Automatic classification method of campus garbage integrated with artificial intelligence image recognition

Abstract

The present invention discloses a method for automatically classifying campus garbage integrated with artificial intelligence image recognition, and relates to the technical field of garbage sorting. The method comprises: analyzing the garbage generation pattern of the target campus to obtain M garbage generation trend information; calling the garbage image archive log; constructing and generating a garbage image recognition model and a garbage classification control model; optimizing resource calls to complete the construction of an automatic classification control unit; constructing a one-way communication channel; receiving and returning the real-time garbage image on the garbage classification conveyor belt to perform automatic garbage classification and real-time sorting. The present invention solves the technical problems of the rough campus garbage classification in the prior art, resulting in large manpower consumption and low processing efficiency in garbage treatment, and achieves the technical effect of realizing automatic classification and real-time sorting according to the characteristics of garbage in different regions, and improving the intelligent level and efficiency of campus garbage treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of garbage sorting, and in particular to a method for automatic classification of campus garbage integrating artificial intelligence image recognition. Background Art

[0002] In the campus environment, the proper disposal of solid waste is an important task. Among them, solid waste can be mainly divided into kitchen waste and domestic waste composed of various components, including plastics, paper, textiles, metals, wood and bamboo, etc. At present, most primary and secondary school campuses generally adopt the method of classifying waste into four categories: kitchen waste, recyclable waste, hazardous waste and other waste. However, in the actual operation process, there are many problems that need to be solved. On the one hand, the proportion of various components (such as plastics, paper, textiles, metals, wood and bamboo, etc.) in campus domestic waste has always been unclear. The lack of such data makes it difficult for the campus to formulate accurate and effective strategies in terms of waste management and resource recycling. For example, if the specific proportion of plastic waste generation is not clear, it is impossible to reasonably plan the recycling process and facility investment of plastic waste. On the other hand, the classification of campus waste has long relied mainly on the manual labor of campus cleaning staff. Cleaning staff need to manually sort various types of garbage. This process not only requires a lot of manpower costs, but also consumes a lot of time and material resources, such as the consumption of garbage bags and garbage cans. More importantly, the manual classification method is greatly affected by subjective factors, and it is difficult to ensure the accuracy and consistency of the classification results. For example, different cleaning staff may make different judgments on some difficult-to-distinguish garbage (such as whether some composite material products belong to recyclable or other garbage categories), resulting in a relatively rough classification result that cannot meet the requirements of modern efficient and accurate garbage disposal.

[0003] The existing technology has the defect of rough classification of campus garbage, which leads to technical problems such as high manpower consumption and low processing efficiency in garbage disposal. Summary of the invention

[0004] The present application provides a method for automatic classification of campus waste that integrates artificial intelligence image recognition, which is used to solve the technical problems in the prior art of rough classification of campus waste, resulting in high manpower consumption and low processing efficiency in waste treatment.

[0005] In view of the above problems, the present application provides a method for automatic classification of campus garbage integrating artificial intelligence image recognition, the method comprising:

[0006] Analyze the garbage generation pattern of the target campus to obtain M garbage generation trend information of M sample functional areas; obtain garbage image archive logs through local data calls; build and generate garbage image recognition models and garbage classification control models based on the garbage image archive logs; optimize the resource calls of the garbage image recognition model and the garbage classification control model based on the M garbage generation trend information to complete the construction of an automatic classification control unit; construct a one-way communication channel between the image collector of the garbage classification conveyor belt and the automatic classification control unit; the automatic classification control unit receives and performs automatic garbage classification and real-time sorting based on the real-time garbage images on the garbage classification conveyor belt sent back by the image collector.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] Analyze the garbage generation pattern of the target campus and obtain M garbage generation trend information of M sample functional areas; obtain garbage image archive logs through local data call; build and generate garbage image recognition model and garbage classification control model; optimize the resource call of the garbage image recognition model and garbage classification control model according to the M garbage generation trend information, and complete the construction of the automatic classification control unit; build a one-way communication channel between the image collector of the garbage classification conveyor belt and the automatic classification control unit; the automatic classification control unit receives and performs automatic garbage classification and real-time sorting based on the real-time garbage image on the garbage classification conveyor belt transmitted back by the image collector. The technical effect of realizing automatic classification and real-time sorting according to the characteristics of garbage in different areas and improving the intelligent level and efficiency of campus garbage disposal has been achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a flow chart of a method for automatic classification of campus garbage with integrated artificial intelligence image recognition provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of obtaining garbage generation trend information in the automatic campus garbage classification method integrating artificial intelligence image recognition provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application provides a method for automatic classification of campus waste that integrates artificial intelligence image recognition, which is used to solve the technical problems in the prior art that the classification of campus waste is rough, resulting in high manpower consumption and low processing efficiency in waste treatment.

[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0014] Examples, such as Figure 1 As shown, the present application provides a method for automatic classification of campus garbage integrating artificial intelligence image recognition, the method comprising:

[0015] Step S100: Analyze the garbage generation pattern of the target campus to obtain M garbage generation trend information of M sample functional areas.

[0016] Specifically, a comprehensive and in-depth analysis of the garbage generation pattern of the target campus is carried out, with the aim of obtaining M garbage generation trend information corresponding to each of the representative M sample functional areas. First, it is necessary to obtain a detailed campus space planning map of the target campus by interacting with the campus-related management system or database, which is the basic blueprint for subsequent regional division and analysis. Then, based on the understanding of the functional layout of the campus and research needs, a variety of campus area types are preset, which cover M sample functional areas, such as teaching areas, student living areas, faculty and staff office areas, public activity areas, etc. Then, using the preset campus area types as classification standards, the campus space planning map is systematically traversed, so that the target campus is carefully divided into multiple campus areas with clear functional attributes, and each area is assigned a functional area identification for subsequent accurate identification and analysis. After completing the regional division, the garbage generation pattern of the campus area corresponding to each functional area identification is deeply explored. The data on the amount of various types of garbage generated, changes in types, and factors that may affect garbage generation in different time periods are collected and organized. After careful data processing and analysis, M garbage generation trend information that precisely corresponds to M sample functional areas is finally summarized. This information will provide an indispensable decision-making basis for subsequent model construction and resource optimization allocation.

[0017] Step S200: Obtain the junk image archive log by calling local data.

[0018] Specifically, the garbage image archive log is obtained by calling local data. The local database stores a large amount of past campus garbage-related image information. These images were taken and recorded at different times, locations, and scenes, covering the actual conditions of garbage dumping and treatment processes in every corner of the campus. Through precise data query and extraction operations, these historical garbage image data are integrated to form a garbage image archive log. This log is not only a visual record of the campus garbage situation, but also provides rich, diverse, and authentic data materials for the construction of subsequent garbage image recognition models, which helps to improve the model's recognition ability and accuracy of various garbage image features, and lays a solid data foundation for accurate classification.

[0019] Step S300: constructing and generating a garbage image recognition model and a garbage classification control model according to the garbage image archive log.

[0020] Specifically, the regional area information of multiple campus areas is obtained through local data calls, and the image sampling weight is calculated based on this. When calculating, the regional area ratio and the garbage generation frequency ratio are comprehensively considered. For example, the regional area ratio is obtained by comparing the area of ​​each area with the total area, and the garbage generation frequency ratio is determined by the ratio of the number of garbage generation times in each area within a specific period to the total number of garbage generation times according to the garbage generation trend information. Then, the weight coefficients (the sum of the two is 1) are set for the two respectively for weighted summation to obtain the image sampling weight. Then, according to these weights, the regional image archive logs in the garbage image archive log are sampled to form multiple sample garbage image sets. At the same time, M garbage generation trend information is analyzed, and the garbage types are aggregated to determine K sample garbage types, such as plastic, paper, textiles, metal, wood and bamboo, etc. With these sample garbage types as constraints, the sample garbage image set is artificially semantically identified to generate a sample identified image set. The garbage image recognition model is constructed using the sample garbage image set and the sample identification image set. The model consists of K parallel garbage image recognition branches. Each branch learns the image features of a sample garbage type, such as shape, color, texture, etc., to achieve directional recognition and marking. The grasping control test is carried out for K types of sample garbage, simulating the grasping process of the robot arm to obtain K grasping control parameters, including grasping force, angle, speed, etc. Finally, the sample garbage type is associated with the grasping control parameters and stored to complete the construction of the garbage classification control model, providing a precise control strategy for subsequent garbage sorting.

[0021] Step S400: optimizing the resource call of the garbage image recognition model and the garbage classification control model according to the M garbage generation trend information, and completing the construction of the automatic classification control unit.

[0022] Specifically, based on the M garbage generation trend information, the resource call optimization work of the garbage image recognition model and the garbage classification control model is comprehensively carried out to build an automatic classification control unit. For the K garbage image recognition branches in the garbage image recognition model, the frequency of occurrence, quantity proportion, etc. of different garbage types in the M garbage generation trend information are deeply analyzed. For those garbage image recognition branches corresponding to the sample garbage types with large generation and high recognition difficulty in various areas of the campus, more computing power resources are allocated to ensure that they can quickly and accurately process image data and identify garbage categories, thereby obtaining M garbage classification recognition models suitable for garbage recognition scenarios in different regions. In terms of the garbage classification control model, based on the M garbage generation trend information, the distribution and processing requirements of different garbage types in various regions are also considered. For the multiple garbage sorting robotic arms on the garbage classification conveyor belt, M robotic arm call strategies are carefully generated. Each strategy contains K call ratio parameters corresponding to K sample garbage types, and clarifies the robotic arm operation ratio that should be allocated to each garbage type in a specific area. Afterwards, based on the K types of sample garbage, the M robot arm call strategies are associated and stored with the K grasping control parameters previously obtained through the grasping control test, thereby obtaining M robot arm control models, which specify in detail the grasping action details and call rules of the robot arm for different types of garbage. Finally, according to the M sample functional areas, a mapping communication connection is established between the M garbage classification recognition models and the M robot arm control models, so that the two can work together. When garbage enters the system from different areas, the corresponding garbage classification recognition model first accurately analyzes and recognizes the garbage image, and then the robot arm control model accurately dispatches the robot arm to perform garbage sorting actions according to the recognition results, according to the preset call ratio and grasping parameters, and successfully completes the construction of the automatic classification control unit, providing a solid and reliable guarantee for the efficient, accurate and automatic classification of campus garbage.

[0023] Step S500: constructing a one-way communication channel between the image collector of the garbage sorting conveyor belt and the automatic sorting control unit.

[0024] Specifically, it is crucial to build a one-way communication channel between the image collector of the garbage sorting conveyor belt and the automatic sorting control unit. First, it is necessary to select appropriate communication technology and hardware equipment according to the actual equipment layout and data transmission requirements, such as using high-speed Ethernet, wireless communication modules (such as Wi-Fi, Bluetooth or wireless communication technology in a specific frequency band), etc., to ensure that the high-speed and stable transmission requirements of image data can be met. Then, the communication protocol is designed and configured, the image collector is defined as the data transmitter, the automatic sorting control unit is defined as the data receiver, and the data format, encoding method, and transmission rate and other parameters are specified to ensure that the real-time garbage image on the garbage sorting conveyor belt collected by the image collector can be accurately encoded and transmitted in the established format. In terms of hardware connection, the output port of the image collector is correctly connected to the selected communication device (such as a network switch, a wireless communication base station, etc.) to establish a one-way data transmission link from the image collector to the communication device; at the same time, the receiving port of the automatic sorting control unit is connected to the corresponding interface of the communication device to realize one-way data access from the communication device to the automatic sorting control unit. To ensure the unidirectionality of the communication channel, strict settings and restrictions are made at the communication equipment and software levels to prevent reverse data transmission or the entry of interference signals. For example, an access control list (ACL) is set in the network configuration to only allow the image collector to send data to the automatic classification control unit, while prohibiting the automatic classification control unit from sending any data to the image collector; in the wireless communication settings, a unidirectional transmission mode is adopted or targeted permissions are set for the receiving end so that it can only receive data from the image collector. Through these measures, a stable, efficient and unidirectional communication channel is successfully built, providing reliable communication guarantee for the subsequent automatic classification control unit to receive real-time garbage image data and perform automatic classification and sorting.

[0025] Step S600: The automatic classification control unit receives and performs automatic classification and real-time sorting of garbage based on the real-time garbage image on the garbage classification conveyor belt transmitted back by the image collector.

[0026] Specifically, before the garbage operation and maintenance equipment is ready to dump campus garbage on the garbage sorting conveyor belt, it will first load the functional area identification of the real-time upload area to the automatic sorting control unit. This identification serves as key information, enabling the automatic sorting control unit to quickly locate and activate the corresponding garbage sorting recognition model and robotic arm control model. Subsequently, the automatic sorting control unit receives the garbage images on the garbage sorting conveyor belt sent back by the image collector in real time, and uses the activated garbage sorting recognition model to quickly and accurately analyze these real-time images. Through deep learning algorithms and pre-trained model parameters, it identifies the types of various types of garbage in the image, such as accurately distinguishing different types of garbage such as plastic, paper, textiles, metal, wood and bamboo. Once the garbage type identification is completed, the real-time robotic arm control model immediately starts the scheduling operation according to the identification results. According to the preset robotic arm action strategies for different types of garbage, multiple garbage sorting robotic arms are controlled to move accurately to the corresponding positions, and the garbage on the conveyor belt is sorted in real time with appropriate parameters such as gripping force, angle and speed. For example, for lighter plastic waste, the robot arm will use a smaller gripping force and a specific gripping angle to ensure accurate gripping without causing the waste to scatter; for heavier metal waste, it will adjust to a larger gripping force and a stable gripping method. The entire process is completed efficiently in a short time, realizing the automatic classification and real-time sorting of garbage, greatly improving the efficiency and accuracy of campus garbage disposal, effectively reducing the errors and workload of manual classification, and improving the intelligent level of campus environmental management.

[0027] In one possible implementation, Figure 2 As shown, step S100 also includes:

[0028] Step S110: interactively obtaining a campus space planning map of the target campus.

[0029] Step S120: Preset the campus area type, wherein the campus area type includes the M types of sample functional areas.

[0030] Step S130: traverse the campus space planning map using the campus area type to divide the target campus into multiple campus areas, wherein the multiple campus areas have multiple functional area identifiers.

[0031] Step S140: Analyze the garbage generation patterns of the multiple campus areas according to the multiple functional area identifiers to obtain M garbage generation trend information corresponding to the M types of sample functional areas.

[0032] Specifically, we first need to interact with the relevant information systems of the campus management department to obtain the campus space planning map of the target campus. This planning map is a detailed blueprint of the campus layout, covering the distribution of various buildings, roads, public areas, etc. on campus. Through specific data interfaces or application processes, the planning map data is extracted from platforms such as the campus geographic information system (GIS) or campus asset management system to ensure the accuracy and completeness of the data. The planning map will serve as the basic framework for subsequent regional division and analysis, providing key support for a comprehensive understanding of the campus space structure.

[0033] After obtaining the campus space planning map, the campus area types are preset according to the functional characteristics of the campus and the characteristics of garbage generation. These types include M sample functional areas, such as teaching area, living area, sports area, administrative area, etc. The teaching area is mainly the area where teaching activities are concentrated, and its garbage generation is related to the use of teaching supplies, teachers and students' diet, etc.; the living area is the area where students and faculty members live their daily lives, and the types of garbage are relatively diverse, including domestic garbage, kitchen waste, etc.; the garbage in the sports area may be related to sports activities, such as beverage bottles, discarded sports equipment, etc.; the administrative area mainly produces office-related garbage, such as paper, ink cartridges, etc. Clarifying these area types will help to more specifically divide the campus area and analyze the garbage generation pattern.

[0034] The preset campus area types are used as classification standards to comprehensively traverse the campus space planning map. Starting from the upper left corner of the planning map, identification and judgment are performed pixel by pixel or region by region, and areas with similar functional characteristics are classified into one category. For example, when it is identified that an area is mainly composed of teaching facilities such as teaching buildings and laboratory buildings, it is marked as a teaching area and assigned a corresponding functional area logo; if it is found that it is an area where living facilities such as dormitories and canteens are concentrated, it is determined to be a living area and assigned a corresponding logo. Through such a detailed traversal operation, the target campus is accurately divided into multiple campus areas, each of which has a clear functional (such as teaching area, living area, sports area, administrative area) area logo, which makes the functional attributes of each area on campus clear at a glance, laying a solid foundation for the subsequent analysis of garbage generation patterns.

[0035] Finally, based on multiple functional area labels, an in-depth analysis of the garbage generation pattern of each campus area is carried out. For each functional area, garbage generation data within a certain period of time (such as one semester or one year) is collected, including information such as the type, quantity, and generation time of garbage. In the teaching area, the changing rules of garbage generation in different time periods (such as during class, between classes, and during exams) and the impact of various teaching activities (such as experimental classes, theoretical classes, etc.) on garbage types are analyzed; in the living area, the impact of residents' living habits on garbage generation in different seasons and different dates (weekdays and weekends) is studied, such as the possibility of more kitchen waste in summer and the possibility of an increase in the total amount of domestic waste on weekends; in the sports area, attention is paid to the relationship between sports events, daily exercise and garbage generation; in the administrative area, the impact of factors such as office rhythm and meeting arrangements on office waste generation is examined. Through comprehensive statistics, analysis and induction of these data, M garbage generation trend information corresponding to M sample functional areas is finally obtained, which will provide important decision-making basis for building an efficient automatic garbage classification system.

[0036] In a possible implementation, step S140 further includes:

[0037] Step S141: According to the multiple functional area identifiers, the multiple campus areas are divided into M groups of campus areas corresponding to the M types of sample functional areas.

[0038] Step S142: calling W regional garbage archive logs of W campus areas in the first group of campus areas.

[0039] Step S143: After the W regional garbage archive logs are time-series aligned, data fusion is performed to obtain a first sample archive log.

[0040] Step S144: outputting a first garbage generation trend graph based on the first sample archive log fitting, wherein the first garbage generation trend graph includes a plurality of garbage time-varying curves of a plurality of sample campus garbage.

[0041] Step S145: Analyze the change in the proportion of multiple types of garbage generated on the first garbage generation trend graph, and output the first garbage generation trend.

[0042] Step S146: Similarly, garbage generation pattern analysis of the multiple campus areas is performed based on the multiple functional area identifiers to obtain information corresponding to the M garbage generation trend information.

[0043] Specifically, based on multiple functional area identifiers, the previously divided multiple campus areas are classified and grouped. Campus areas with the same or similar functional area identifiers are integrated together to form M groups of campus areas corresponding to M sample functional areas. For example, all campus areas marked as teaching areas are grouped together, all living areas are grouped together, and so on. This grouping method helps to conduct separate and in-depth research on the garbage generation situation in different types of functional areas, making the analysis results more targeted and accurate.

[0044] For the first group of campus areas (taking the teaching area as an example, the subsequent operations for other groups are the same), select W representative campus areas (the value of W can be determined according to the actual situation, such as selecting areas with large areas, frequent personnel activities, or areas with typical teaching activity characteristics), and then call the W regional garbage archive logs corresponding to each of these W campus areas. These regional garbage archive logs record the detailed garbage generation data of each area in the past period of time (such as the past year), including the time, type, and quantity of garbage generation, and are an important source of original data for analyzing the garbage generation pattern of the area.

[0045] After obtaining the garbage archive logs of W regions, since the time range and time interval of these log records may be different, in order to accurately analyze the time variation law of garbage generation, a time alignment operation is required. By identifying the timestamp information in each log, the garbage generation data of different regions are accurately aligned on the time axis to ensure that all data can be compared and analyzed at the same time scale. After completing the time alignment, perform a data fusion operation to merge and integrate the garbage generation data of these W regions at the same time point or time period. For example, add up the amount of paper garbage generated in different teaching areas in the same week, add up the amount of plastic garbage, etc., to obtain a first sample archive log that comprehensively reflects the garbage generation situation of this group of campus areas (teaching areas). This sample archive log integrates garbage data from multiple regions and can more comprehensively reflect the garbage generation characteristics of this type of functional area.

[0046] Based on the first sample archive log, the data fitting technology is used to generate the first garbage generation trend chart. First, the data in the sample archive log is classified and sorted according to the garbage type, and the generation data of various sample campus garbage such as paper, plastic, metal, and kitchen waste at different time points are extracted. Then, for each sample campus garbage, a suitable mathematical model (such as a linear regression model, a polynomial fitting model, etc., selected according to the characteristics of the data) is used to fit its generation data that changes over time to obtain the corresponding garbage time-varying curve. These curves intuitively show the changing trend of the generation of each sample campus garbage within a certain time range (such as each month of the year). For example, paper garbage may be generated more in the school season and the examination season, while plastic garbage increases in summer due to increased beverage consumption. By integrating multiple garbage time-varying curves of various sample campus garbage in the same chart, the first garbage generation trend chart is formed.

[0047] The first garbage generation trend chart is analyzed in depth, focusing on the changes in the proportion of multiple types of garbage generation. Calculate the proportion of each sample campus garbage generation to the total garbage generation at different time points or time periods, and analyze the changing trends of these proportions over time. For example, it is observed that over a period of time, as the school implements environmental protection policies, the proportion of plastic garbage generation gradually decreases, while the increase in paper recycling rate leads to a relatively stable or slightly decreased proportion of paper garbage generation. Through such analysis, we can clearly understand the importance of different types of garbage in the total amount of garbage in this group of campus areas (teaching areas) and their dynamic changes, thereby outputting the first garbage generation trend. This trend not only includes the changing trend of the generation of each type of garbage, but also reflects the changing trend of the proportional relationship between them, which provides a key basis for the subsequent optimization of garbage disposal strategies and resource allocation.

[0048] According to the above analysis process for the first group of campus areas (teaching areas), the same operations are performed on other groups of campus areas (living areas, sports areas, administrative areas, etc.). That is, first group the areas, call the regional garbage archive log, perform time series alignment and data fusion, fit the garbage generation trend chart, and finally analyze the change in garbage generation ratio and output the corresponding garbage generation trend. Through this one-by-one analysis method, M garbage generation trend information corresponding to M sample functional areas is finally obtained, and the garbage generation mode and change law of different functional areas of the entire target campus are fully grasped, laying a solid data foundation for building an accurate and efficient garbage automatic classification system.

[0049] In a possible implementation, step S144 further includes:

[0050] Step S1441: Split the first sample archive log according to a preset time series division scale to obtain a plurality of first-stage archive logs.

[0051] Step S1442: Decompose the plurality of first-stage archive logs according to garbage types to obtain a plurality of groups of garbage generation ratios of a plurality of garbage generation nodes.

[0052] Step S1443: reorganize the multiple groups of garbage generation ratios according to the garbage types to obtain multiple garbage ratio time series data of the multiple sample campus garbage.

[0053] Step S1444: fitting the multiple garbage ratio time series data to obtain the multiple garbage time-varying curves.

[0054] Step S1445: fusing the multiple garbage time-varying curves by image, and outputting the first garbage generation trend graph.

[0055] Specifically, the first sample archive log is split according to a pre-set time series division scale. This time series division scale can be determined based on time units (such as weeks, months, quarters, etc.) or specific events (such as the beginning and end of a semester, holidays, etc.), with the purpose of segmenting continuous garbage generation data according to a certain time pattern. For example, if the time series division scale is selected as the month, the garbage generation data recorded in the first sample archive log within a year will be split into 12 parts, each part corresponding to the garbage generation situation in one month, thereby obtaining multiple first-stage archive logs. These first-stage archive logs respectively record the garbage generation information in each time period in detail, providing basic data for the subsequent in-depth analysis of the time change characteristics of garbage generation.

[0056] For the multiple first-stage archive logs obtained, they are broken down in detail according to the garbage type. The garbage generation data in the archive logs of each stage are classified and counted according to different garbage types (such as plastic, paper, metal, kitchen waste, other garbage, etc.), and the proportion of each type of garbage generated in the total garbage generated at each garbage generation node (i.e., a specific time point or time period in each stage) is calculated. In this way, multiple sets of garbage generation ratio data for multiple garbage generation nodes are obtained. For example, in a certain month (first-stage archive log), it is counted that a total of 1,000 kilograms of garbage is generated in that month, of which 200 kilograms are plastic garbage, so the proportion of plastic garbage in this month is 20%. In this way, the relative contribution of various types of garbage to the total garbage generation at different time points can be clearly understood, providing key data support for the subsequent analysis of the changing trend of garbage types.

[0057] After obtaining multiple sets of garbage generation ratios, these ratio data are reorganized according to the garbage type. The generation ratios of the same garbage type at different garbage generation nodes (different stages) are recombined together to form a time series data corresponding to each sample campus garbage, that is, multiple garbage ratio time series data. For example, by arranging the generation ratio of plastic garbage in each month (different stages) in chronological order, the garbage ratio time series data of plastic garbage is obtained. Such a reorganization operation enables us to observe the changes in the generation ratio of each type of garbage from a time dimension, which helps to discover the time laws and trends in the garbage generation pattern.

[0058] The mathematical fitting method is used to process multiple garbage ratio time series data to obtain multiple garbage time-varying curves. For each sample campus garbage ratio time series data, a suitable mathematical model (such as linear function, polynomial function, exponential function, etc., determined according to the distribution characteristics and change trends of the data) is selected for fitting operations. For example, if the proportion of plastic garbage production shows a linear growth trend over time, a linear regression model can be used for fitting. Through the fitting operation, a mathematical function expression that can accurately describe the law of change of the proportion of each type of garbage production over time is obtained. Plotting these functions forms multiple garbage time-varying curves. These curves intuitively show the changes in the proportion of each sample campus garbage at different time points, and are important visualization tools for analyzing garbage production trends.

[0059] Finally, multiple garbage time-varying curves are subjected to image fusion operations, and they are integrated in the same coordinate system to form a comprehensive chart, that is, the output of the first garbage generation trend chart. In this trend chart, the horizontal axis usually represents time (according to the previously set time series division scale), and the vertical axis represents the garbage generation ratio. Different curves represent different sample campus garbage types. By observing the trends, heights, and cross-relationships between these curves, we can clearly understand how the generation ratios of various sample campus garbage change over time within the time range covered by the entire first sample archive log, as well as the relative change relationship between them. For example, we can intuitively see which type of garbage generation ratio has increased faster in a certain period of time, which type of garbage is relatively stable or declining, etc., so as to fully grasp the garbage generation trend characteristics of this group of campus areas (such as teaching areas), and provide a strong basis for subsequent garbage management decisions (such as resource allocation, classification strategy formulation, etc.).

[0060] Finally, multiple garbage time-varying curves are subjected to image fusion operations, and they are integrated in the same coordinate system to form a comprehensive chart, that is, the output of the first garbage generation trend chart. In this trend chart, the horizontal axis usually represents time (according to the previously set time series division scale), and the vertical axis represents the garbage generation ratio. Different curves represent different sample campus garbage types. By observing the trends, heights, and cross-relationships between these curves, we can clearly understand how the generation ratios of various sample campus garbage change over time within the time range covered by the entire first sample archive log, as well as the relative change relationship between them. For example, we can intuitively see which type of garbage generation ratio has increased faster in a certain period of time, which type of garbage is relatively stable or declining, etc., so as to fully grasp the garbage generation trend characteristics of this group of campus areas (such as teaching areas), and provide a strong basis for subsequent garbage management decisions (such as resource allocation, classification strategy formulation, etc.).

[0061] In a possible implementation, step S300 further includes:

[0062] Step S310: obtaining the area information of multiple campus areas through local data calling.

[0063] Step S320: Calculate and obtain multiple image sampling weights based on the multiple region area information.

[0064] Step S330: performing image sampling of multiple regional image archive logs in the junk image archive log according to the multiple image sampling weights to obtain multiple sample junk image sets.

[0065] Step S340: Aggregate the M pieces of garbage generation trend information by garbage type to obtain K sample garbage types.

[0066] Step S350: performing artificial semantic identification of the plurality of sample garbage image sets based on the K sample garbage types as constraints to obtain a plurality of sample identified image sets.

[0067] Step S360: constructing and generating the junk image recognition model based on the multiple sample junk image sets and the multiple sample identification image sets, wherein the junk image recognition model is composed of K junk image recognition branches connected in parallel, and the K junk image recognition branches are used for directional recognition and marking of the K types of sample junk types.

[0068] Step S370: Perform a grab control test on the K types of sample garbage, and after obtaining K grab control parameters, associate and store the K types of sample garbage and the K grab control parameters to complete the construction of the garbage classification control model.

[0069] Specifically, by interacting with local data sources such as the campus geographic information system or campus asset management system, the regional area information of multiple campus areas is extracted. This information accurately records the size of the space occupied by each campus area and is one of the important bases for the subsequent calculation of image sampling weights. There are significant differences in the size of different areas, and this difference is often related to the amount of garbage generated and the distribution of garbage types. For example, larger areas may generate more types and quantities of garbage. Therefore, the impact of area factors on sample representativeness needs to be considered during the sampling process.

[0070] After obtaining the regional area information, multiple image sampling weights are calculated based on these data. The calculation process comprehensively considers multiple factors such as regional area factors and garbage generation trends to ensure the scientificity and rationality of the sampling. First, calculate the proportion of each area in the total area of ​​the campus. The larger the proportion, the higher the weight may be given in the sampling because its garbage generation situation has a greater impact on the overall situation. At the same time, it may also be combined with the garbage generation trend information obtained from the previous analysis. For example, although some areas are not large in area, garbage is generated frequently and of special types. Appropriate consideration will also be given when calculating the weight, so that the sampling weight can accurately reflect the importance of each area in the garbage image archive, and provide guidance for the subsequent targeted extraction of representative image samples.

[0071] According to the calculated multiple image sampling weights, image sampling operations are performed on multiple regional image archive logs in the garbage image archive log. According to the weight size, the corresponding proportion of images are extracted from each regional image archive log to form multiple sample garbage image sets. The sampling process aims to screen out image samples that best represent the garbage characteristics of different areas of the campus from a large amount of historical garbage image data. These samples cover various possible garbage scenes, types and states, providing a rich and diverse data basis for subsequent model training. Through reasonable sampling weight allocation, it is ensured that the sample garbage image set not only contains images of common garbage types and scenes, but also highlights images of special areas or special garbage generation situations, thereby improving the model's recognition ability and adaptability to the overall situation of campus garbage.

[0072] Aggregating the types of garbage for M garbage generation trend information is a key step in building an accurate garbage image recognition model and garbage classification control model. First, we deeply analyze each garbage generation trend information, which records in detail the changing trend of the amount of garbage generated in different sample functional areas within a certain period of time. From these rich data, we sort out all the types of garbage involved to form an initial set of garbage types. Then, we carefully classify and sort this initial set. Group garbage types with similar characteristics, properties or treatment methods together, such as classifying various plastic products into plastics, various types of paper into paper, textiles of different materials into textiles, metal products into metals, and wood and bamboo products into wood and bamboo. In the classification process, we fully consider the physical properties (such as material, shape, size, etc.), chemical properties (whether it is recyclable, harmful, etc.) of the garbage, as well as the common forms and sources in the campus environment. After this system aggregation operation, K representative sample garbage types are finally determined. These sample garbage types cover the main categories of campus garbage and can comprehensively reflect the composition of campus garbage. They provide a clear and scientific classification basis for subsequent work such as artificial semantic identification of sample garbage image sets, model construction, and garbage sorting control based on these types, ensuring that the entire automatic garbage classification system can accurately and efficiently identify and process campus garbage.

[0073] Manual semantic labeling of multiple sample garbage image sets with K sample garbage types as constraints is an important step in building an effective garbage image recognition model. First, multiple sample garbage image sets obtained through careful sampling are obtained. These image sets contain rich and diverse garbage images extracted from garbage image archive logs, reflecting the garbage conditions in different areas and at different times on campus. Then, each image in the image set is viewed one by one according to the predetermined K sample garbage types, such as plastic, paper, textiles, metal, wood and bamboo, etc. Carefully observe the morphology, material, color and other characteristics of the garbage in the image, and use professional knowledge and experience to determine the sample garbage type to which it belongs. For example, when an object with a certain degree of transparency and flexibility is seen in the image, it is identified as plastic garbage in combination with its appearance characteristics and common garbage types; for objects that appear as paper materials, with paper texture and shape in the image, they are marked as paper garbage. Through such a meticulous manual judgment and labeling process, each image is given an accurate semantic label, that is, the sample garbage type to which it belongs. After comprehensive processing of the entire sample garbage image set, multiple sample labeling image sets are finally obtained. These image sets with precise semantic identification provide high-quality, clearly labeled training data for the subsequent training of garbage image recognition models, enabling the models to learn the unique features of different sample garbage types in images, thus laying a solid foundation for accurate and efficient garbage image recognition.

[0074] Using multiple sample garbage image sets and multiple sample identification image sets to build a garbage image recognition model is the core step to achieve intelligent garbage classification. First, multiple sample garbage image sets are used as input data for the model. These image sets contain rich and diverse garbage images, covering various garbage forms under different angles, lighting conditions, and backgrounds. At the same time, multiple sample identification image sets are used as corresponding output labels to clearly indicate the sample garbage type to which each image belongs. During the construction process, deep learning technology and convolutional neural network (CNN) architecture are used. The model consists of K parallel garbage image recognition branches, each of which focuses on identifying a specific type of sample garbage. For each branch, the model is trained with image data of the corresponding type in a large number of sample garbage image sets, so that the model learns the unique image features of the sample garbage type, such as plastic garbage may have smooth surface, specific color and shape, and paper garbage has paper texture, folded shape and other features. During the training process, the model continuously adjusts internal parameters to minimize the error between the predicted results and the true labels in the sample identification image set. After multiple rounds of training, the model completes the training process when it reaches a high accuracy on the validation set. The final garbage image recognition model, whose K garbage image recognition branches can accurately perform directional identification and marking of K types of sample garbage when faced with new garbage images, provides key type judgment basis for subsequent automatic garbage classification, greatly improving the efficiency and accuracy of garbage classification.

[0075] To build a garbage classification control model, it is necessary to obtain key parameters by performing grasping control tests on K types of sample garbage types. First, for each type of sample garbage (such as plastic, paper, textile, metal, wood and bamboo, etc.), the actual garbage sorting scene is simulated, and the grasping operation test is performed using specially designed test equipment and robotic arms. During the test, various control parameters of the robotic arm are systematically adjusted, including but not limited to grasping force, grasping angle, grasping speed, and the motion trajectory of the robotic arm. For different types of sample garbage, due to their different characteristics such as material, shape, weight and texture, the optimal grasping control parameters required are also different. For example, when grasping lighter and softer textiles, a smaller grasping force and a specific grasping angle may be required to prevent them from slipping or being damaged; while when grasping heavier and harder metal garbage, a larger grasping force and a stable grasping method are required. Through a large number of repeated grasping test experiments, the data such as the grasping success rate, grasping stability, and the impact on the garbage itself and the surrounding environment of the robotic arm for each type of sample garbage under different parameter settings are recorded and analyzed. Based on these experimental data, the optimal grab control parameters for each type of sample garbage are determined, that is, K grab control parameters are obtained. Finally, the K types of sample garbage are associated with the corresponding K grab control parameters and stored to form a complete parameter correspondence library, thereby completing the construction of the garbage classification control model. In actual operation, this model can quickly and accurately call the corresponding grab control parameters according to the type of garbage determined by the garbage image recognition model, thereby accurately controlling the movement of the garbage sorting robot arm, achieving efficient and accurate garbage classification operations, and ensuring the stable and reliable operation of the entire automatic garbage classification system.

[0076] In a possible implementation, step S400 further includes:

[0077] Step S410: adjusting the computing power allocation of the K garbage image recognition branches in the garbage image recognition model according to the M garbage generation trend information to obtain M garbage classification recognition models.

[0078] Step S420: performing robot arm call allocation of the garbage classification control model according to the M garbage generation trend information to obtain M robot arm control models.

[0079] Step S430: The M garbage classification recognition models and the M robotic arm control models are communicatively connected according to the M sample functional area mappings to complete the construction of the automatic classification control unit.

[0080] Specifically, by deeply analyzing the M garbage generation trend information, the characteristics and distribution trends of different types of garbage can be accurately grasped. For those specific types of garbage that account for a large proportion of the overall garbage generation and have a high recognition complexity, such as plastic garbage, the corresponding recognition branches in the garbage image recognition model will be allocated more computing power resources. Because plastic garbage has various forms and complex material textures, more powerful computing power is required to accurately identify its characteristics and categories. For garbage types that are relatively easy to identify and have a small amount of generation, such as certain specific metal garbage, the computing power allocated is relatively small. In this way, based on the characteristics of different garbage types, the computing power of the K garbage image recognition branches is reasonably allocated and adjusted, and finally M garbage classification and recognition models are successfully obtained. Each model can be based on its optimized computing power configuration. More efficiently identify various types of garbage in the area.

[0081] Next, work is carried out based on the M garbage generation trend information, and the occurrence patterns and quantity scales of different garbage types in different scenarios are analyzed in detail, so as to determine the call allocation strategy of the robotic arm in the garbage classification control model. Taking paper garbage as an example, if paper garbage is frequently generated in a certain scenario and the single generation amount is large, then the call ratio parameter of the robotic arm for this type of garbage will be increased accordingly to ensure that there are a sufficient number of robotic arms with appropriate action parameter settings to handle the sorting of paper garbage in a timely manner. By carefully planning the call allocation of robotic arms for different types of garbage, M robotic arm control models were successfully generated, which accurately specified the operation details and call rules of the robotic arms for various types of garbage.

[0082] Finally, according to the division framework of M sample functional areas, the M garbage classification recognition models and M robotic arm control models obtained previously are mapped and connected. When garbage enters from a specific functional area, the garbage image is quickly and accurately identified according to the garbage classification recognition model corresponding to the area. After determining the type of garbage, the corresponding robotic arm control model is immediately activated through the mapping communication link, so that the robotic arm can efficiently complete the garbage sorting task according to the preset precise action parameters and calling rules, thereby successfully completing the construction of the automatic classification control unit, laying a solid foundation for the stable operation of the entire campus garbage automatic classification system.

[0083] In a possible implementation, step S420 further includes:

[0084] Step S421: generating M robot arm calling strategies for multiple garbage sorting robot arms on the garbage sorting conveyor belt according to the M garbage generation trend information, wherein each robot arm calling strategy includes K calling ratio parameters corresponding to the K types of sample garbage types.

[0085] Step S422: The M robot arm calling strategies and the K grasping control parameters are stored in association with each other according to the K types of sample garbage to obtain the M robot arm control models.

[0086] Specifically, we deeply explore the M garbage generation trend information, and this process uses data mining and statistical analysis algorithms. By mining the massive garbage generation data, we can accurately analyze the distribution patterns and quantity changes of different garbage types in different situations. We use cluster analysis algorithms to classify data with similar garbage generation patterns into one category, so as to gain a clearer insight into the generation characteristics of each garbage type. Based on these analysis results, we start to formulate M robot arm call strategies for multiple garbage sorting robot arms on the garbage sorting conveyor belt. For K types of sample garbage, we set K call ratio parameters for each robot arm call strategy. This process uses a multi-objective optimization algorithm to determine the optimal operation ratio of each robot arm for different garbage types, taking into account multiple objective factors such as garbage processing efficiency, robot arm load balancing, and sorting priority of different garbage types. For example, for garbage types with large generation volume and relatively regular and easy to grasp, such as certain types of plastic bottles, a higher call ratio parameter will be allocated to ensure that there are enough robot arm resources to invest in its sorting work to achieve the efficiency of the overall sorting process.

[0087] According to K types of sample garbage, M robot arm calling strategies are associated with K grasping control parameters and stored to construct M robot arm control models. When determining the grasping control parameters, motion planning algorithms and mechanical simulation algorithms are used. Through the motion planning algorithm, the optimal motion path of the robot arm is planned according to the shape, size, position and other information of the garbage to ensure that the robot arm can accurately grasp the garbage. At the same time, with the help of the mechanical simulation algorithm, the force and angle required for the robot arm to grasp different types of garbage are simulated to avoid the garbage slipping or damage to the robot arm due to excessive or insufficient grasping force. In this way, the robot arm calling strategy is closely combined with the grasping control parameters obtained through careful calculation and simulation to form a complete robot arm control model, so that each robot arm control model can perform intelligent and precise sorting operations for different types of garbage in a specific area, greatly improving the reliability and effectiveness of the automatic garbage sorting system on campus.

[0088] In a possible implementation, step S600 further includes:

[0089] Step S610: Before dumping campus garbage onto the garbage sorting conveyor belt, the garbage operation and maintenance equipment loads the functional area identification of the real-time upload area into the automatic sorting control unit.

[0090] Step S620: the automatic classification control unit receives and activates the real-time garbage classification recognition model and the real-time robotic arm control model according to the functional area identification of the real-time upload area and mapping the M garbage classification recognition models and the M robotic arm control models.

[0091] Step S630: After the real-time garbage classification recognition model identifies the garbage type based on the real-time garbage image on the garbage classification conveyor belt transmitted back by the image collector, the real-time robotic arm control model dispatches the multiple garbage sorting robotic arms to perform real-time sorting of the garbage on the garbage classification conveyor belt based on the garbage type recognition result.

[0092] Specifically, before dumping campus garbage on the garbage sorting conveyor belt, the garbage operation and maintenance equipment will perform a key preparatory work, that is, loading the functional area identification of the real-time upload area into the automatic classification control unit. First, the garbage operation and maintenance equipment accurately determines the real-time upload area where it is currently located through its built-in positioning system or communication connection with the campus area management system. This area can be one of the areas with specific functions such as the teaching area, living area, administrative area, and sports area on campus. Once the real-time upload area is determined, the garbage operation and maintenance equipment will obtain the corresponding functional area identification information from its storage unit or through interaction with the campus area identification database. This identification information is a unique code or description of the functional attributes of the area, such as "Teaching Area 01", "Living Area Building B", etc., which is used to clearly distinguish campus areas with different functions. Subsequently, the garbage operation and maintenance equipment uses wireless communication technology (such as Wi-Fi, Bluetooth, ZigBee or specific campus IoT communication protocols) to establish a connection link with the automatic classification control unit. Through this stable communication link, the functional area identification of the real-time upload area is accurately sent to the automatic classification control unit. After receiving the identification information, the automatic classification control unit immediately stores it in a specific cache area or configuration register so that it can be quickly called in the subsequent automatic classification and real-time sorting of garbage. The importance of this operation lies in that it provides key pre-information for the automatic classification control unit, allowing it to know in advance which functional area the garbage to be processed comes from, so that it can quickly make targeted preparations based on the pre-set garbage generation trend information, garbage classification identification model, and robot arm control model parameters of the area, so as to achieve accurate classification and processing of garbage in different areas, and improve the operating efficiency and accuracy of the entire campus garbage automatic classification system.

[0093] After receiving the functional area identification of the real-time upload area uploaded by the garbage operation and maintenance equipment, the automatic classification control unit quickly starts a series of key operations to achieve accurate garbage classification identification and robotic arm control. First, the automatic classification control unit maintains a mapping table, which records in detail the one-to-one correspondence between M sample functional areas and M garbage classification identification models and M robotic arm control models. When receiving the functional area identification, the automatic classification control unit immediately searches for matching operations in the mapping table. For example, if the received functional area identification is "teaching area A", the corresponding garbage classification identification model "teaching area A garbage classification identification model" and robotic arm control model "teaching area A robotic arm control model" are found by quickly searching the mapping table. Once the corresponding model is determined, the automatic classification control unit sends an activation instruction to the relevant computing resources and hardware devices. For the garbage classification identification model, the activation operation includes allocating sufficient computing resources (such as CPU cores, GPU resources, etc.) to it, loading pre-trained model parameters and algorithm logic, so that it enters the working state and is ready to perform real-time recognition and analysis on the garbage images that will be transmitted from the garbage classification conveyor belt. At the same time, for the robot control model, the activation process involves initializing the hardware driver related to the robot control, setting the initial working parameters, and establishing a communication link with the garbage sorting robot to ensure that the robot can receive and execute subsequent control instructions. Through such a precise mapping activation mechanism, the automatic classification control unit can quickly activate the classification recognition model and robot control model that best suits the characteristics of the garbage in the area where the garbage comes from, laying a solid foundation for the subsequent efficient and accurate automatic classification and real-time sorting of garbage, ensuring that the entire system can quickly respond to the garbage disposal needs of different areas and realize intelligent campus garbage management.

[0094] The image collector installed on the garbage sorting conveyor belt continuously collects images of the garbage on the conveyor belt, and quickly transmits the real-time garbage image back to the automatic sorting control unit through the previously constructed one-way communication channel. After receiving the image, the automatic sorting control unit immediately transmits it to the activated real-time garbage sorting recognition model. The real-time garbage sorting recognition model quickly and accurately analyzes and recognizes the input real-time garbage image based on its internal pre-trained deep learning algorithm and a large amount of sample data. It extracts and compares multiple features of the garbage in the image, such as shape, color, texture, size, etc., and uses advanced image recognition technology, convolutional neural network (CNN), to accurately determine the type of garbage, such as plastic bottles, paper, metal cans or other types of garbage.

[0095] Once the garbage type identification is completed, the real-time garbage classification identification model transmits the identification results to the real-time robotic arm control model. Based on the garbage type information received, the real-time robotic arm control model quickly searches for the corresponding robotic arm scheduling instructions in its internally stored control strategy library. These instructions are formulated based on the optimization parameters obtained from the previous grasping control tests on different sample garbage types, including detailed control information such as the optimal grasping force, grasping angle, grasping speed, and robotic arm motion trajectory for each type of garbage. According to the scheduling instructions, the real-time robotic arm control model accurately controls the actions of multiple garbage sorting robotic arms. For example, when the garbage is identified as a plastic bottle, the robotic arm will extend the robotic arm at a suitable angle and force according to the preset grasping parameters for the plastic bottle, accurately grasp the plastic bottle, and place it in the corresponding garbage classification collection area. The whole process is completed in a very short time, realizing the real-time sorting operation of garbage on the garbage sorting conveyor belt, ensuring that different types of garbage can be quickly and accurately separated and collected, greatly improving the efficiency and accuracy of campus garbage disposal, effectively reducing the burden of manual classification, and improving the intelligent level of campus environmental management.

[0096] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0097] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0098] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A campus garbage automatic classification method integrating artificial intelligence image recognition, characterized in that: The method comprises: Analyze the garbage generation pattern of the target campus and obtain M garbage generation trend information of M sample functional areas; Obtain junk image archive logs through local data calls; Constructing and generating a garbage image recognition model and a garbage classification control model according to the garbage image archive log; Optimizing resource calls of the garbage image recognition model and the garbage classification control model according to the M garbage generation trend information, and completing the construction of the automatic classification control unit; Constructing a one-way communication channel between the image collector of the garbage sorting conveyor belt and the automatic sorting control unit; The automatic classification control unit receives and performs automatic classification and real-time sorting of garbage according to the real-time garbage image on the garbage classification conveyor belt transmitted back by the image collector; Analyzing the garbage generation pattern of a target campus to obtain M garbage generation trend information of M sample functional areas, the method comprising: Interactively obtain a campus space planning map of the target campus; Preset campus area types, wherein the campus area types include the M sample functional areas; The campus space planning map is traversed using the campus area type to divide the target campus into a plurality of campus areas, wherein the plurality of campus areas have a plurality of functional area identifiers; The garbage generation patterns of the multiple campus areas are analyzed according to the multiple functional area identifiers to obtain M garbage generation trend information corresponding to the M types of sample functional areas.

2. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 1 is characterized in that: The method includes: performing garbage generation pattern analysis on the multiple campus areas according to the multiple functional area identifiers to obtain M garbage generation trend information corresponding to the M sample functional areas. According to the multiple functional area identifiers, the multiple campus areas are divided into M groups of campus areas corresponding to the M types of sample functional areas; Calling W regional garbage archive logs of W campus regions in the first group of campus regions; After performing time-series alignment on the W regional garbage archive logs, performing data fusion to obtain a first sample archive log; Outputting a first garbage generation trend graph according to the first sample archive log fitting, wherein the first garbage generation trend graph includes a plurality of garbage time-varying curves of a plurality of sample campus garbage; Performing a change analysis on the proportion of multiple types of garbage generated on the first garbage generation trend graph, and outputting a first garbage generation trend; By analogy, the garbage generation patterns of the multiple campus areas are analyzed according to the multiple functional area identifiers to obtain information corresponding to the M garbage generation trends.

3. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 2 is characterized in that: Outputting a first garbage generation trend graph according to the first sample archive log fitting, the method comprising: Splitting the first sample archive log according to a preset time series division scale to obtain a plurality of first-stage archive logs; Decomposing the plurality of first-stage archive logs according to garbage types to obtain a plurality of groups of garbage generation ratios of a plurality of garbage generation nodes; Reorganize the multiple groups of garbage generation ratios according to garbage types to obtain multiple garbage ratio time series data of the multiple sample campus garbage; The plurality of garbage ratio time series data are used to fit the plurality of garbage time-varying curves; The first garbage generation trend graph is output by fusing the multiple garbage time-varying curves through image.

4. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 1 is characterized in that: A garbage image recognition model and a garbage classification control model are constructed and generated according to the garbage image archive log, and the method includes: By calling local data, obtaining area information of multiple regions of the multiple campus regions; Calculate and obtain multiple image sampling weights according to the multiple region area information; Perform image sampling of multiple regional image archive logs in the junk image archive log according to the multiple image sampling weights to obtain multiple sample junk image sets; Aggregating the M pieces of garbage generation trend information by garbage type to obtain K types of sample garbage; Performing manual semantic identification of the plurality of sample garbage image sets based on the K types of sample garbage as constraints to obtain a plurality of sample identified image sets; The junk image recognition model is constructed based on the multiple sample junk image sets and the multiple sample identification image sets, wherein the junk image recognition model is composed of K junk image recognition branches connected in parallel, and the K junk image recognition branches are used to perform directional recognition and marking of the K types of sample junk types; The K types of sample garbage are subjected to a grasping control test, and after obtaining K grasping control parameters, the K types of sample garbage and the K grasping control parameters are associated and stored to complete the construction of the garbage classification control model.

5. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 4 is characterized in that: According to the M pieces of garbage generation trend information, resource call optimization of the garbage image recognition model and the garbage classification control model is performed to complete the construction of the automatic classification control unit. The method includes: According to the M garbage generation trend information, the computing power allocation of the K garbage image recognition branches in the garbage image recognition model is adjusted to obtain M garbage classification recognition models; According to the M pieces of garbage generation trend information, the robot arm call allocation of the garbage classification control model is performed to obtain M robot arm control models; The M garbage classification recognition models and the M robotic arm control models are communicatively connected according to the M sample functional area mappings to complete the construction of the automatic classification control unit.

6. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 5 is characterized in that: According to the M pieces of garbage generation trend information, the robot arm call allocation of the garbage classification control model is performed to obtain M robot arm control models, and the method includes: Generate M robot arm calling strategies for multiple garbage sorting robot arms on the garbage sorting conveyor belt according to the M garbage generation trend information, wherein each robot arm calling strategy includes K calling ratio parameters corresponding to the K types of sample garbage; The M robot arm calling strategies and the K grasping control parameters are stored in association with the K types of sample garbage to obtain the M robot arm control models.

7. The automatic campus garbage classification method integrating artificial intelligence image recognition as claimed in claim 6 is characterized in that: The automatic classification control unit receives and performs automatic classification and real-time sorting of garbage according to the real-time garbage image on the garbage classification conveyor belt transmitted back by the image collector, and the method includes: Before dumping campus garbage onto the garbage sorting conveyor belt, the garbage operation and maintenance equipment loads the functional area identification of the real-time upload area into the automatic sorting control unit; The automatic classification control unit receives and maps and activates the real-time garbage classification recognition model and the real-time robotic arm control model in the M garbage classification recognition models and the M robotic arm control models according to the functional area identification of the real-time upload area; After the real-time garbage classification recognition model identifies the garbage type based on the real-time garbage image on the garbage classification conveyor belt sent back by the image collector, the real-time robotic arm control model dispatches the multiple garbage sorting robotic arms to perform real-time sorting of the garbage on the garbage classification conveyor belt based on the garbage type recognition result.

Images