Visual union monitoring system and method for household garbage classification source based on AI analysis

Abstract

The invention provides an AI analysis-based visual union monitoring system and method for a household garbage classification source, and relates to the technical field of garbage classification monitoring. A preset monitoring area is obtained, in-can monitoring and out-of-can monitoring are carried out, acquired data are obtained, data fusion is carried out, and area fusion acquisition information is obtained; obtaining and fusing abnormal data outside a region barrel and abnormal data inside the region barrel of the preset monitoring region, calculating a region anomaly classification coefficient, and carrying out depth anomaly classification judgment to obtain region anomaly classification judgment information; according to the method, classification monitoring and comprehensive monitoring of the inside and outside of the barrel can be carried out on the preset monitoring area, and then comprehensive classification management of the area monitoring network is achieved; and the efficiency and the accuracy of garbage classification linkage monitoring are improved.

Description

Technical Field

[0001] The present invention proposes a visual monitoring system and method for the source of household waste classification based on AI analysis, which relates to the technical field of waste classification monitoring, and specifically to the technical field of visual monitoring of the source of household waste classification based on AI analysis. Background Art

[0002] As environmental awareness grows, garbage classification has become an important part of residents' daily lives. Traditional garbage classification monitoring uses image recognition technology to accurately capture the behavior of lost packages outside the bin and upload it to the backend system in real time. However, it is difficult to classify and identify garbage inside the bin, lost packages outside the bin, and conduct comprehensive monitoring and linkage analysis inside and outside the bin. The monitoring efficiency is low and the accuracy is low. It is difficult to conduct targeted monitoring and analysis and lacks flexibility. Summary of the invention

[0003] The present invention provides a visual monitoring system and method for the source of household waste classification based on AI analysis to solve the above problems:

[0004] The present invention proposes a visual monitoring system and method for the source of household waste classification based on AI analysis, the method comprising:

[0005] S1. Obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the area and the data collected inside the area, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network;

[0006] S2. Obtain the abnormal data outside the regional bucket and the abnormal data inside the regional bucket of the preset monitoring area, perform data fusion to obtain regional abnormal fusion data, calculate the regional abnormal classification coefficient, perform regional abnormality determination, perform deep abnormality classification determination on the obtained regional abnormality determination information, and obtain regional abnormality classification determination information;

[0007] S3. Obtain the ranking of classified abnormal areas, obtain the ranking change of the preset monitoring area according to the preset monitoring time series, calculate the time series evaluation abnormality coefficient, and perform abnormal marking and warning of the abnormal preset monitoring area.

[0008] Furthermore, the S1 includes:

[0009] Acquire regional camera information, and divide the regional camera information into external camera information and internal camera information according to the regional camera position;

[0010] The camera outside the barrel collects data from the area outside the barrel to obtain the collected data outside the barrel.

[0011] The camera in the barrel collects data from the area in the barrel to obtain the collected data in the regional barrel;

[0012] Fusing the data collected outside the regional bucket with the data collected inside the regional bucket to obtain regional fusion collection information;

[0013] Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network;

[0014] Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

[0015] Further, the S2 includes:

[0016] Performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information;

[0017] Extract feature data of out-of-bucket packet loss data from the regional out-of-bucket collection data of the preset monitoring area to obtain abnormal data outside the regional bucket;

[0018] Extract feature data of unclassified data in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket;

[0019] Merging the abnormal data outside the regional bucket with the abnormal data inside the regional bucket to obtain regional abnormal fusion data;

[0020] Calculating a regional anomaly classification coefficient based on the regional anomaly fusion data;

[0021] Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result;

[0022] Acquiring regional anomaly determination information according to the regional anomaly comparison result;

[0023] Perform deep abnormality classification and determination on the regional abnormality determination information to obtain the regional abnormality classification and determination information.

[0024] Furthermore, the performing of deep abnormality classification determination on the regional abnormality determination information to obtain the regional abnormality classification determination information includes:

[0025] According to the abnormal data outside the regional bucket, the abnormal data inside the regional bucket and the regional abnormal classification coefficient, the abnormal coefficient outside the regional bucket and the abnormal coefficient inside the regional bucket are calculated;

[0026] Compare the regional bucket out-of-bucket abnormality coefficient with the preset regional bucket out-of-bucket abnormality threshold, obtain the bucket out-of-bucket comparison result, perform an abnormality determination on the regional bucket out-of-bucket according to the bucket out-of-bucket comparison result, and obtain the regional bucket out-of-bucket determination information;

[0027] Compare the regional bucket anomaly coefficient with a preset regional bucket anomaly threshold to obtain an intra-bucket comparison result, perform anomaly determination on the regional bucket according to the intra-bucket comparison result, and obtain intra-bucket determination information;

[0028] The regional bucket outside determination information and the regional bucket inside determination information are the regional anomaly classification determination information.

[0029] Further, the S3 includes:

[0030] All regional anomaly coefficients are ranked from small to large to obtain the ranking of classified anomaly regions;

[0031] Obtain a preset monitoring time sequence, and obtain a ranking change of a classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence;

[0032] Calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change;

[0033] Compare the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing determination abnormality comparison result;

[0034] According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

[0035] Furthermore, the system comprises:

[0036] The regional monitoring and collection module is used to obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the regional bucket and the data collected inside the regional bucket, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network;

[0037] The inside-outside-bucket anomaly analysis module is used to obtain the outside-outside-bucket anomaly data and the inside-outside-bucket anomaly data of the preset monitoring area, obtain the regional anomaly fusion data after data fusion, calculate the regional anomaly classification coefficient, make regional anomaly judgment, perform deep anomaly classification judgment on the obtained regional anomaly judgment information, and obtain regional anomaly classification judgment information;

[0038] The regional abnormal change analysis module is used to obtain the ranking of classified abnormal areas, obtain the ranking change of the preset monitoring area according to the preset monitoring time sequence, calculate the time sequence evaluation abnormal coefficient, and perform abnormal marking and early warning of the abnormal preset monitoring area.

[0039] Furthermore, the regional monitoring and acquisition module includes:

[0040] A camera division module is used to obtain regional camera information and divide the regional camera information into external camera information and internal camera information according to the regional camera position;

[0041] The outside-barrel collection module is used to collect data from the outside-barrel area through the outside-barrel camera to obtain the outside-barrel collection data of the area;

[0042] The in-barrel collection module is used to collect data from the area in the barrel through the in-barrel camera to obtain the collected data in the regional barrel;

[0043] A regional network acquisition module, used to fuse the data collected outside the regional bucket with the data collected inside the regional bucket to obtain regional fusion collection information;

[0044] Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network;

[0045] Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

[0046] Furthermore, the inside and outside bucket abnormality analysis module includes:

[0047] A feature monitoring module, used for performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information;

[0048] The out-of-bucket anomaly extraction module is used to extract feature data of out-of-bucket packet loss data from the regional out-of-bucket collection data of the preset monitoring area to obtain the regional out-of-bucket anomaly data;

[0049] The in-bucket anomaly extraction module is used to extract feature data of unclassified data in the bucket from the data collected in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket;

[0050] An anomaly fusion module, used to fuse the anomaly data outside the regional bucket with the anomaly data inside the regional bucket to obtain regional anomaly fusion data;

[0051] A regional anomaly analysis module, used for calculating a regional anomaly classification coefficient based on the regional anomaly fusion data;

[0052] Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result;

[0053] Acquiring regional anomaly determination information according to the regional anomaly comparison result;

[0054] The deep classification analysis module is used to perform deep anomaly classification and determination on the regional anomaly determination information to obtain the regional anomaly classification and determination information.

[0055] Furthermore, the deep classification analysis module includes:

[0056] A deep anomaly calculation module is used to calculate an anomaly coefficient outside the regional bucket and an anomaly coefficient inside the regional bucket based on the anomaly data outside the regional bucket and the anomaly data inside the regional bucket combined with the regional anomaly classification coefficient;

[0057] The out-of-bucket anomaly analysis module is used to compare the out-of-bucket anomaly coefficient of the region with the preset out-of-bucket anomaly threshold of the region, obtain the out-of-bucket comparison result, make an anomaly judgment on the out-of-bucket region according to the out-of-bucket comparison result, and obtain the out-of-bucket judgment information of the region;

[0058] The in-bucket anomaly analysis module is used to compare the in-bucket anomaly coefficient of the regional bucket with the preset in-bucket anomaly threshold of the regional bucket to obtain an in-bucket comparison result, and perform an anomaly determination on the regional bucket according to the in-bucket comparison result to obtain in-bucket determination information of the regional bucket;

[0059] The classification anomaly determination module is used for the determination information outside the regional bucket and the determination information inside the regional bucket, that is, the regional anomaly classification determination information.

[0060] Furthermore, the regional abnormal change analysis module includes:

[0061] The anomaly ranking module is used to rank all regional anomaly coefficients in ascending order to obtain the ranking of classified anomaly regions;

[0062] A ranking change monitoring module is used to obtain a preset monitoring time sequence, and obtain the ranking change amount of the classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence;

[0063] A time series anomaly calculation module, used to calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change;

[0064] A judgment and early warning module, used for comparing the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing judgment abnormality comparison result;

[0065] According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

[0066] Beneficial effects of the present invention: Through real-time monitoring and AI analysis, the system can accurately identify errors inside and outside the garbage classification bin and improve the accuracy of garbage classification. The system can automatically identify and mark abnormal areas, reducing the workload of manual inspections and improving supervision efficiency. By improving the accuracy and efficiency of garbage classification, resource waste and environmental pollution in the garbage disposal process can be reduced. The system's early warning and feedback mechanism can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification. The large amount of data collected by the system can provide comprehensive analysis data for institutions or platforms to formulate garbage classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 Schematic diagram of the visual monitoring method for the source of household waste classification based on AI analysis. DETAILED DESCRIPTION

[0068] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0069] In one embodiment of the present invention, a visual monitoring system and method for the source of household waste classification based on AI analysis is proposed in the present invention, and the method includes:

[0070] S1. Obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the area and the data collected inside the area, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network;

[0071] S2. Obtain the abnormal data outside the regional bucket and the abnormal data inside the regional bucket of the preset monitoring area, perform data fusion to obtain regional abnormal fusion data, calculate the regional abnormal classification coefficient, perform regional abnormality determination, perform deep abnormality classification determination on the obtained regional abnormality determination information, and obtain regional abnormality classification determination information;

[0072] S3, obtain the ranking of the classified abnormal area, obtain the ranking change of the preset monitoring area according to the preset monitoring time sequence, calculate the time sequence evaluation abnormal coefficient, and perform abnormal marking and warning of the abnormal preset monitoring area, such as Figure 1 shown.

[0073] The working principle of the above technical solution is as follows: obtain the preset monitoring areas, which include the garbage delivery points or the locations of the classified garbage bins. Perform real-time monitoring inside and outside the garbage bins in the preset monitoring area. The monitoring inside the bin mainly focuses on the classification of garbage, while the monitoring outside the bin focuses on the delivery behavior, the environment around the garbage bin, etc. The data collected inside and outside the bin are fused to form regional fusion collection information. This step can comprehensively analyze the classification and delivery of garbage in the monitoring area. Extract abnormal data from the fused data, including abnormal data inside the bin and abnormal data outside the bin. The abnormal data inside and outside the bin are fused again to form regional abnormal fusion data. Then, calculate the regional abnormal classification coefficient, which is used to evaluate the severity and type of abnormality. Use AI methods to deeply analyze the abnormal data, make more detailed classification judgments, and obtain regional abnormal classification judgment information. According to the abnormal classification judgment information, rank each monitoring area to reflect the degree of abnormality in each area. Combined with the preset monitoring time series (such as daily, weekly, etc.), obtain the ranking change of each area, and calculate the time series evaluation abnormality coefficient. This step can evaluate the persistence and development trend of abnormalities. Based on the time series evaluation anomaly coefficient, the abnormal preset monitoring area is marked and the early warning mechanism is triggered so that timely measures can be taken to intervene.

[0074] As environmental awareness grows, garbage classification has become an important part of residents' daily lives. The present invention proposes an intelligent monitoring solution for lost packages outside the garbage bin based on AI technology. The solution uses image recognition technology to accurately capture the behavior of lost packages outside the garbage bin and upload it to the background system in real time, aiming to strengthen residents' awareness of garbage classification, improve garbage disposal efficiency, and jointly maintain a beautiful environment.

[0075] By installing cameras and using AI technology to monitor the loss of packages outside the bin in real time, relevant data will be uploaded to the backend system for processing in a timely manner. Through the implementation of this solution, it is expected to significantly improve the accuracy and efficiency of garbage classification and reduce littering.

[0076] The camera is used to photograph the garbage in the bin, and AI technology is used to detect whether the garbage has been sorted and whether it is qualified.

[0077] Cooperate with the camera at the garbage recycling site to monitor the garbage classification situation at the current site and whether there is any bag lost outside the garbage bin. The playback function can accurately detect the individual behavior of bag lost outside the bin.

[0078] Administrators can use the backend management program to set the community's recycling point name, point type, operation mode and status.

[0079] The system will automatically report the captured packet loss outside the barrel and collect anomalies for today, this week, and this month.

[0080] The abnormal display of packet loss outside the bucket includes the community, the delivery point, the warning type and the abnormal picture.

[0081] The system supports multi-channel input function and can adapt to monitoring under multiple cameras, ensuring that the system processing speed is not less than 5 frames per second to meet real-time requirements, ensure the timeliness and accuracy of data, and the output data is a picture with a marked box.

[0082] Real-time monitoring of site staff can be achieved, and the work of site advocates can be supervised through existing cameras to determine whether the advocates are online at the specified time to provide garbage sorting services to community residents.

[0083] The data and information involved in this application have been obtained with full consent and authorization, and the collection, use and processing of relevant information comply with the relevant laws, regulations and standards of relevant countries and regions.

[0084] The technical effects of the above technical solutions are as follows: through real-time monitoring and AI analysis, the system can accurately identify errors inside and outside the garbage classification bin and improve the accuracy of garbage classification. The system can automatically identify and mark abnormal areas, reducing the workload of manual inspections and improving supervision efficiency. By improving the accuracy and efficiency of garbage classification, resource waste and environmental pollution in the garbage disposal process can be reduced. The system's early warning and feedback mechanism can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification. The large amount of data collected by the system can provide comprehensive analysis data for institutions or platforms to formulate garbage classification.

[0085] In one embodiment of the present invention, the S1 includes:

[0086] Acquire regional camera information, and divide the regional camera information into external camera information and internal camera information according to the regional camera position;

[0087] The camera outside the barrel collects data from the area outside the barrel to obtain the collected data outside the barrel.

[0088] The camera inside the barrel collects data from the area inside the barrel to obtain the collected data in the area barrel; the camera inside the barrel is a camera that takes pictures of the garbage in the barrel, not a camera installed in the barrel;

[0089] The data collected outside the regional barrel is fused with the data collected inside the regional barrel to obtain regional fusion collection information; the regional fusion collection information includes overlapping the same parts and splicing the different parts to obtain regional fusion collection data, aiming to reduce or avoid monitoring blind spots. The regional camera includes a barrel camera and a barrel camera of a preset monitoring area, and the regional fusion collection information includes every combined frame image without blind spots inside and outside the barrel;

[0090] Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network;

[0091] Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

[0092] The working principle of the above technical solution is: obtain the location information of all cameras in the preset monitoring area. According to the location of the camera, the camera information is divided into the camera information outside the bin and the camera information inside the bin. It should be noted here that the camera inside the bin is not directly installed inside the trash can, but refers to the camera whose shooting range covers the internal area of ​​the trash can. The camera outside the bin collects data from the external area of ​​the trash can, including the placement behavior, personnel flow, etc., to obtain the regional external bin collection data. The camera inside the bin collects data from the internal area of ​​the trash can, including the classification of garbage, the overflow status of the trash can, etc., to obtain the regional internal bin collection data. The external bin collection data and the internal bin collection data are fused. The fusion process includes overlapping the same parts and splicing different parts to obtain regional fusion collection information without blind spots. This step is to reduce or avoid monitoring blind spots and ensure the comprehensiveness and accuracy of monitoring. The regional fusion collection information is presented in the form of each combined frame image, including a blind spot view inside and outside the bin. After obtaining the regional fusion collection information of all preset monitoring areas, the system builds a monitoring area network. The monitoring area network consists of multiple monitoring areas, each of which contains complete information collected from inside and outside the bin. Comprehensive monitoring of the monitoring area network is carried out to analyze the comprehensive monitoring data of each monitoring area in real time. The comprehensive monitoring data includes information on the overflow status of the garbage bin, classification accuracy, and placement behavior.

[0093] The technical effect of the above technical solution is: through the combined use of cameras inside and outside the barrel, the system can realize comprehensive monitoring of the source of garbage classification to ensure coverage without blind spots. And it can carry out targeted monitoring and processing of classification inside the barrel and packet loss outside the barrel, while considering comprehensive abnormal analysis of classification inside the barrel and packet loss outside the barrel; data fusion processing enables the system to obtain more comprehensive monitoring information, thereby improving the judgment accuracy of garbage classification accuracy. The system can monitor the situation of each monitoring area in real time, discover and deal with abnormal problems in time, and improve supervision efficiency. By building a monitoring area network, the system can realize intelligent management. The system can detect the overflowing state of the garbage bin in time to avoid waste of resources and environmental pollution caused by garbage overflow. The real-time monitoring and feedback mechanism of the system can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification.

[0094] In one embodiment of the present invention, S2 includes:

[0095] Performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information;

[0096] Extract feature data of out-of-bucket packet loss data from the regional out-of-bucket collection data of the preset monitoring area to obtain abnormal data outside the regional bucket;

[0097] Extract feature data of unclassified data in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket;

[0098] Merging the abnormal data outside the regional bucket with the abnormal data inside the regional bucket to obtain regional abnormal fusion data;

[0099] Calculating a regional anomaly classification coefficient based on the regional anomaly fusion data;

[0100] The calculation formula of the regional anomaly classification coefficient is:

[0101]

[0102] Among them, QY is the regional anomaly classification coefficient, N is the total number of regional monitoring (camera batches), and w out is the preset weight of the abnormal data outside the bucket, which is used to adjust the importance of the abnormal data outside the bucket in the calculation. in It is the preset weight of the abnormal data in the bucket, which is used to adjust the importance of the abnormal data in the bucket in the calculation. outi is the amount of abnormal data outside the bucket monitored for the i-th time (obtained from data collection), which can be quantified according to specific abnormality types, frequency, or duration, etc. ini is the amount of abnormal data in the bucket monitored for the i-th time (obtained from data collection), which can also be quantified based on specific abnormality types, frequency, duration, and other factors. i is the amount of data overlapping the data collected inside and outside the bucket during the i-th monitoring, which is used to measure the contribution of fusion information to reducing or avoiding monitoring blind spots. z The amount of data that overlaps the total monitored data collected inside and outside the bucket; 0< <1.

[0103] Quantification of out-of-bucket anomaly types includes packet loss data anomalies and non-delivery time anomalies;

[0104] Packet loss data anomalies include garbage that is directly discarded without being classified in the data collected outside the bin. The type and quantity of the garbage can be determined through image recognition technology. For example, the system can identify which garbage is recyclable, hazardous waste, wet garbage (kitchen waste) or dry garbage (other garbage), and count the quantity of each type of garbage. According to the importance and quantity of the garbage type, a quantitative standard can be set, such as "Every time more than 5 recyclables are found to be directly discarded without being classified, it will be recorded as a packet loss data anomaly."

[0105] The non-delivery time anomaly includes the timestamp information of the monitoring video. The system can determine whether the garbage delivery occurs within the specified delivery time. For the garbage delivery behavior during the non-delivery time, it can be recorded as one non-delivery time anomaly. If this behavior occurs frequently, such as more than 3 times a week, it can be regarded as three non-delivery time anomalies.

[0106] When there is a packet loss data anomaly or a non-delivery time anomaly, it is an out-of-bucket anomaly type.

[0107] The quantification of anomaly types within a bucket includes unclassified data anomalies and placement location anomalies;

[0108] Unclassified data anomalies include various types of mixed garbage in the collected data in the barrel, which are identified and classified by the system through image recognition technology. Quantitative standards can be set according to the severity of mixed placement (such as the number and proportion of mixed garbage types, etc.). For example, "each time monitoring finds that more than 3 different types of garbage are mixed in the barrel, it is an unclassified data anomaly."

[0109] Abnormal placement includes judging whether the garbage is placed correctly according to the preset placement standards. If the garbage is placed in the wrong location, it can be recorded as an abnormality. If the same type of abnormal placement occurs frequently, such as more than 5 times in the same location per week, it is considered a placement abnormality.

[0110] When there is an unclassified data anomaly or an abnormal placement location, it is an in-bucket anomaly type.

[0111] Abnormal frequency quantification includes the number of abnormal events that occur per unit time. For abnormal types outside the bucket and inside the bucket, you can set the quantification standard of abnormal frequency separately. For example, "more than 3 abnormal packet loss data outside the bucket occurs per week" or "more than 5 abnormal unclassified data inside the bucket occurs per month" are abnormal frequency data outside the bucket and abnormal frequency data inside the bucket respectively.

[0112] The abnormal duration includes the length of time from the start to the end of the abnormal event. For the abnormal types outside the bucket and inside the bucket, you can set the quantitative standards for the abnormal duration separately. For example, you can set "single abnormal duration of packet loss data outside the bucket exceeds 30 minutes" or "single abnormal duration of unclassified data inside the bucket exceeds 1 hour" as the abnormal duration data outside the bucket or inside the bucket (based on the data that can be collected).

[0113] The differences between the quantification of anomalies inside and outside the bucket include:

[0114] The quantification of out-of-bucket anomalies includes the disposal of garbage outside the disposal time, direct disposal without classification, etc. These anomaly types mainly reflect the out-of-bucket anomaly analysis data.

[0115] Abnormal quantification in the bucket includes mixed delivery, incorrect delivery location, etc. These abnormal types mainly reflect the monitoring data during the garbage classification collection and processing process.

[0116] The abnormal data volume can be one of the abnormal data of type, frequency and time. The specific one to be selected can be determined according to actual needs.

[0117] Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result;

[0118] Acquiring regional anomaly determination information according to the regional anomaly comparison result;

[0119] Perform deep abnormality classification and determination on the regional abnormality determination information to obtain the regional abnormality classification and determination information.

[0120] The working principle of the above technical solution is as follows: based on the regional fusion collection information, the system performs feature monitoring on each preset monitoring area in the monitoring area network. Feature monitoring mainly focuses on the preset abnormal features inside and outside the bucket (packet loss data outside the bucket and unclassified data inside the bucket). For the regional outside-bucket collection data of the preset monitoring area, the system extracts feature data of the packet loss data outside the bucket. Packet loss data may include behavior data such as garbage directly discarded without classification and garbage delivery during non-delivery time, which are all considered as abnormal data outside the bucket. For the regional inside-bucket collection data of the preset monitoring area, the system extracts feature data of the unclassified data inside the bucket. Unclassified data may include various types of mixed garbage, garbage delivered in incorrect locations, etc., which are all considered as abnormal data inside the bucket. The abnormal data outside the bucket and the abnormal data inside the bucket are fused to form regional abnormal fusion data. This step can understand the abnormal behavior or status in the monitoring area separately or comprehensively. Based on the regional abnormal fusion data, the system calculates the regional abnormal classification coefficient. This coefficient is used to evaluate the severity and type of the abnormality. The numerator is the sum of the weighted evaluation values ​​of the abnormal data inside and outside the bucket for each monitoring, and multiplying it by the corresponding evaluation value of the overlap degree of the fusion information.

[0121] The denominator of the above formula is the sum of the weight factors for each monitoring. out and w in is constant, so the denominator is equal to M×(w out +w in ).pass The comprehensive abnormal classification evaluation coefficients inside and outside the buckets were calculated comprehensively under the corresponding weights of each monitoring of the region;

[0122] Compare the regional anomaly classification coefficient with the preset regional anomaly threshold to obtain the regional anomaly comparison result. If the anomaly classification coefficient exceeds the threshold, it is determined that the monitoring area is abnormal. Based on the regional anomaly comparison result, the system obtains regional anomaly determination information, which includes whether the area is abnormal. Perform in-depth analysis of the regional anomaly determination information and make more detailed classification determinations. This step can more accurately identify the specific type of abnormal data, etc.

[0123] The technical effect of the above technical solution is: through feature monitoring and abnormal data extraction, the system can more accurately identify abnormal behaviors or states at the source of garbage classification and improve the accuracy of monitoring. Data fusion and calculation of abnormal classification coefficients enable the system to comprehensively evaluate abnormal conditions in the monitoring area and provide comprehensive abnormal judgment information. Through abnormal comparison and judgment, the system can promptly discover and mark abnormal areas and trigger early warning mechanisms so that timely measures can be taken to intervene. Deep abnormal classification and judgment provide support for intelligent management, enabling the system to formulate targeted treatment measures according to the type of abnormality and improve management efficiency. The system's real-time monitoring and early warning mechanism can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification. By promptly discovering and handling abnormal behaviors, the system helps to reduce resource waste and environmental pollution.

[0124] In one embodiment of the present invention, the step of performing deep abnormality classification determination on the regional abnormality determination information to obtain the regional abnormality classification determination information includes:

[0125] According to the abnormal data outside the regional bucket, the abnormal data inside the regional bucket and the regional abnormal classification coefficient, the abnormal coefficient outside the regional bucket and the abnormal coefficient inside the regional bucket are calculated;

[0126] The calculation formula of the abnormal coefficient outside the regional bucket is:

[0127]

[0128] Among them, TW is the regional out-of-bucket anomaly coefficient, QY is the regional anomaly classification coefficient, and w out is the preset weight of the abnormal data outside the bucket, which is used to adjust the importance of the abnormal data outside the bucket in the calculation.outd is the amount of abnormal data outside the bucket in the current region, M is the total number of all regions, and D outk is the amount of abnormal data outside the bucket of the kth region;

[0129] The calculation formula of the anomaly coefficient in the regional bucket is:

[0130]

[0131] Among them, TN is the regional bucket anomaly coefficient, QY is the regional anomaly classification coefficient, and w in It is the preset weight of the abnormal data in the bucket, which is used to adjust the importance of the abnormal data in the bucket in the calculation. ind is the amount of abnormal data in the bucket of the current region, M is the total number of all regions, and D ink is the amount of abnormal data in the bucket of the kth region;

[0132] Compare the regional bucket out-of-bucket abnormality coefficient with the preset regional bucket out-of-bucket abnormality threshold, obtain the bucket out-of-bucket comparison result, perform an abnormality determination on the regional bucket out-of-bucket according to the bucket out-of-bucket comparison result, and obtain the regional bucket out-of-bucket determination information;

[0133] Compare the regional bucket anomaly coefficient with a preset regional bucket anomaly threshold to obtain an intra-bucket comparison result, perform anomaly determination on the regional bucket according to the intra-bucket comparison result, and obtain intra-bucket determination information;

[0134] The regional bucket outside determination information and the regional bucket inside determination information are the regional anomaly classification determination information.

[0135] The working principle of the above technical solution is: according to the abnormal data outside the regional bucket and the abnormal data inside the regional bucket, combined with the regional abnormal classification coefficient calculated previously, the abnormal coefficient outside the regional bucket and the abnormal coefficient inside the regional bucket are calculated respectively. These two coefficients reflect the severity or importance of the abnormalities outside the bucket and inside the bucket respectively.

[0136] The numerator of the above-mentioned regional bucket anomaly coefficient and regional bucket anomaly coefficient mainly reflects the relative importance or scale of the abnormal data inside and outside the bucket of a specific region. in) by multiplying. The larger the numerator, the higher the proportion or importance of the abnormal data inside and outside the bucket of the current region in the overall abnormal data. It can be identified which regions have more significant or worthy of attention in and out of the bucket. The denominator is calculated by summing up the abnormal data inside and outside the bucket of all regions. It ensures that the regional in and out of the bucket anomaly coefficient is a proportional coefficient or relative value, rather than an absolute value. The calculation of the denominator can compare the in and out of the bucket anomaly data of different regions on the same scale. As a sum term, the denominator plays a normalization role, making the regional in and out of the bucket anomaly coefficient a dimensionless proportional coefficient. The relative size and distribution of the abnormal data inside and outside the bucket of different regions can be understood more intuitively. By comparing the in and out of the bucket anomaly coefficients of different regions, it is easier to identify which regions have more significant anomalies.

[0137] Combining the numerator and denominator, the calculation of the regional bucket in-and-out anomaly coefficient is significant in that it provides a quantitative indicator to measure the relative importance or proportion of the in-bucket anomaly data in a specific region in the overall anomaly data. By comparing the in-bucket anomaly coefficients of different regions, it is possible to identify which regions have more significant in-bucket anomalies, and take corresponding measures to deal with or improve them.

[0138] w in and w out The existence of emphasizes the importance of abnormal data inside and outside the bucket in calculating the regional abnormal classification coefficient and the regional bucket abnormal coefficient. Even though it may be eventually eliminated, the original intention of introducing this weight factor is to show the degree of attention paid to the abnormal data in the bucket. In practical applications, w in and w out The value of may not be fixed, but can be adjusted according to the specific situation. For example, if you want to increase the impact of abnormal data in the bucket on the final result, you can increase w in and w out Although in the final formula w in and w out may be cancelled, but in the process of setting the weight factor, it has been adjusted by adjusting w in and w out If the formula needs to be expanded or modified in the future (for example, by introducing new abnormal data types or weighting factors), keep w in The presence in the formula can provide more flexibility and room for expansion.

[0139] The system compares the regional bucket-outside anomaly coefficient with the preset regional bucket-outside anomaly threshold to obtain the bucket-outside comparison result. If the bucket-outside anomaly coefficient exceeds the threshold, it is determined that there is an anomaly outside the bucket. The system compares the regional bucket-inside anomaly coefficient with the preset regional bucket-inside anomaly threshold to obtain the bucket-inside comparison result. If the bucket-inside anomaly coefficient exceeds the threshold, it is determined that there is an anomaly inside the bucket. Based on the bucket-outside comparison result, the system makes an anomaly determination outside the regional bucket and obtains the regional bucket-outside determination information. Based on the bucket-inside comparison result, the system makes an anomaly determination inside the regional bucket and obtains the regional bucket-inside determination information. This information also contains the specific details of the anomaly and possible handling measures. The regional bucket-outside determination information and the regional bucket-inside determination information are integrated to form the regional anomaly classification determination information.

[0140] The technical effect of the above technical solution is: by calculating the abnormal coefficients outside and inside the bucket respectively and comparing them with the corresponding thresholds, the system can more accurately identify the source and type of the abnormality, thereby refining the abnormal classification. The calculation of the abnormal coefficient and the comparison process with the threshold enable the system to more scientifically evaluate the severity of the abnormality and improve the accuracy of abnormal judgment. Based on the regional abnormal classification judgment information, the system can formulate more targeted treatment measures, effectively respond to different types of abnormalities, and improve management efficiency. By refining the abnormal classification and improving the judgment accuracy, the system can better adapt to the complex and changeable monitoring environment and enhance the robustness and stability of the system. The system's deep abnormal classification judgment function can promote the refined management of garbage classification work and improve the accuracy and efficiency of garbage classification. By timely discovering and handling abnormal behavior.

[0141] In one embodiment of the present invention, S3 includes:

[0142] All regional anomaly coefficients are ranked from small to large to obtain the ranking of classified anomaly regions;

[0143] Obtain a preset monitoring time sequence, and obtain a ranking change of a classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence;

[0144] Calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change;

[0145] The calculation formula of the timing evaluation coefficient is:

[0146]

[0147] Among them, XP is the timing evaluation coefficient, ΔB q is the upward ranking change of the region in the classification anomaly ranking, B q is the downward change of the region in the classification anomaly ranking; when B q=0, ΔB q The larger the value, the smaller the XP value, and the smaller the timing evaluation coefficient. B q -ΔB q The larger it is, the greater the XP.

[0148] Compare the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing determination abnormality comparison result;

[0149] According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

[0150] The working principle of the above technical solution is: rank all regional anomaly coefficients in ascending order to obtain the ranking of classified abnormal regions. This step can intuitively understand the degree of abnormality of each monitoring area and identify the area with the most serious abnormality. Obtain a preset monitoring time sequence, which is each time point of the preset time period to be monitored. According to the preset monitoring time sequence, the system obtains the change in the ranking of classified abnormal regions of each preset monitoring area within the continuous monitoring period. This change reflects the dynamic change of the degree of regional abnormality. The system calculates the time series evaluation abnormality coefficient of each preset monitoring area based on the ranking change. This coefficient comprehensively considers the absolute value of the regional abnormality coefficient and its change trend over time, and can more comprehensively evaluate the abnormal condition of the area. Compare the time series evaluation abnormality coefficient with the preset time series abnormality threshold to obtain the time series judgment abnormality comparison result. If the time series evaluation abnormality coefficient exceeds the threshold, it is determined that the monitoring area has abnormal changes in the time series. According to the time series judgment abnormality comparison result, the system marks the corresponding preset monitoring area as abnormal and triggers the early warning mechanism. This step can timely discover and handle regional abnormalities to prevent the problem from further deteriorating.

[0151] The technical effect of the above technical solution is: by calculating the ranking change and the time series evaluation anomaly coefficient, the system can realize dynamic monitoring and early warning of the abnormality of the monitoring area. This function can timely discover and deal with potential problems and improve monitoring efficiency. The time series evaluation anomaly coefficient comprehensively considers the absolute value of the regional anomaly coefficient and its changing trend over time, and can more comprehensively evaluate the abnormal conditions of the region. This evaluation method can more accurately identify abnormal areas and abnormal types. Through the abnormal marking and early warning mechanism, the system can guide managers to quickly locate and deal with abnormal areas and improve management efficiency. At the same time, the system's automated monitoring function reduces the workload of managers. The system's time series monitoring function can promote the refined management of garbage classification and improve the accuracy and efficiency of garbage classification. By timely discovering and handling abnormal behaviors, the system can enhance the public's awareness and participation in garbage classification. By comprehensively considering the dynamic changes of the regional anomaly coefficient, the system can better adapt to the complex and changing monitoring environment and enhance the robustness and stability of the system. This function can ensure that the system can operate accurately and reliably under various circumstances.

[0152] In one embodiment of the present invention, the system comprises:

[0153] The regional monitoring and collection module is used to obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the regional bucket and the data collected inside the regional bucket, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network;

[0154] The inside-outside-bucket anomaly analysis module is used to obtain the outside-outside-bucket anomaly data and the inside-outside-bucket anomaly data of the preset monitoring area, obtain the regional anomaly fusion data after data fusion, calculate the regional anomaly classification coefficient, make regional anomaly judgment, perform deep anomaly classification judgment on the obtained regional anomaly judgment information, and obtain regional anomaly classification judgment information;

[0155] The regional abnormal change analysis module is used to obtain the ranking of classified abnormal areas, obtain the ranking change of the preset monitoring area according to the preset monitoring time sequence, calculate the time sequence evaluation abnormal coefficient, and perform abnormal marking and early warning of the abnormal preset monitoring area.

[0156] The working principle of the above technical solution is as follows: obtain the preset monitoring areas, which include the garbage delivery points or the locations of the classified garbage bins. Perform real-time monitoring inside and outside the garbage bins in the preset monitoring area. The monitoring inside the bin mainly focuses on the classification of garbage, while the monitoring outside the bin focuses on the delivery behavior, the environment around the garbage bin, etc. The data collected inside and outside the bin are fused to form regional fusion collection information. This step can comprehensively analyze the classification and delivery of garbage in the monitoring area. Extract abnormal data from the fused data, including abnormal data inside the bin and abnormal data outside the bin. The abnormal data inside and outside the bin are fused again to form regional abnormal fusion data. Then, calculate the regional abnormal classification coefficient, which is used to evaluate the severity and type of abnormality. Use AI methods to deeply analyze the abnormal data, make more detailed classification judgments, and obtain regional abnormal classification judgment information. According to the abnormal classification judgment information, rank each monitoring area to reflect the degree of abnormality in each area. Combined with the preset monitoring time series (such as daily, weekly, etc.), obtain the ranking change of each area, and calculate the time series evaluation abnormality coefficient. This step can evaluate the persistence and development trend of abnormalities. Based on the time series evaluation anomaly coefficient, the abnormal preset monitoring area is marked and the early warning mechanism is triggered so that timely measures can be taken to intervene.

[0157] The technical effects of the above technical solutions are as follows: through real-time monitoring and AI analysis, the system can accurately identify errors inside and outside the garbage classification bin and improve the accuracy of garbage classification. The system can automatically identify and mark abnormal areas, reducing the workload of manual inspections and improving supervision efficiency. By improving the accuracy and efficiency of garbage classification, resource waste and environmental pollution in the garbage disposal process can be reduced. The system's early warning and feedback mechanism can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification. The large amount of data collected by the system can provide comprehensive analysis data for institutions or platforms to formulate garbage classification.

[0158] In one embodiment of the present invention, the regional monitoring and acquisition module includes:

[0159] A camera division module is used to obtain regional camera information and divide the regional camera information into external camera information and internal camera information according to the regional camera position;

[0160] The outside-barrel collection module is used to collect data from the outside-barrel area through the outside-barrel camera to obtain the outside-barrel collection data of the area;

[0161] The in-barrel collection module is used to collect data from the area in the barrel through the in-barrel camera to obtain the collected data in the regional barrel; the in-barrel camera is a camera that takes pictures of the garbage in the barrel, not a camera installed in the barrel;

[0162] The regional network acquisition module is used to fuse the data collected outside the regional barrel with the data collected inside the regional barrel to obtain regional fusion acquisition information; the regional fusion acquisition information includes overlapping the same parts and splicing the different parts to obtain regional fusion acquisition data, aiming to reduce or avoid monitoring blind spots. The regional camera includes a barrel camera and a barrel camera of a preset monitoring area, and the regional fusion acquisition information includes every combined frame image without blind spots inside and outside the barrel;

[0163] Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network;

[0164] Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

[0165] The working principle of the above technical solution is: obtain the location information of all cameras in the preset monitoring area. According to the location of the camera, the camera information is divided into the camera information outside the bin and the camera information inside the bin. It should be noted here that the camera inside the bin is not directly installed inside the trash can, but refers to the camera whose shooting range covers the internal area of ​​the trash can. The camera outside the bin collects data from the external area of ​​the trash can, including the placement behavior, personnel flow, etc., to obtain the regional external bin collection data. The camera inside the bin collects data from the internal area of ​​the trash can, including the classification of garbage, the overflow status of the trash can, etc., to obtain the regional internal bin collection data. The external bin collection data and the internal bin collection data are fused. The fusion process includes overlapping the same parts and splicing different parts to obtain regional fusion collection information without blind spots. This step is to reduce or avoid monitoring blind spots and ensure the comprehensiveness and accuracy of monitoring. The regional fusion collection information is presented in the form of each combined frame image, including a blind spot view inside and outside the bin. After obtaining the regional fusion collection information of all preset monitoring areas, the system builds a monitoring area network. The monitoring area network consists of multiple monitoring areas, each of which contains complete information collected from inside and outside the bin. Comprehensive monitoring of the monitoring area network is carried out to analyze the comprehensive monitoring data of each monitoring area in real time. The comprehensive monitoring data includes information on the overflow status of the garbage bin, classification accuracy, and placement behavior.

[0166] The technical effect of the above technical solution is: through the combined use of cameras inside and outside the barrel, the system can realize comprehensive monitoring of the source of garbage classification to ensure coverage without blind spots. And it can carry out targeted monitoring and processing of classification inside the barrel and packet loss outside the barrel, while considering comprehensive abnormal analysis of classification inside the barrel and packet loss outside the barrel; data fusion processing enables the system to obtain more comprehensive monitoring information, thereby improving the judgment accuracy of garbage classification accuracy. The system can monitor the situation of each monitoring area in real time, discover and deal with abnormal problems in time, and improve supervision efficiency. By building a monitoring area network, the system can realize intelligent management. The system can detect the overflowing state of the garbage bin in time to avoid waste of resources and environmental pollution caused by garbage overflow. The real-time monitoring and feedback mechanism of the system can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification.

[0167] In one embodiment of the present invention, the inside-outside-bucket abnormality analysis module includes:

[0168] A feature monitoring module, used for performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information;

[0169] The out-of-bucket anomaly extraction module is used to extract feature data of out-of-bucket packet loss data from the regional out-of-bucket collection data of the preset monitoring area to obtain the regional out-of-bucket anomaly data;

[0170] The in-bucket anomaly extraction module is used to extract feature data of unclassified data in the bucket from the data collected in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket;

[0171] An anomaly fusion module, used to fuse the anomaly data outside the regional bucket with the anomaly data inside the regional bucket to obtain regional anomaly fusion data;

[0172] A regional anomaly analysis module, used for calculating a regional anomaly classification coefficient based on the regional anomaly fusion data;

[0173] The calculation formula of the regional anomaly classification coefficient is:

[0174]

[0175] Among them, QY is the regional anomaly classification coefficient, N is the total number of regional monitoring (camera batches), and w out is the preset weight of the abnormal data outside the bucket, which is used to adjust the importance of the abnormal data outside the bucket in the calculation. in It is the preset weight of the abnormal data in the bucket, which is used to adjust the importance of the abnormal data in the bucket in the calculation. outi is the amount of abnormal data outside the bucket monitored for the i-th time (obtained from data collection), which can be quantified according to specific abnormality types, frequency, or duration, etc.ini is the amount of abnormal data in the bucket monitored for the i-th time (obtained from data collection), which can also be quantified based on specific abnormality types, frequency, duration, and other factors. i is the amount of data overlapping the data collected inside and outside the bucket during the i-th monitoring, which is used to measure the contribution of fusion information to reducing or avoiding monitoring blind spots. z The amount of data that overlaps the total monitored data collected inside and outside the bucket; 0< <1.

[0176] Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result;

[0177] Acquiring regional anomaly determination information according to the regional anomaly comparison result;

[0178] The deep classification analysis module is used to perform deep anomaly classification and determination on the regional anomaly determination information to obtain the regional anomaly classification and determination information.

[0179] The working principle of the above technical solution is as follows: based on the regional fusion collection information, the system performs feature monitoring on each preset monitoring area in the monitoring area network. Feature monitoring mainly focuses on the preset abnormal features inside and outside the bucket (packet loss data outside the bucket and unclassified data inside the bucket). For the regional outside-bucket collection data of the preset monitoring area, the system extracts feature data of the packet loss data outside the bucket. Packet loss data may include garbage directly discarded without classification, garbage delivery during non-delivery time, etc., which are all regarded as abnormal data outside the bucket. For the regional inside-bucket collection data of the preset monitoring area, the system extracts feature data of the unclassified data inside the bucket. Unclassified data may include various types of mixed garbage, garbage delivered in incorrect locations, etc., which are all regarded as abnormal data inside the bucket. The abnormal data outside the bucket and the abnormal data inside the bucket are fused to form regional abnormal fusion data. This step can understand the abnormal behavior or state in the monitoring area separately or comprehensively. Based on the regional abnormal fusion data, the system calculates the regional abnormal classification coefficient. This coefficient is used to evaluate the severity and type of the abnormality. The regional abnormal classification coefficient is compared with the preset regional abnormal threshold to obtain the regional abnormal comparison result. If the abnormal classification coefficient exceeds the threshold, it is determined that the monitoring area is abnormal. Based on the regional abnormality comparison results, the system obtains regional abnormality determination information, which includes whether the area is abnormal. The regional abnormality determination information is deeply analyzed and classified more carefully. This step can more accurately identify the specific type of abnormal data, etc.

[0180] The technical effect of the above technical solution is: through feature monitoring and abnormal data extraction, the system can more accurately identify abnormal behaviors or states at the source of garbage classification and improve the accuracy of monitoring. Data fusion and calculation of abnormal classification coefficients enable the system to comprehensively evaluate abnormal conditions in the monitoring area and provide comprehensive abnormal judgment information. Through abnormal comparison and judgment, the system can promptly discover and mark abnormal areas and trigger early warning mechanisms so that timely measures can be taken to intervene. Deep abnormal classification and judgment provide support for intelligent management, enabling the system to formulate targeted treatment measures according to the type of abnormality and improve management efficiency. The system's real-time monitoring and early warning mechanism can remind the public to pay attention to the correctness of garbage classification, thereby enhancing the public's awareness of garbage classification. By promptly discovering and handling abnormal behaviors, the system helps to reduce resource waste and environmental pollution.

[0181] In one embodiment of the present invention, the deep classification analysis module includes:

[0182] A deep anomaly calculation module is used to calculate an anomaly coefficient outside the regional bucket and an anomaly coefficient inside the regional bucket based on the anomaly data outside the regional bucket and the anomaly data inside the regional bucket combined with the regional anomaly classification coefficient;

[0183] The calculation formula of the abnormal coefficient outside the regional bucket is:

[0184]

[0185] Among them, TW is the regional out-of-bucket anomaly coefficient, QY is the regional anomaly classification coefficient, and w out is the preset weight of the abnormal data outside the bucket, which is used to adjust the importance of the abnormal data outside the bucket in the calculation. outd is the amount of abnormal data outside the bucket in the current region, M is the total number of all regions, and D outk is the amount of abnormal data outside the bucket of the kth region;

[0186] The calculation formula of the anomaly coefficient in the regional bucket is:

[0187]

[0188] Among them, TN is the regional bucket anomaly coefficient, QY is the regional anomaly classification coefficient, and w in It is the preset weight of the abnormal data in the bucket, which is used to adjust the importance of the abnormal data in the bucket in the calculation. ind is the amount of abnormal data in the bucket of the current region, M is the total number of all regions, and D ink is the amount of abnormal data in the bucket of the kth region;

[0189] The out-of-bucket anomaly analysis module is used to compare the out-of-bucket anomaly coefficient of the region with the preset out-of-bucket anomaly threshold of the region, obtain the out-of-bucket comparison result, make an anomaly judgment on the out-of-bucket region according to the out-of-bucket comparison result, and obtain the out-of-bucket judgment information of the region;

[0190] The in-bucket anomaly analysis module is used to compare the in-bucket anomaly coefficient of the regional bucket with the preset in-bucket anomaly threshold of the regional bucket to obtain an in-bucket comparison result, and perform an anomaly determination on the regional bucket according to the in-bucket comparison result to obtain in-bucket determination information of the regional bucket;

[0191] The classification anomaly determination module is used for the determination information outside the regional bucket and the determination information inside the regional bucket, that is, the regional anomaly classification determination information.

[0192] The working principle of the above technical solution is: according to the abnormal data outside the regional bucket and the abnormal data inside the regional bucket, combined with the regional abnormal classification coefficient calculated previously, the abnormal coefficient outside the regional bucket and the abnormal coefficient inside the regional bucket are calculated respectively. These two coefficients reflect the severity or importance of the abnormalities outside the bucket and inside the bucket respectively. The system compares the abnormal coefficient outside the regional bucket with the preset abnormal threshold outside the regional bucket to obtain the comparison result outside the bucket. If the abnormal coefficient outside the bucket exceeds the threshold, it is determined that there is an abnormality outside the bucket. The system compares the abnormal coefficient inside the regional bucket with the preset abnormal threshold inside the regional bucket to obtain the comparison result inside the bucket. If the abnormal coefficient inside the bucket exceeds the threshold, it is determined that there is an abnormality inside the bucket. According to the comparison result outside the bucket, the system makes an abnormality judgment outside the regional bucket and obtains the judgment information outside the regional bucket. According to the comparison result inside the bucket, the system makes an abnormality judgment inside the regional bucket and obtains the judgment information inside the regional bucket. This information also contains the specific details of the abnormality and possible treatment measures. The judgment information outside the regional bucket and the judgment information inside the regional bucket are integrated to form the regional abnormality classification judgment information.

[0193] The technical effect of the above technical solution is: by calculating the abnormal coefficients outside and inside the bucket respectively and comparing them with the corresponding thresholds, the system can more accurately identify the source and type of the abnormality, thereby refining the abnormal classification. The calculation of the abnormal coefficient and the comparison process with the threshold enable the system to more scientifically evaluate the severity of the abnormality and improve the accuracy of abnormal judgment. Based on the regional abnormal classification judgment information, the system can formulate more targeted treatment measures, effectively respond to different types of abnormalities, and improve management efficiency. By refining the abnormal classification and improving the judgment accuracy, the system can better adapt to the complex and changeable monitoring environment and enhance the robustness and stability of the system. The system's deep abnormal classification judgment function can promote the refined management of garbage classification work and improve the accuracy and efficiency of garbage classification. By timely discovering and handling abnormal behavior.

[0194] In one embodiment of the present invention, the regional abnormal change analysis module includes:

[0195] The anomaly ranking module is used to rank all regional anomaly coefficients in ascending order to obtain the ranking of classified anomaly regions;

[0196] A ranking change monitoring module is used to obtain a preset monitoring time sequence, and obtain the ranking change amount of the classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence;

[0197] A time series anomaly calculation module, used to calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change;

[0198] The calculation formula of the timing evaluation coefficient is:

[0199]

[0200] Among them, XP is the timing evaluation coefficient, ΔB q is the upward ranking change of the region in the classification anomaly ranking, B q is the downward change of the region in the classification anomaly ranking; when B q =0, ΔB q The larger the value, the smaller the XP value, and the smaller the timing evaluation coefficient. B q -ΔB q The larger it is, the greater the XP.

[0201] A judgment and early warning module, used for comparing the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing judgment abnormality comparison result;

[0202] According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

[0203] The working principle of the above technical solution is: rank all regional anomaly coefficients in ascending order to obtain the ranking of classified abnormal regions. This step can intuitively understand the degree of abnormality of each monitoring area and identify the area with the most serious abnormality. Obtain a preset monitoring time sequence, which is each time point of the preset time period to be monitored. According to the preset monitoring time sequence, the system obtains the change in the ranking of classified abnormal regions of each preset monitoring area within the continuous monitoring period. This change reflects the dynamic change of the degree of regional abnormality. The system calculates the time series evaluation abnormality coefficient of each preset monitoring area based on the ranking change. This coefficient comprehensively considers the absolute value of the regional abnormality coefficient and its change trend over time, and can more comprehensively evaluate the abnormal condition of the area. Compare the time series evaluation abnormality coefficient with the preset time series abnormality threshold to obtain the time series judgment abnormality comparison result. If the time series evaluation abnormality coefficient exceeds the threshold, it is determined that the monitoring area has abnormal changes in the time series. According to the time series judgment abnormality comparison result, the system marks the corresponding preset monitoring area as abnormal and triggers the early warning mechanism. This step can timely discover and handle regional abnormalities to prevent the problem from further deteriorating.

[0204] The technical effect of the above technical solution is: by calculating the ranking change and the time series evaluation anomaly coefficient, the system can realize dynamic monitoring and early warning of the abnormality of the monitoring area. This function can timely discover and deal with potential problems and improve monitoring efficiency. The time series evaluation anomaly coefficient comprehensively considers the absolute value of the regional anomaly coefficient and its changing trend over time, and can more comprehensively evaluate the abnormal conditions of the region. This evaluation method can more accurately identify abnormal areas and abnormal types. Through the abnormal marking and early warning mechanism, the system can guide managers to quickly locate and deal with abnormal areas and improve management efficiency. At the same time, the system's automated monitoring function reduces the workload of managers. The system's time series monitoring function can promote the refined management of garbage classification and improve the accuracy and efficiency of garbage classification. By timely discovering and handling abnormal behaviors, the system can enhance the public's awareness and participation in garbage classification. By comprehensively considering the dynamic changes of the regional anomaly coefficient, the system can better adapt to the complex and changing monitoring environment and enhance the robustness and stability of the system. This function can ensure that the system can operate accurately and reliably under various circumstances.

[0205] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A visual monitoring method for the source of household waste classification based on AI analysis, characterized in that: The method comprises: S1. Obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the area and the data collected inside the area, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network; S2. Obtain the abnormal data outside the regional bucket and the abnormal data inside the regional bucket of the preset monitoring area, perform data fusion to obtain regional abnormal fusion data, calculate the regional abnormal classification coefficient, perform regional abnormality determination, perform deep abnormality classification determination on the obtained regional abnormality determination information, and obtain regional abnormality classification determination information; S3. Obtain the ranking of classified abnormal areas, obtain the ranking change of the preset monitoring area according to the preset monitoring time series, calculate the time series evaluation abnormality coefficient, and perform abnormal marking and warning of the abnormal preset monitoring area.

2. According to claim 1, the visual monitoring method for the source of household waste classification based on AI analysis is characterized in that: The S1 includes: Acquire regional camera information, and divide the regional camera information into external camera information and internal camera information according to the regional camera position; The camera outside the barrel collects data from the area outside the barrel to obtain the collected data outside the barrel. The camera in the barrel collects data from the area in the barrel to obtain the collected data in the regional barrel; Fusing the data collected outside the regional bucket with the data collected inside the regional bucket to obtain regional fusion collection information; Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network; Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

3. According to claim 1, the visual monitoring method for the source of household waste classification based on AI analysis is characterized in that: The S2 includes: Performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information; Extract feature data of packet loss data outside the regional bucket from the data collected outside the regional bucket of the preset monitoring area to obtain abnormal data outside the regional bucket; Extract feature data of unclassified data in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket; Merging the abnormal data outside the regional bucket with the abnormal data inside the regional bucket to obtain regional abnormal fusion data; Calculating a regional anomaly classification coefficient based on the regional anomaly fusion data; Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result; Acquiring regional anomaly determination information according to the regional anomaly comparison result; Perform deep abnormality classification and determination on the regional abnormality determination information to obtain the regional abnormality classification and determination information.

4. According to claim 3, the visual monitoring method for the source of household waste classification based on AI analysis is characterized in that: The performing deep abnormality classification determination on the regional abnormality determination information to obtain the regional abnormality classification determination information includes: According to the abnormal data outside the regional bucket, the abnormal data inside the regional bucket and the regional abnormal classification coefficient, the abnormal coefficient outside the regional bucket and the abnormal coefficient inside the regional bucket are calculated; Compare the regional bucket out-of-bucket abnormality coefficient with the preset regional bucket out-of-bucket abnormality threshold, obtain the bucket out-of-bucket comparison result, perform an abnormality determination on the regional bucket out-of-bucket according to the bucket out-of-bucket comparison result, and obtain the regional bucket out-of-bucket determination information; Compare the regional bucket anomaly coefficient with a preset regional bucket anomaly threshold to obtain an intra-bucket comparison result, perform anomaly determination on the regional bucket according to the intra-bucket comparison result, and obtain intra-bucket determination information; The regional bucket outside determination information and the regional bucket inside determination information are the regional anomaly classification determination information.

5. According to claim 1, the visual monitoring method for the source of household waste classification based on AI analysis is characterized in that: The S3 includes: Rank all regional anomaly coefficients from small to large to obtain the ranking of classified anomaly regions; Obtain a preset monitoring time sequence, and obtain a ranking change amount of a classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence; Calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change; Compare the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing determination abnormality comparison result; According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

6. The visual monitoring system for the source of household waste classification based on AI analysis is characterized by: The system comprises: The regional monitoring and collection module is used to obtain the preset monitoring area, perform in-bucket monitoring and out-bucket monitoring on the preset monitoring area, obtain the data collected outside the regional bucket and the data collected inside the regional bucket, perform data fusion, obtain regional fusion collection information, and then build a monitoring area network; The inside-outside-bucket anomaly analysis module is used to obtain the outside-outside-bucket anomaly data and the inside-outside-bucket anomaly data of the preset monitoring area, obtain the regional anomaly fusion data after data fusion, calculate the regional anomaly classification coefficient, make regional anomaly judgment, perform deep anomaly classification judgment on the obtained regional anomaly judgment information, and obtain regional anomaly classification judgment information; The regional abnormal change analysis module is used to obtain the ranking of classified abnormal areas, obtain the ranking change of the preset monitoring area according to the preset monitoring time sequence, calculate the time sequence evaluation abnormal coefficient, and perform abnormal marking and early warning of the abnormal preset monitoring area.

7. According to claim 6, the visual monitoring system for the source of household waste classification based on AI analysis is characterized in that: The regional monitoring and acquisition module includes: A camera division module is used to obtain regional camera information and divide the regional camera information into external camera information and internal camera information according to the regional camera position; The outside-barrel collection module is used to collect data from the outside-barrel area through the outside-barrel camera to obtain the outside-barrel collection data of the area; The in-barrel collection module is used to collect data from the area in the barrel through the in-barrel camera to obtain the collected data in the regional barrel; A regional network acquisition module, used to fuse the data collected outside the regional bucket with the data collected inside the regional bucket to obtain regional fusion collection information; Obtain regional fusion collection information of all preset monitoring areas and build a monitoring area network; Comprehensive monitoring is performed on the monitoring area network to obtain comprehensive monitoring data.

8. The visual monitoring system for the source of household waste classification based on AI analysis according to claim 6 is characterized in that: The inside and outside bucket abnormality analysis module includes: A feature monitoring module, used for performing feature monitoring on each preset monitoring area of ​​the monitoring area network according to the regional fusion collection information; The out-of-bucket anomaly extraction module is used to extract feature data of out-of-bucket packet loss data from the regional out-of-bucket collection data of the preset monitoring area to obtain the regional out-of-bucket anomaly data; The in-bucket anomaly extraction module is used to extract feature data of unclassified data in the bucket from the data collected in the regional bucket of the preset monitoring area to obtain abnormal data in the regional bucket; An anomaly fusion module, used to fuse the anomaly data outside the regional bucket with the anomaly data inside the regional bucket to obtain regional anomaly fusion data; A regional anomaly analysis module, used for calculating a regional anomaly classification coefficient based on the regional anomaly fusion data; Comparing the regional anomaly classification coefficient with a preset regional anomaly threshold to obtain a regional anomaly comparison result; Acquiring regional anomaly determination information according to the regional anomaly comparison result; The deep classification analysis module is used to perform deep anomaly classification and determination on the regional anomaly determination information to obtain the regional anomaly classification and determination information.

9. The visual monitoring system for the source of household waste classification based on AI analysis according to claim 8 is characterized in that: The deep classification analysis module includes: A deep anomaly calculation module is used to calculate an anomaly coefficient outside the regional bucket and an anomaly coefficient inside the regional bucket based on the anomaly data outside the regional bucket and the anomaly data inside the regional bucket combined with the regional anomaly classification coefficient; The out-of-bucket anomaly analysis module is used to compare the out-of-bucket anomaly coefficient of the region with the preset out-of-bucket anomaly threshold of the region, obtain the out-of-bucket comparison result, make an anomaly judgment on the out-of-bucket region according to the out-of-bucket comparison result, and obtain the out-of-bucket judgment information of the region; The in-bucket anomaly analysis module is used to compare the in-bucket anomaly coefficient of the regional bucket with the preset in-bucket anomaly threshold of the regional bucket to obtain an in-bucket comparison result, and perform an anomaly determination on the regional bucket according to the in-bucket comparison result to obtain in-bucket determination information of the regional bucket; The classification anomaly determination module is used for the determination information outside the regional bucket and the determination information inside the regional bucket, that is, the regional anomaly classification determination information.

10. According to claim 6, the visual monitoring system for the source of household waste classification based on AI analysis is characterized in that: The regional abnormal change analysis module includes: The anomaly ranking module is used to rank all regional anomaly coefficients in ascending order to obtain the ranking of classified anomaly regions; A ranking change monitoring module is used to obtain a preset monitoring time sequence, and obtain the ranking change amount of the classified abnormal area ranking of each preset monitoring area according to the preset monitoring time sequence; A time series anomaly calculation module, used to calculate the time series evaluation anomaly coefficient of each preset monitoring area according to the ranking change; A judgment and early warning module, used for comparing the timing evaluation abnormality coefficient with a preset timing abnormality threshold to obtain a timing judgment abnormality comparison result; According to the abnormal comparison result of the time sequence determination, abnormal marking and early warning are performed for the corresponding preset monitoring area.

Images