-priority conflicting items are removed. After processing, an intermediate feature table is generated, in which only one conflict-free feature value is retained for each feature dimension on the same entity sample.

[0073] Duplicate samples are identified and processed based on the time series identifier and entity code for each sample in the intermediate feature table. The time series identifier is typically a timestamp accurate to the second, and the entity code uniquely identifies each garbage entity, such as "20240501-001," generated by combining batch and serial number. A time window is set, which can be based on the actual garbage processing cycle, such as 24 hours. Within this window, samples with the same composite primary key index, composed of the time series identifier and entity code, are identified. When multiple records have the same composite primary key index, they are considered duplicates. In this case, only the record with the latest timestamp is retained, and the remaining records are physically deleted. For example, if the same batch of garbage is recorded three times within 30 minutes, only the latest record is retained. This deduplication process is performed on the entire intermediate feature table, ensuring that each entity sample has only one latest record within the time window, ultimately forming a standardized feature set. The standardized feature set meets unified standards for feature names, data structure, and sample uniqueness.

[0074] Example 2: See Figure 3 The operation of constructing a garbage migration trajectory topology map begins with parsing the initial location coordinates and sorting target location coordinates of the garbage entity from a standardized feature set. The standardized feature set contains the location information of the garbage entity during its generation, transportation, and sorting. This information may exist in different forms, some of which are structured coordinate data, such as latitude and longitude, regional grid codes, and some are unstructured descriptive text, such as "XX Street Garbage Transfer Station", "Chengdong Sorting Center Area A", "Community Garbage Bin No. 3", etc. Converting unstructured location descriptions into spatial coordinate processing results requires the use of a geographic information mapping system, which contains a precise spatial coordinate database of garbage disposal-related facilities. By matching name keywords or address codes, the physical location corresponding to the text description is converted into a three-dimensional coordinate value, where the X-axis and Y-axis correspond to planar geographic coordinates, and the Z-axis corresponds to the level or height information of the facility, such as the floor number of the sorting center, the placement height of the garbage bin, etc.

[0075] Through the entity codes in the garbage flow dataset, the garbage flow records within each time slice are associated and mapped with the spatial coordinate processing results. The garbage flow dataset is divided into time slices, and each time slice corresponds to a fixed time interval, such as 1 hour or 6 hours. It records the transfer of garbage entities within this time period, including the entity code, transfer time, and the processing nodes involved. The entity code is the unique identifier of each garbage entity and runs through the entire garbage processing process. Through this code, the flow records within the time slice can be associated with the corresponding initial position coordinates and target position coordinates, forming a coordinate trajectory dataset in which each record contains both time attributes (the time slice to which it belongs) and spatial attributes (the position coordinates).

[0076] All spatial coordinate pairs in the coordinate trajectory dataset are converted to a three-dimensional spatial grid system. The three-dimensional spatial grid system is a three-dimensional grid divided according to the geographical scope of the garbage disposal area. Each grid cell has a unique code, and the size of the grid can be adjusted according to the distribution density of garbage disposal facilities in the area. The grid cells in areas with dense facilities are smaller, and the grid cells in areas with sparse facilities are larger. During the conversion process, the original coordinate values ​​are first mapped to the corresponding grid cells, and then the specific coordinate values ​​are replaced by grid codes, making the spatial location information more convenient for batch processing and analysis. Based on the conversion results, a garbage migration path with a flow vector is constructed. The flow vector points from the starting grid to the end grid and contains direction information and distance parameters. After summarizing the migration paths of all garbage entities, a spatial trajectory data set is formed.

[0077] A weighted directed network is constructed based on the spatial trajectory data set, and then a garbage migration trajectory topology map is generated. The operation of generating a dynamic indicator matrix starts by splitting the garbage flow dataset into multiple material category subsets based on the physical state identifier. The physical state identifier contains physical property information such as the hardness, density, conductivity, and magnetism of the garbage. This information can be used to distinguish different material categories, such as metals have conductivity and magnetism, plastics have a range of densities, and paper has lower hardness. Based on these properties, the garbage flow dataset is divided into material category subsets such as metal, plastic, paper, glass, and kitchen waste. Each subset contains the flow records of the corresponding material in each time slice.

[0078] By associating the sorting starting and ending areas in the garbage migration trajectory topology map, the complete path of each material from generation to final sorting is determined, forming a set of migration paths divided by material category and region. The input and output of each type of material in the migration path set in the target region are calculated. Input refers to the total amount of material entering the target region from other regions within a specific time slice, and output refers to the total amount of material transferred from the target region to other regions within the same time slice. This data is organized by material category, region, and time slice to construct a three-dimensional tensor of material category and region period. Each element of this tensor corresponds to the input or output of a certain type of material in a specific region and time slice.

[0079] Based on a three-dimensional tensor of material categories and regions, a flow intensity function is used to calculate the flow intensity values ​​of different materials in each region and period. The flow intensity values ​​reflect the level of material flow activity between regions. After the calculation is completed, the flow intensity values ​​of the same material in the same region at different time slices are arranged in chronological order, and a flow intensity sequence is output. The cross-periodic variation characteristics of the flow intensity sequence are analyzed, and the flow intensity change rate is calculated using a time series fluctuation algorithm. The time series fluctuation algorithm captures the numerical differences between adjacent periods in the sequence. A positive change rate indicates an increase in flow intensity, while a negative change rate indicates a decrease in flow intensity. Combined with standard deviation normalization, the change rate is converted into a standardized indicator value, eliminating dimensional differences between different materials and regions. The resulting dynamic indicator matrix has rows representing material categories, columns representing regions and time periods, and matrix elements representing standardized flow intensity-related indicators, which comprehensively reflects the dynamic changes in material sorting behavior.

[0080] Example 3: See Figure 4During the conflict resolution process, the initial feature table must be labeled with the source identifier of the original data corresponding to each feature value. Sources of original data include smart sensors at waste generation points, manual data entry terminals, onboard systems in transport vehicles, and detection equipment at sorting centers. Each source has a unique identification code, such as "Sensor-001" for a smart trash can sensor in a community, "Manual-002" for a manual data entry terminal on a street, and "Truck-003" for an onboard system in a transport vehicle. A pre-defined trust rating system evaluates different sources based on three dimensions: the degree of automation in data collection, the frequency of device calibration, and the data verification mechanism. Highly automated equipment, such as regularly calibrated smart sensors, receives a relatively high score. Manual data entry terminals, due to potential input errors, receive a relatively low score. Onboard systems equipped with real-time verification receive a higher score than those without. Based on the evaluation results, priority weights are assigned to different sources, ranging from 0 to 1. For example, a smart sensor might receive a weight of 0.85, a manual data entry terminal 0.6, and an onboard system equipped with real-time verification 0.75.

[0081] Scan all feature dimensions with the same name in the initial feature table and compare the numerical differences from different data sources. For example, in the feature dimension of "garbage weight", the value recorded by the smart sensor is 5.2kg, the value entered manually is 5kg, and the value recorded by the on-board system is 5.3kg. These values ​​differ due to different measurement accuracy and methods, constituting a numerical conflict. When a numerical conflict is detected, extract the priority weight value of each data source and only retain the feature value of the data source with the highest weight value. If the weight value of the smart sensor is the highest, retain the value of 5.2kg, and remove the manually entered 5kg and the 5.3kg of the on-board system as low-priority conflicting items. Perform the same operation on all conflicting feature dimensions in the initial feature table, reintegrate all feature data that have cleared conflicts by dimension, and arrange all data in the same feature dimension in the order of entity coding to generate an intermediate feature table.

[0082] The deduplication process begins by extracting the time series identifier field and the junk entity code field from the intermediate feature table. The time series identifier field represents the precise time of data record generation, accurate to the millisecond level, such as "2024-05-2008:30:25.123." The entity code field uniquely identifies each junk entity, consisting of the date the junk was generated, the region code, and a random sequence, such as "20240520-A03-789." These two fields are combined to form a composite primary key index, such as "2024-05-2008:30:25.123+20240520-A03-789." Each composite primary key index corresponds to a unique junk entity's state at a specific moment in time.

[0083] Set a time difference tolerance threshold interval, which is determined based on the actual operating cycle of the garbage disposal process. For example, the interval from garbage generation to transportation is usually no more than 2 hours, so the time difference tolerance threshold interval can be set to 2 hours. When the entity codes in the composite primary key index of two records are the same, and the time difference of the time series identifier is within this interval, it is determined to be a duplicate sample set. For example, two records corresponding to the same entity code have time series identifiers of "2024-05-2008:30:25.123" and "2024-05-2009:15:30.456" respectively, with a time difference of 45 minutes, which is within the threshold interval of 2 hours, and belongs to a duplicate sample set.

[0084] In the duplicate sample set, only the data record with the latest timestamp is retained, and the remaining records are physically deleted. For example, in the duplicate sample set above, the record with the latest timestamp "2024-05-2009:15:30.456" is retained, and the record corresponding to "2024-05-2008:30:25.123" is deleted. This deduplication process is performed on the entire intermediate feature table, checking the composite primary key index of each record row by row to identify and process all duplicate sample sets.

[0085] During the processing, if a special case is encountered where the timestamp is the same but the eigenvalue is different, the deviation rate of the eigenvalue is calculated to assist in the judgment. The calculation formula for the deviation rate is:

[0086]

[0087] Where D represents the deviation rate, V1 and V2 are the feature values ​​of the two records, and max(V1, V2) represents the maximum value of V1 and V2. When the deviation rate is below the set threshold, the two records are considered to have a slight difference due to measurement error, and the record with the feature value closer to the historical average of the entity is retained. When the deviation rate is above the threshold, both records are marked as outliers and temporarily excluded from the standardized feature set until manual verification. After deduplication and anomaly processing are completed for the entire table, a standardized feature set is output. Each record in this feature set represents the latest status data of a unique entity within the valid time range.

[0088] Example 4: Mapping the starting grid code and the ending grid code to standard sorting area nodes. Standard sorting area nodes are fixed areas divided according to the waste disposal process, covering waste generation points, transfer stations, sorting centers, recycling plants, etc. Each area has a unique code, such as "G01" represents a waste collection point in a residential area, "Z03" represents a waste transfer station in the western part of the city, "F02" represents a plastic sorting workshop, and "C05" represents a metal recycling plant. The starting grid code and the ending grid code are derived from a spatial trajectory data set. Each grid code corresponds to a specific grid cell in a three-dimensional spatial grid system. By querying the grid-area correspondence table, the grid code can be converted to the corresponding standard sorting area node code. For example, the spatial location corresponding to the grid code "3-15-2" belongs to the "G01" area, and the spatial location corresponding to the grid code "7-9-1" belongs to the "Z03" area. Then, the starting point and end point of the migration path from "3-15-2" to "7-9-1" are mapped to the two standard sorting area nodes "G01" and "Z03", respectively.

[0089] Aggregate migration paths with the same starting and ending areas within the same time slice, and accumulate their flow frequencies as edge weights. A time slice is the time division unit of the garbage flow dataset. For example, if a time slice is 4 hours, each day can be divided into 6 time slices. Within the same time slice, there may be multiple migration paths from "G01" to "Z03", each path corresponding to a migration record of a garbage entity. For example, in the time slice numbered "T08" (corresponding to 8:00-12:00 on a certain day), the migration path from "G01" to "Z03" appears 12 times, indicating that 12 batches of garbage were transported from area "G01" to area "Z03" during this time period. The cumulative frequency of these 12 flows, 12, is the edge weight connecting "G01" and "Z03". If the migration path from "Z03" to "F02" appears 8 times and the migration path from "F02" to "C05" appears 5 times in the same time slice, the corresponding edge weights are 8 and 5 respectively.

[0090] A weighted directed network is constructed, with sorting area nodes as vertices, aggregated migration directions as directed edges, and the accumulated flow frequency as edge weight. Vertices are the standard sorting area nodes described above, such as "G01," "Z03," "F02," "C05," and so on. Each vertex is represented by a node in the network, and the size of the node can be adjusted based on the total migration frequency of that area within a given period. Nodes in areas with high migration frequency appear visually larger. Directed edges point from the starting area to the ending area. For example, a directed edge from "G01" to "Z03" represents the flow of garbage from a residential collection point to the West District transfer station; a directed edge from "Z03" to "F02" represents the flow of garbage from the transfer station to the plastic sorting workshop. The thickness of the edge is proportional to the edge weight. The larger the weight, the thicker the edge appears visually. For example, an edge with a weight of 12 is thicker than an edge with a weight of 8, visually reflecting the difference in flow frequency along different migration paths.

[0091] Visualizing a weighted directed network outputs a topological map of garbage migration trajectories. During visualization, the network layout is optimized to avoid excessive overlap of nodes and edges, making the topology clear and easy to read. The topology map also features interactive time slices. By switching between different time slices, you can view changes in garbage migration trajectories within that time period. For example, if you switch to time slice "T14" (corresponding to 2:00 PM - 6:00 PM on a specific day), if the flow frequency of the migration path from "G01" to "Z03" increases to 15 times, the edge thickness will be adjusted accordingly. If a reverse migration path from "Z03" to "G01" is added within this time slice (e.g., when unqualified garbage is returned), a new directed edge from "Z03" to "G01" will be added, with a weight equal to the frequency of this reverse migration. In this way, the garbage migration trajectory topological map dynamically displays the garbage migration between sorting areas within different time slices, including migration direction and flow frequency, providing an intuitive visual representation for analyzing the path patterns of garbage sorting.

[0092] Example 5: The calculation process of the flow intensity function begins by extracting the input and output of each type of material to the target area at each time slice from the material category area period three-dimensional tensor. The material category area period three-dimensional tensor is a data structure containing three dimensions: material category, area, and time slice, with specific classification items under each dimension. Material categories include plastics, metals, paper, glass, kitchen waste, etc.; areas cover various nodes in the garbage disposal process, such as community collection points, regional transfer stations, different workshops in the sorting center, and final recycling plants; time slices are time periods divided at fixed time intervals, such as one slice every 6 hours, and a day can be divided into 4 slices.

[0093] Taking the flow of plastic materials in "Area A of the South City Sorting Center" as an example, it can be extracted from the three-dimensional tensor that in the first time slice (0:00-6:00), the input amount is the total amount of plastic transported into Area A from other areas during this time period. Assuming that these plastics come from the East City Transfer Station and the North City Collection Point, which are 200kg and 150kg respectively, then the total input amount is 350kg; the output amount is the total amount of plastic transported out of Area A in the same time slice. Assuming that they are transported to the plastic recycling plant and the quality inspection workshop respectively, which are 180kg and 120kg respectively, then the total output amount is 300kg. Similarly, it can be extracted that the input amount of the second time slice (6:00-12:00) is 420kg and the output amount is 380kg, the input amount of the third time slice (12:00-18:00) is 500kg and the output amount is 450kg, and the input amount of the fourth time slice (18:00-24:00) is 300kg and the output amount is 250kg.

[0094] Obtain the target area's sorting equipment processing capacity parameter as the normalization coefficient. This parameter is the maximum amount of material that the equipment in that area can handle per unit time. This parameter is determined by the equipment's model, power, and operating efficiency. For example, if the plastic sorting equipment in "Chengnan Sorting Center Area A" has a maximum processing capacity of 500 kg per six hours, the normalization coefficient for that area is 500 kg.

[0095] Divide the input and output by the normalization coefficient for the corresponding area to obtain the material flow intensity value. For the plastic material in "Chengnan Sorting Center Area A," the input flow intensity value for the first time slice is 350÷500=0.7, and the output flow intensity value is 300÷500=0.6; the input flow intensity value for the second time slice is 420÷500=0.84, and the output flow intensity value is 380÷500=0.76; the input flow intensity value for the third time slice is 500÷500=1, and the output flow intensity value is 450÷500=0.9; and the input flow intensity value for the fourth time slice is 300÷500=0.6, and the output flow intensity value is 250÷500=0.5.

[0096] Arranging flow intensity values ​​in time slice order forms a flow intensity sequence. For plastic materials in "Chengnan Sorting Center Area A," the input flow intensity sequence is [0.7, 0.84, 1, 0.6], and the output flow intensity sequence is [0.6, 0.76, 0.9, 0.5]. The same approach can be applied to other material categories and other areas, such as metal materials in "Chengbei Sorting Center Area B" and paper materials in "Chengdong Sorting Center Area C." These sequences can be used to subsequently analyze the changing characteristics of material flow intensity and generate a dynamic indicator matrix.

[0097] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0098] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A garbage sorting method based on artificial intelligence, characterized in that: The method comprises the following steps: Perform heterogeneous data fusion operations on multi-source garbage feature datasets, unify the feature dimensions of data from different sources through the semantic alignment module, remove dimensional redundant items, eliminate duplicate samples based on time series identification, and output a standardized feature set; Continuously extract the spatial location identifier and physical state identifier of garbage entities based on a standardized feature set; A transfer learning model is used to implement dynamic time slicing on the standardized feature set to generate a garbage flow dataset with periodic attributes. Combining the garbage flow dataset with spatial location identifiers, the coordinates of the starting and ending points of garbage sorting are converted into a three-dimensional spatial grid system to construct a topological map of garbage migration trajectories. The garbage migration trajectory topology map, garbage flow dataset and physical state identifier are associated to perform cluster analysis on garbage sorting paths by material category, quantify the flow intensity of different materials in a specific period in the target sorting area, and generate a dynamic indicator matrix of material sorting behavior.

2. The artificial intelligence-based garbage sorting method according to claim 1, characterized in that: The heterogeneous data fusion operation specifically includes: Analyze the semantic relevance of all feature labels in the multi-source garbage feature dataset, map heterogeneous feature names to a unified concept space through the domain ontology library, and output the initial feature table; Detect feature dimensions with value conflicts in the initial feature table, retain feature values ​​of high-priority data sources based on data source credibility weights, remove low-priority conflicting items, and generate an intermediate feature table; Based on the time series identifier and entity code of each sample in the intermediate feature table, the entity samples that appear repeatedly in the time window are identified, and the data records of the latest time series version are retained to form a standardized feature set.

3. The artificial intelligence-based garbage sorting method according to claim 1, characterized in that: The operation of constructing the garbage migration trajectory topology map includes: Analyze the initial location coordinates and sorting target location coordinates of garbage entities from the standardized feature set, and convert the unstructured location description into spatial coordinate processing results; By using the entity codes in the garbage flow dataset, the garbage flow records in each time slice are associated and mapped with the spatial coordinate processing results to generate a coordinate trajectory dataset with spatiotemporal attributes. Convert all spatial coordinate pairs in the coordinate trajectory dataset into a three-dimensional spatial grid system, and construct garbage migration paths with flow direction vectors based on the conversion results to form a spatial trajectory data set; A weighted directed network is constructed based on the spatial trajectory data set, in which network nodes represent the sorting area codes, and network edges represent the migration direction and frequency, and finally a garbage migration trajectory topology map is generated.

4. The artificial intelligence-based garbage sorting method according to claim 3, characterized in that: The operation of generating a dynamic indicator matrix further includes: The waste flow dataset is split into multiple material category subsets based on physical state identifiers, and the sorting starting and ending areas in the waste migration trajectory topology map are associated to form a set of migration paths divided by material category and area. The input and output of each type of material in the target area in the migration path set are counted to construct a three-dimensional tensor of material category region period. Based on this tensor, the flow intensity function is used to calculate the flow intensity value of different materials in each area and each period, and the flow intensity sequence is output; The cross-period variation characteristics of the flow intensity sequence are analyzed, the flow intensity change rate is calculated by the time series fluctuation algorithm, and the dynamic indicator matrix of the material sorting behavior is generated by combining the standard deviation normalization process.

5. The artificial intelligence-based garbage sorting method according to claim 2, characterized in that: The operation process of the semantic alignment module is achieved through the following operations: The names of all feature dimensions in the multi-source garbage feature dataset are collected to build a global feature name set pool; through the predefined industry terminology mapping table, each name in the global feature name set pool is semantically parsed to identify feature items with the same meaning but different names between different data sources, and a standardized naming rule library is determined; the identified synonymous feature names are replaced with word meanings according to the standardized naming rule library, so that all data sources adopt unified feature naming; finally, the feature set with completed name replacement is dimensionally aligned according to the dimension order defined by the preset feature structure template to generate an initial feature table.

6. The artificial intelligence-based garbage sorting method according to claim 5, characterized in that: The conflict item handling process includes the following steps: Mark the original data source corresponding to each feature value in the initial feature table, and assign priority weight values ​​to different sources based on the preset trust rating system; scan all feature dimensions with the same name in the initial feature table, compare the differences in values ​​from different data sources, and when value conflicts are detected, only retain the feature value from the data source with the highest priority weight value; Perform a clearing operation on feature values ​​with low priority weights or inconsistent values; reintegrate all feature data that have completed conflict clearing by dimension to generate an intermediate feature table.

7. The artificial intelligence-based garbage sorting method according to claim 6, characterized in that: The specific execution process of the operation of eliminating duplicate samples is as follows: Extract the time series identifier field and the junk entity code field from the intermediate feature table and combine them to form a composite primary key index. Set a time difference tolerance threshold interval. When the timestamp difference between multiple records with the same composite primary key index is within this tolerance interval, it is determined to be a duplicate sample set. Only the data record with the latest timestamp is retained in the duplicate sample set, and the remaining records are physically deleted. After executing this deduplication process on the entire table, a standardized feature set is output.

8. The artificial intelligence-based garbage sorting method according to claim 3, characterized in that: The spatial coordinate processing result conversion process includes: Convert the original coordinate coefficient values ​​in the coordinate trajectory data set to the standardized coordinates of the three-dimensional space grid system; Apply spatial grid coding rules to perform regional discretization coding on standardized coordinates; By integrating entity coding and time slice identification, the garbage migration path is represented as a vector line segment from the starting grid code to the end grid code, and the material category and flow frequency attributes are added; All vector segments are integrated at the time slice level to form a spatial trajectory data set.

9. The artificial intelligence-based garbage sorting method according to claim 8, characterized in that: The construction logic of the weighted directed network is: Map the starting grid code and the ending grid code to the standard sorting area nodes; Aggregate the migration paths with the same starting and ending areas in the same time slice, and accumulate their flow frequencies as edge weights; A weighted directed network is constructed with the sorting area nodes as vertices, the aggregation migration direction as the directed edge, and the flow frequency accumulation value as the edge weight. Visualize the topology of garbage migration trajectories output by a weighted directed network.

10. The artificial intelligence-based garbage sorting method according to claim 4, characterized in that: The flow intensity function calculation process includes: Extract the input and output of each type of material to the target area at each time slice from the three-dimensional tensor of the material category area period; Obtaining the processing capacity parameter of the sorting equipment in the target area as a normalization coefficient; The input and output quantities are divided by the normalization coefficient of the corresponding area to obtain the material flow intensity value; the flow intensity values ​​are arranged in time slice order to form a flow intensity sequence.

Images