A renewable resource sorting and management system based on image analysis

Abstract

The present invention discloses a renewable resource sorting and management system based on image analysis, comprising: an image acquisition module, an image analysis module, a scheduling and control module, a real-time monitoring module, and a secondary analysis module; the image acquisition module comprises a multispectral imaging unit and a 3D point cloud acquisition unit; the image analysis module performs the following operations: (a) obtaining material types by matching multispectral data with a material fingerprint library; (b) calculating the spatial density distribution of the 3D point cloud; and (c) fusing the material type and spatial density distribution to generate renewable resource partition information; the scheduling and control module schedules sorting equipment based on the partition information; the real-time monitoring module collects equipment status and sorting information; and the secondary analysis module generates equipment management information and sorting evaluation information. The present invention achieves efficient, intelligent, and sustainable renewable resource sorting through economic-oriented strategies, intelligent equipment maintenance, and data-driven optimization.

Description

Technical Field

[0001] The present invention relates to the field of renewable resource management systems, and in particular to a renewable resource sorting and management system based on image analysis. Background Art

[0002] As a key link in resource recycling, the core goal of renewable resource sorting is to improve the purity and value of recyclables through efficient classification and reduce subsequent processing costs. The traditional sorting model that relies on manual visual inspection and experience-based judgment has significant limitations, such as a high rate of material misjudgment, lack of spatial distribution analysis, and delayed response to equipment failures. These lead to low resource recovery rates and insufficient economic benefits, making it difficult to adapt to the large-scale and refined needs of the modern circular economy. Therefore, the industry urgently needs to introduce automated and intelligent technologies to achieve precision, efficiency, and standardization of the sorting process. Therefore, a renewable resource sorting management system based on image analysis is proposed. Summary of the Invention

[0003] The technical problem to be solved by the present invention is: how to solve the problems of high material misjudgment rate, lack of spatial distribution analysis, and delayed response to equipment failure in the existing management system, and provide a renewable resource sorting management system based on image analysis.

[0004] The present invention solves the above-mentioned technical problems through the following technical solutions, which include:

[0005] Image acquisition module, image analysis module, scheduling control module, real-time monitoring module, and re-analysis module;

[0006] The image acquisition module includes a multispectral imaging unit and a 3D point cloud acquisition unit;

[0007] The image analysis module performs:

[0008] (a): Obtain material type by matching multispectral data through the material fingerprint library;

[0009] (b): Calculate the spatial density distribution of 3D point cloud;

[0010] (c) Fusion of material type and spatial density distribution to generate renewable resource partition information. The specific process is as follows:

[0011] The fusion of material type and spatial density distribution is achieved through the fusion evaluation function, and the renewable resource zoning information is generated by optimizing the weight coefficient of the fusion evaluation function;

[0012] The real-time monitoring module collects equipment status and sorting information;

[0013] The re-analysis module generates equipment management information and sorting evaluation information.

[0014] Furthermore, when the image analysis module performs step (c):

[0015] Constructing fusion evaluation function , the specific process is as follows:

[0016] ,

[0017] in is the material recognition confidence, is the density zoning rationality index, and are the weight coefficients of the corresponding items, which are used to adjust the importance of material identification confidence and density zoning rationality index respectively;

[0018] choose The partition scheme corresponding to the maximum value is the final renewable resource partition information.

[0019] Furthermore, the dispatch control module includes an economic decision-making submodule, which is used to:

[0020] Access the price database to obtain material unit prices ;

[0021] Then calculate the value density of each region , where U is the area of ​​the region, is the mass of the i-th type of material in the corresponding area;

[0022] Finally, according to the value density of each region Generate sorting priorities in descending order;

[0023] The scheduling control module is also used to adjust the conveyor belt speed and receive the instructions of the re-analysis module by adjusting the weight coefficient of the fusion evaluation function of the image analysis module. and , optimize the renewable resource zoning plan.

[0024] Furthermore, the real-time monitoring module is also used to collect the air pressure of the pneumatic sorting device. , Robotic arm motor current , color sorter response time ;

[0025] Then calculate the air pressure health index , Current Health Index and Response Time Index :

[0026] Air Pressure Health Index : ;

[0027] Current Health Index : ;

[0028] Response Time Index : ;

[0029] in is standard atmospheric pressure, is the standard current, is the standard response time;

[0030] Computing Equipment Health Index The specific process is as follows:

[0031] ;

[0032] when When the value falls below the threshold, an early warning signal is generated and the abnormal parameter type is simultaneously marked, such as abnormal air pressure, abnormal current or abnormal response time.

[0033] Furthermore, the real-time monitoring module is further configured to:

[0034] Obtain sorting quality through weighing units and component detectors Mixing rate with impurities ;

[0035] Counting missed detections using industrial cameras and wrong fractions ;

[0036] Generate a sorting information matrix based on the above data in time series , and transmitted to the re-analysis module for sorting evaluation, where T refers to the transpose of the matrix.

[0037] Furthermore, the process of the re-analysis module generating device management information includes:

[0038] Based on sorting information matrix Time series data to calculate the comprehensive sorting efficiency , where the matrix The row vectors of correspond to different batches of sorting data, specifically:

[0039] ,λ is the penalty coefficient;

[0040] when When the missed detection rate is greater than the preset value, the device management information is sent to the scheduling control module to adjust the weight coefficient of the fusion evaluation function of the image analysis module. and ;

[0041] If the misclassification rate is greater than the preset value, adjust the conveyor belt speed.

[0042] Furthermore, the process of generating device management information by the re-analysis module further includes:

[0043] Comprehensive performance of computing equipment ;

[0044] in, 、 and is the weight coefficient, is the actual running time, is the total time, is the standard energy consumption, is the actual energy consumption;

[0045] according to Value generation equipment management information triggers task allocation, maintenance or shutdown instructions in a hierarchical manner;

[0046] Calculating the risk of nozzle clogging for pneumatic sorting devices , specifically , where k is a constant, is the change in air pressure, is the air pressure threshold;

[0047] when When the value is greater than the preset value, additional nozzle detection requirements are added;

[0048] The process of generating device management information by the reanalysis module also includes:

[0049] Fault location is performed based on the independent parameter health index. The specific process is as follows: Air pressure health index When it is less than the preset value, such as the air pressure health index When it is less than 0.8, it is determined that the pneumatic system is abnormal, and the air pressure pipeline leakage detection of the pneumatic sorting device is triggered;

[0050] When the current health index When the current health index is less than the preset value, If the value is less than 0.7, it is determined that the robot arm load is abnormal, and the motor current waveform is retrieved to analyze the cause of the jam or overload.

[0051] When the response time index When it is greater than the preset value, such as the response time index If it is greater than 0.9, the response time is too long, which is judged to be a delay in the image processing of the color sorter. The algorithm of the automatic optimization image analysis module is time-consuming.

[0052] Furthermore, the specific process of the re-analysis module generating the sorting evaluation information includes:

[0053] Get sorting information matrix from real-time monitoring module , from the sorting information matrix Extract the qualified quality of sorting , impurity mixing rate , the number of missed detections and wrong fractions ;

[0054] Get the material unit price from the scheduling control module ;

[0055] Calculate the data obtained from the real-time monitoring module and the scheduling control module to obtain the batch value recovery rate , the specific process is: ,in is the mass of the jth type of input material before sorting, Refers to the unit price of the jth type of material in the input materials;

[0056] Then, the multispectral features and 3D contour features of the missed materials in the image acquisition module are extracted; the missed features are grouped using the K-means clustering algorithm, that is, the multispectral features and 3D contour features of the missed materials in the image acquisition module are grouped, and the frequency of occurrence of each type of material is counted;

[0057] Mark the top three material categories with the highest frequency of occurrence as high-frequency loss materials;

[0058] For high-frequency loss materials, in the image analysis module, the material spectral feature matching weight is increased by a preset value, and the 3D point cloud segmentation sensitivity threshold of such materials is lowered by a preset value;

[0059] Then diagnose the cause of the missorting, analyze the sorting logs of the missorted materials, and classify and count them. The specific statistical process is as follows:

[0060] Material misjudgment: The image feature matching similarity is between the threshold range , and is the material matching similarity threshold;

[0061] Position offset: The Euclidean distance between the actual sorting coordinate and the target coordinate is greater than the preset value;

[0062] Equipment malfunction: The pressure sensor reading of the robotic arm is lower than the standard value;

[0063] Then the batch value recovery rate , integrate and classify the information of high-frequency loss materials and misclassification reasons;

[0064] The classification assessment information includes a visual report of the following:

[0065] Value recovery rate and its historical trend curve;

[0066] Radar map of characteristic distribution of high-frequency loss materials;

[0067] Pie chart of the proportion of reasons for misclassification.

[0068] Furthermore, the process of diagnosing the cause of the misclassification further includes:

[0069] When the material misjudgment ratio exceeds 60%, the material matching threshold is dynamically adjusted. , the specific process is:

[0070] ;

[0071] in is the preset adjustment step size, and the misjudgment rate is the ratio of the number of material misjudgments to the total error score. The original material matching similarity threshold before adjustment, Match the similarity threshold for the adjusted new material;

[0072] When the position offset exceeds 30%, a coordinate calibration instruction is sent to the scheduling control module;

[0073] When the proportion of device malfunctions exceeds 20%, the device will be marked as requiring mandatory maintenance in the device management information.

[0074] Furthermore, the invention further comprises a dynamic calibration unit, which is used to:

[0075] Regularly calibrate the spectrometer and 3D camera of the image acquisition module;

[0076] Reset the standard parameters of the sorting equipment, that is, adjust is standard atmospheric pressure, is the standard current, is the standard response time;

[0077] Update the matching threshold of the material fingerprint library.

[0078] Compared with the existing technology, the present invention has the following advantages: the renewable resource sorting and management system based on image analysis can accurately identify the type of renewable resource materials and analyze the spatial density distribution, thereby generating reasonable zoning information, improving sorting accuracy and resource classification rationality, determining sorting priority according to material value density, and enabling priority processing of high-value materials, thereby improving the economic benefits of sorting operations, and real-time monitoring of equipment status parameters and calculation of health indexes, thereby providing timely warnings of equipment anomalies, ensuring stable equipment operation, and reducing downtime due to faults; comprehensive analysis and evaluation based on sorting data can dynamically optimize and integrate evaluation functions, equipment parameters, and sorting strategies, continuously improving sorting efficiency and quality, regularly calibrating and updating equipment parameters to ensure that the system maintains high-precision operation for a long time, and presenting sorting evaluation information through visual reports, making it easy for managers to intuitively grasp the system operation status and achieve refined management and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0080] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0081] like Figure 1 As shown, this embodiment provides a technical solution: a renewable resource sorting and management system based on image analysis, comprising:

[0082] Image acquisition module, image analysis module, scheduling control module, real-time monitoring module, and re-analysis module; the image acquisition module includes a multispectral imaging unit and a 3D point cloud acquisition unit;

[0083] The image analysis module performs:

[0084] (a): Obtain material type by matching multispectral data through the material fingerprint library;

[0085] (b): Calculate the spatial density distribution of 3D point cloud;

[0086] (c) Fusion of material type and spatial density distribution to generate renewable resource partition information. The specific process is as follows:

[0087] The fusion of material type and spatial density distribution is achieved through the fusion evaluation function, and the renewable resource zoning information is generated by optimizing the weight coefficient of the fusion evaluation function;

[0088] The scheduling control module schedules the sorting equipment according to the partition information;

[0089] The real-time monitoring module collects equipment status and sorting information;

[0090] The re-analysis module generates equipment management information and sorting evaluation information.

[0091] When the image analysis module performs step (c):

[0092] Constructing fusion evaluation function , the specific process is as follows:

[0093] ,

[0094] in is the material recognition confidence, is the density zoning rationality index, and are the weight coefficients of the corresponding items, which are used to adjust the importance of material identification confidence and density zoning rationality index respectively;

[0095] choose The partition scheme corresponding to the maximum value is the final renewable resource partition information.

[0096] By constructing a fusion evaluation function and weightedly integrating the material identification confidence and density zoning rationality index, the material credibility and spatial distribution rationality of recycled resources can be comprehensively considered, avoiding the one-sidedness of single-dimensional evaluation, thereby generating more accurate and reliable recycled resource zoning information, and improving the sorting system's adaptability to complex material scenarios and the scientific nature of decision-making;

[0097] For example, suppose that in a certain renewable resource pile, the multispectral imaging unit identifies that the material of area A is metal with a high confidence level. =0.9, but the 3D point cloud shows that its spatial density distribution is scattered and the rationality index is low. =0.5;

[0098] The material recognition confidence of area B is medium. =0.7, but the density distribution is compact and uniform =0.8;

[0099] If sorting is based solely on material identification, area A may be prioritized, but the actual density dispersion may contain mixed materials;

[0100] If sorted only by density distribution, region B may be better, but the reliability of material identification is insufficient.

[0101] By fusion evaluation function Calculate, assume = =0.5;

[0102] S of area A = 0.5 × 0.9 + 0.5 × 0.5 = 0.7;

[0103] S of area B = 0.5 × 0.7 + 0.5 × 0.8 = 0.75;

[0104] Finally, area B with a larger S is selected as the priority partition, which balances the material credibility and distribution rationality and avoids misjudgment caused by a single indicator.

[0105] The dispatch control module includes an economic decision-making submodule, which is used to:

[0106] Access the price database to obtain material unit prices ;

[0107] Then calculate the value density of each region , where U is the area of ​​the region, is the mass of the i-th type of material in the corresponding area;

[0108] Finally, according to the value density of each region Generate sorting priorities in descending order;

[0109] The scheduling control module is also used to adjust the conveyor belt speed and receive the instructions of the re-analysis module by adjusting the weight coefficient of the fusion evaluation function of the image analysis module. and , optimize the renewable resource zoning plan;

[0110] By accessing the price database through the economic decision-making submodule, the value density is calculated by combining the unit price of the material and the regional quality. The sorting priority is generated in descending order of value density. This allows the sorting system to prioritize high-value recycled resources, optimize the economic benefits of resource sorting, prevent low-value materials from occupying sorting resources, and improve the input-output ratio and resource recovery efficiency of the overall sorting operation. Assume that there are two material areas in the recycled resource sorting scenario:

[0111] Area 1: 1 square meter, including 50 kg of scrap iron, priced at 1 yuan / kg; 30 kg of waste plastic, priced at 0.5 yuan / kg;

[0112] Value Density =(50×1+30×0.5) / 1=65 yuan / square meter;

[0113] Area 2: 1.5 square meters, containing 20 kg of scrap copper at a price of 10 yuan / kg; 10 kg of scrap aluminum at a price of 5 yuan / kg;

[0114] Value Density =(20×10+10×5) / 1.5=166.67 yuan / square meter.

[0115] According to the descending order of value density, the system prioritizes the dispatching of sorting equipment to process area 2, which is the sorting of high-value scrap copper and scrap aluminum, to ensure that high-yield materials are recycled first and improve the economy of the sorting operation.

[0116] The real-time monitoring module is also used for:

[0117] Collect air pressure of pneumatic sorting device , Robotic arm motor current , color sorter response time ;

[0118] Then calculate the air pressure health index , Current Health Index and Response Time Index :

[0119] Air Pressure Health Index : ;

[0120] Current Health Index : ;

[0121] Response Time Index : ;

[0122] in is standard atmospheric pressure, is the standard current, is the standard response time;

[0123] Computing Equipment Health Index The specific process is as follows:

[0124] ;

[0125] when When the value falls below the threshold, an early warning signal is generated and the abnormal parameter type is simultaneously marked, such as abnormal air pressure, abnormal current or abnormal response time;

[0126] By collecting key equipment operating parameters in real time, such as air pressure, current, and response time, calculating standardized health indexes and providing dynamic warnings, we can achieve:

[0127] Accurate monitoring of equipment status: Convert complex equipment parameters into intuitive health indexes to reflect equipment operation quality in real time;

[0128] Early warning of faults: triggering warnings before equipment performance deteriorates or fails, thus avoiding sudden downtime that affects sorting efficiency;

[0129] Efficient anomaly location: Synchronously mark the specific abnormal parameter type, shorten troubleshooting time, and reduce maintenance costs;

[0130] For example, in a sorting line:

[0131] Standard atmospheric pressure =0.6MPa, the measured air pressure at a certain moment =0.5MPa,

[0132] Air Pressure Health Index =1-|0.5-0.6| / 0.6≈0.83, the threshold is assumed to be 0.8, and no warning is triggered at this time;

[0133] The subsequent measured air pressure continued to drop to =0.48MPa, HP=1-|0.48-0.6| / 0.6=0.8, when the threshold is reached, the system generates an air pressure abnormality warning and automatically triggers the air pressure pipeline leakage detection of the pneumatic sorting device, prompting maintenance personnel to check the leakage point in time to avoid sorting action failure due to insufficient air pressure.

[0134] The real-time monitoring module is also used for:

[0135] Obtain sorting quality through weighing units and component detectors Mixing rate with impurities ;

[0136] Counting missed detections using industrial cameras and wrong fractions ;

[0137] Generate a sorting information matrix based on the above data in time series , and transmitted to the re-analysis module for sorting evaluation, T refers to the transpose of the matrix;

[0138] By collecting multi-dimensional data such as sorting quality, impurity mixing rate, missed detection number, error score, etc. in real time, generating a sorting information matrix in time series and transmitting it to the analysis module, it can realize data-based monitoring of the entire process, convert the key performance indicators of the sorting operation into structured data, and reflect the sorting quality and efficiency in real time;

[0139] This module provides raw data support for subsequent comprehensive sorting efficiency calculations, equipment performance analysis, and sorting strategy optimization. It also traces the time points and characteristics of sorting anomalies through historical data sequences, helping to locate the root causes of problems such as missed inspections and missorting. Assume that within a certain period of time, the real-time monitoring module collects three batches of sorting data as follows:

[0140] Batch 1: sorting qualified mass W1=1000kg, impurity mixing rate Z1=3%, missed inspection number =5, wrong fraction =3;

[0141] Batch 2: W2=1200kg, Z2=5%, =10, =8;

[0142] Batch 3: W3=900kg, Z3=2%, =2, =1;

[0143] The generated sorting information matrix is: ;

[0144] By analyzing the matrix, it was found that the impurity mixing rate and missed detection / error score of batch 2 were significantly higher. The re-analysis module can retrieve the image data and equipment operation log of this batch in a targeted manner to trace whether the sorting quality has declined due to misjudgment of the image analysis algorithm or drift of equipment parameters, and then optimize the fusion evaluation function or calibrate the equipment parameters.

[0145] The process of generating device management information by the re-analysis module includes:

[0146] Based on sorting information matrix Time series data to calculate the comprehensive sorting efficiency , where the matrix The row vectors of correspond to different batches of sorting data, specifically:

[0147] ,λ is the penalty coefficient;

[0148] when When the missed detection rate is greater than the preset value, the device management information is sent to the scheduling control module to adjust the weight coefficient of the fusion evaluation function of the image analysis module. and ;

[0149] If the error rate is greater than the preset value, adjust the conveyor belt speed, that is, reduce the conveyor belt speed to prompt the recognition time to reduce the error rate;

[0150] The missed detection rate is the ratio of missed detections to the total sorting volume, and the wrong sorting rate is the ratio of wrong detections to the total sorting volume;

[0151] By calculating the comprehensive sorting efficiency By comparing it with the threshold, it can dynamically evaluate the actual efficiency of the sorting operation, accurately locate batches or links with reduced efficiency based on time series data, automatically trigger fusion evaluation function optimization or conveyor belt speed adjustment, and realize intelligent dynamic calibration of the sorting strategy, avoiding efficiency bottlenecks caused by fixed parameters, and improving the overall efficiency and stability of recycled resource sorting.

[0152] If the comprehensive sorting efficiency within a certain period of time Continuously below the preset threshold, such as comprehensive sorting efficiency If the value is continuously lower than 90%, the system determines that the current fusion evaluation function or conveyor speed is unreasonable and automatically adjusts the partition fusion weight of the image analysis module, such as increasing the weight of the material recognition confidence. , or reduce the conveyor belt speed to extend the material identification time, so that the sorting efficiency of subsequent batches returns to a reasonable range.

[0153] The process of generating device management information by the re-analysis module further includes:

[0154] Comprehensive performance of computing equipment ;

[0155] in, 、 and is the weight coefficient, is the actual running time, is the total time, is the standard energy consumption, is the actual energy consumption;

[0156] according to Value generation equipment management information triggers task allocation, maintenance or shutdown instructions in a hierarchical manner;

[0157] Calculating the risk of nozzle clogging for pneumatic sorting devices , specifically , where k is a constant, is the change in air pressure, is the air pressure threshold;

[0158] when When the value is greater than the preset value, additional nozzle detection requirements are added;

[0159] The process of generating device management information by the reanalysis module also includes:

[0160] Fault location is performed based on the independent parameter health index. The specific process is as follows: Air pressure health index When it is less than the preset value, such as the air pressure health index When it is less than 0.8, it is determined that the pneumatic system is abnormal, and the air pressure pipeline leakage detection of the pneumatic sorting device is triggered;

[0161] When the current health index When the current health index is less than the preset value, When it is less than the preset value of 0.7, it is determined that the robot arm load is abnormal, and the motor current waveform is retrieved to analyze the cause of the jam or overload;

[0162] When the response time index When it is greater than the preset value, such as the response time index If it is greater than 0.9, the response time is too long, which is judged to be a delay in the image processing of the color sorter. The algorithm of the automatic optimization image analysis module takes time.

[0163] By integrating the equipment health index, actual operating time percentage, and energy efficiency, a comprehensive quantitative analysis of equipment performance can be achieved, avoiding deviations from a single indicator and assisting in formulating hierarchical maintenance strategies, such as task allocation, maintenance, or shutdown.

[0164] The nozzle blockage risk model provides early warning of potential failures in pneumatic equipment, reducing unexpected downtime; based on independent parameter health indexes such as air pressure, current, and response time, the fault type is accurately located, shortening the troubleshooting time.

[0165] Automatically trigger algorithm optimization based on parameter anomalies, such as optimizing the image analysis algorithm when image processing of the color sorter is delayed, to improve system adaptability;

[0166] Such as comprehensive equipment performance evaluation:

[0167] Assume that the parameters of a sorting robot arm are as follows:

[0168] The device health index H=0.7, the threshold is 0.6, which is normal;

[0169] Actual running time ratio =0.8, higher than the standard 0.7;

[0170] Energy efficiency =0.9, lower than the standard of 1.0, and the energy consumption is high;

[0171] Weight coefficient =0.5, =0.3, =0.2;

[0172] Calculate the overall efficiency: K = 0.5 × 0.7 + 0.3 × 0.8 + 0.2 × 0.9 = 0.77;

[0173] Because K is higher than the threshold of 0.7, the equipment is operating normally. If K continues to be lower than the threshold, the system will trigger an abnormal energy consumption warning and prompt you to check the motor load or conveyor belt resistance.

[0174] Warning of nozzle blockage risk:

[0175] Standard air pressure for pneumatic sorting devices =0.6MPa, the pressure change at a certain moment =0.15MPa, =0.1MPa, constant k=5, calculate the blockage risk:

[0176] ;

[0177] When the preset congestion risk threshold is 0.5, >0.5, the system adds a nozzle detection task to clear potential blockages in advance to avoid sorting failure;

[0178] Such as fault precise location:

[0179] The air pressure health index = 0.75, and the preset threshold is 0.8, which means that the pneumatic system is abnormal and the air pressure line leak detection is triggered;

[0180] Current Health Index =0.6, the preset threshold is 0.7, which means that the load of the robot arm is abnormal, and the current waveform is retrieved to analyze whether the overload is caused by obstruction;

[0181] Response Time Index =0.95, the preset threshold is 0.9, and the response time is long, which means that the color sorter image processing is delayed, and the automatic optimization of the image analysis algorithm is time-consuming.

[0182] The specific process of the re-analysis module generating the sorting evaluation information includes:

[0183] Get sorting information matrix from real-time monitoring module , from the sorting information matrix Extract the qualified quality of sorting , impurity mixing rate , the number of missed detections and wrong fractions ;

[0184] Get the material unit price from the scheduling control module ;

[0185] Calculate the data obtained from the real-time monitoring module and the scheduling control module to obtain the batch value recovery rate , the specific process is: ,in is the mass of the jth type of input material before sorting, Refers to the unit price of the jth type of material in the input materials;

[0186] Then extract the multi-spectral features and 3D contour features of the missed materials in the image acquisition module;

[0187] The K-means clustering algorithm is used to group the missed detection features. That is, the multispectral features and 3D contour features of the missed detection materials in the image acquisition module are grouped, and the frequency of occurrence of each type of material is counted.

[0188] Mark the top three material categories with the highest frequency of occurrence as high-frequency loss materials;

[0189] For high-frequency loss materials, in the image analysis module, the material spectral feature matching weight is increased by a preset value, such as 20%, and the 3D point cloud segmentation sensitivity threshold of such materials is reduced by a preset value, such as 10%;

[0190] Then diagnose the cause of the missorting, analyze the sorting logs of the missorted materials, and classify and count them. The specific statistical process is as follows:

[0191] Material misjudgment: The image feature matching similarity is between the threshold range , and is the material matching similarity threshold;

[0192] Position offset: The Euclidean distance between the actual sorting coordinate and the target coordinate is greater than the preset value;

[0193] Equipment malfunction: The pressure sensor reading of the robotic arm is lower than the standard value;

[0194] Then the batch value recovery rate , integrate and classify the information of high-frequency loss materials and misclassification reasons;

[0195] The classification assessment information includes a visual report of the following:

[0196] Value recovery rate and its historical trend curve;

[0197] Radar map of characteristic distribution of high-frequency loss materials;

[0198] The batch value recovery rate is used to comprehensively measure the resource value retention capacity of the sorting operation. Combined with missed detection / misclassification data, a multi-dimensional quantitative assessment of sorting quality is achieved to avoid the one-sidedness of relying on a single indicator. The K-means clustering algorithm is used to analyze the image characteristics of missed detection materials, identify the categories of frequently lost materials, and optimize the matching weight and segmentation sensitivity of the image analysis module in a targeted manner to reduce the repeated missed detection rate of similar materials. Through sorting log analysis, the causes of misclassification are classified as material misjudgment, position offset, or equipment malfunction. Combined with the proportion statistics, dynamic adjustments are triggered, such as threshold correction, coordinate calibration, or equipment maintenance, to improve the adaptability and stability of the sorting system. Through visual reports such as the value recovery rate trend curve, the radar chart of the characteristics of frequently lost materials, and the pie chart of the reasons for misclassification, managers are provided with an intuitive sorting efficiency analysis tool to assist in the rapid formulation of optimization strategies.

[0199] For example, when a recycling resource sorting line processes mixed materials, such as waste plastic bottles, aluminum cans, and cartons, it is found that the batch value recovery rate C is lower than the industry average for three consecutive times, and the number of missed inspections and error scores are high. At this time, the value recovery rate calculation is performed:

[0200] Input materials before sorting: 1000kg of waste plastic bottles, unit price 0.8 yuan / kg; 500kg of aluminum cans, unit price 2 yuan / kg; 800kg of cartons, unit price 0.5 yuan / kg,

[0201] Total input value = 1000 × 0.8 + 500 × 2 + 800 × 0.5 = 2200 yuan;

[0202] Qualified quality after sorting: 900kg of waste plastic bottles, impurity mixing rate 5%; 450kg of aluminum cans, impurity mixing rate 3%; 750kg of cartons, impurity mixing rate 2%; qualified value = 900×0.8×(1-5%)+450×2×(1-3%)+750×0.5×(1-2%)=1812 yuan;

[0203] The batch value recovery rate C=1812 / 2200≈82.4%<85%, which means the sorting efficiency does not meet the standard.

[0204] Identification of high-frequency lost materials:

[0205] After image feature extraction of missed items, K-means clustering revealed that transparent PET plastic bottles and irregular-shaped aluminum cans accounted for 70% of the total number of missed items and were marked as high-frequency lost items.

[0206] The system automatically increases the material spectral matching weight for PET plastic bottles by 25% and reduces the 3D point cloud segmentation sensitivity threshold for irregular-shaped aluminum cans by 12%, enhancing the recognition accuracy of both types of materials.

[0207] Diagnosis of the cause of misclassification:

[0208] Analysis of misclassification logs revealed that 60% of misclassifications were due to material misjudgment. For example, the spectral similarity between transparent PET plastic bottles and glass bottles fell within the threshold range of [0.6, 0.8]. 30% of misclassifications were due to positional offset, with sorting coordinate deviations exceeding 5cm.

[0209] The system dynamically adjusts the material matching threshold: = - × (60% / 100%), assuming =0.1, the threshold is adjusted from 0.7 to 0.64, and a coordinate calibration instruction is sent to the scheduling module to reduce misclassification caused by position offset.

[0210] Generates a historical trend curve for value recovery rate, showing the decline in C-value fluctuations in the past week, a radar chart of high-frequency material loss characteristics, highlights the spectral / contour characteristics of transparent PET plastic bottles and special-shaped aluminum cans, and a pie chart of the reasons for misclassification, with material misjudgment accounting for 60%. This helps managers quickly locate problems and initiate optimization strategies.

[0211] The process of diagnosing the cause of the misclassification further includes:

[0212] When the material misjudgment ratio exceeds 60%, the material matching threshold is dynamically adjusted. , the specific process is:

[0213] ;

[0214] in is the preset adjustment step size, and the misjudgment rate is the ratio of the number of material misjudgments to the total error score. The original material matching similarity threshold before adjustment, Match the similarity threshold for the adjusted new material;

[0215] When the position offset exceeds 30%, a coordinate calibration instruction is sent to the scheduling control module;

[0216] When the proportion of device malfunctions exceeds 20%, the device will be marked as requiring mandatory maintenance in the device management information;

[0217] By conducting a quantitative analysis of the proportion of causes of misclassification, targeted automated optimization strategies are triggered to achieve dynamic self-calibration of the sorting system. When the proportion of material misclassification is too high, the matching threshold is adaptively adjusted to reduce misclassification caused by similar spectral features. Equipment coordinates are corrected in a timely manner to address position offset issues, improving sorting and positioning accuracy. Maintenance needs are forcibly marked based on the proportion of equipment malfunctions to prevent sorting anomalies caused by degraded mechanical performance and reduce the risk of systemic failures.

[0218] For example, in the continuous operation of a recycling resource sorting line, the reasons for missorting materials are statistically as follows:

[0219] Material misjudgment rate: 65%, exceeding the 60% threshold;

[0220] Position offset ratio: 25%, which does not exceed the 30% threshold;

[0221] Device malfunction rate: 10%, which does not exceed the 20% threshold;

[0222] At this point, automatic optimization is triggered. The specific process of automatic optimization is as follows:

[0223] Make material matching threshold adjustments:

[0224] The original material matching threshold range is [0.6, 0.8], and the error rate corresponds to the adjustment step size. =0.1;

[0225] Then calculate the threshold: ;

[0226] The system automatically tightens material matching standards to improve the accuracy of distinguishing materials with small differences in spectral characteristics, such as clear PET plastic and clear glass.

[0227] Because the position offset and device malfunction ratio have not reached the threshold, coordinate calibration or forced maintenance will not be triggered for the time being, and only subsequent data will be continuously monitored.

[0228] The system also includes a dynamic calibration unit for:

[0229] Regularly calibrate the spectrometer and 3D camera of the image acquisition module;

[0230] Reset the standard parameters of the sorting equipment, that is, adjust is standard atmospheric pressure, is the standard current, is the standard response time;

[0231] Update the matching threshold of the material fingerprint library;

[0232] The above process prevents parameter drift of image acquisition equipment such as spectrometers and 3D cameras due to wear and tear or environmental changes, ensures the accuracy of material identification and spatial density analysis, unifies the operating parameters of sorting equipment, such as air pressure, current, and response time, to avoid inconsistent sorting actions or equipment failures due to parameter deviations, updates the matching threshold of the material fingerprint library, adapts to changes in the types of recycled resources or the introduction of new materials, and maintains high sorting accuracy.

[0233] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0234] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0235] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A renewable resource sorting and management system based on image analysis, characterized in that: include: Image acquisition module, image analysis module, scheduling control module, real-time monitoring module, and re-analysis module; The image acquisition module includes a multispectral imaging unit and a 3D point cloud acquisition unit; The image analysis module performs: (a): Obtain material type by matching multispectral data through the material fingerprint library; (b): Calculate the spatial density distribution of 3D point cloud; (c) Fusion of material type and spatial density distribution to generate renewable resource partition information. The specific process is as follows: The fusion of material type and spatial density distribution is achieved through the fusion evaluation function, and the renewable resource zoning information is generated by optimizing the weight coefficient of the fusion evaluation function; The scheduling control module schedules the sorting equipment according to the partition information; The real-time monitoring module collects equipment status and sorting information; The re-analysis module generates equipment management information and sorting evaluation information; When the image analysis module performs step (c): Constructing fusion evaluation function , the specific process is as follows: ； in is the material recognition confidence, is the density zoning rationality index, and are the weight coefficients of the corresponding items, which are used to adjust the importance of material identification confidence and density zoning rationality index respectively; choose The partition scheme corresponding to the maximum value is the final renewable resource partition information.

2. The renewable resource sorting and management system based on image analysis according to claim 1, characterized in that: The dispatch control module includes an economic decision-making submodule, which is used to: Access the price database to obtain material unit prices , then calculate the value density of each region , and finally by the value density of each region Generate sorting priorities in descending order; The scheduling control module is also used to adjust the conveyor belt speed and receive the instructions of the re-analysis module by adjusting the weight coefficient of the fusion evaluation function of the image analysis module. and , optimize the renewable resource zoning plan.

3. The renewable resource sorting and management system based on image analysis according to claim 2, characterized in that: The real-time monitoring module is also used for: Collect air pressure of pneumatic sorting device , Robotic arm motor current , color sorter response time ; Then calculate the air pressure health index , Current Health Index and Response Time Index ; Then check the air pressure health index , Current Health Index and Response Time Index Calculate and obtain the device health index ; when When the value falls below the threshold, an early warning signal is generated and the abnormal parameter type is marked simultaneously.

4. The renewable resource sorting and management system based on image analysis according to claim 3, characterized in that: The real-time monitoring module is also used for: Obtain sorting quality through weighing units and component detectors Mixing rate with impurities ; Counting missed detections using industrial cameras and wrong fractions ; Generate a sorting information matrix based on the above data in time series , and transmitted to the re-analysis module for sorting evaluation, where T refers to the transpose of the matrix.

5. The renewable resource sorting and management system based on image analysis according to claim 4, characterized in that: The process of generating device management information by the re-analysis module includes: Based on sorting information matrix Time series data to calculate the comprehensive sorting efficiency , where the matrix The row vectors of correspond to different batches of sorting data; when When the missed detection rate is greater than the preset value, the device management information is sent to the scheduling control module to adjust the weight coefficient of the fusion evaluation function of the image analysis module. and ; If the misclassification rate is greater than the preset value, adjust the conveyor belt speed.

6. The renewable resource sorting and management system based on image analysis according to claim 5, characterized in that: The process of generating device management information by the re-analysis module further includes: Comprehensive performance of computing equipment ,according to Value generation equipment management information triggers task allocation, maintenance or shutdown instructions in a hierarchical manner; Calculating the risk of nozzle clogging for pneumatic sorting devices , specifically , where k is a constant, is the change in air pressure, is the air pressure threshold; when When the value is greater than the preset value, additional nozzle detection requirements are added; The process of generating device management information by the reanalysis module also includes: Fault location is performed based on the independent parameter health index. The specific process is as follows: Air pressure health index When the pressure is less than the preset value, it is determined that the pneumatic system is abnormal, and the air pressure pipeline leakage detection of the pneumatic sorting device is triggered; When the current health index If the value is less than the preset value, it is determined that the robot arm load is abnormal, and the motor current waveform is retrieved to analyze the cause of the jam or overload. When the response time index When it is greater than the preset value, it is determined that the color sorter image processing is delayed, and the algorithm time consumption of the image analysis module is automatically optimized.

7. The renewable resource sorting and management system based on image analysis according to claim 6, characterized in that: The specific process of the re-analysis module generating the sorting evaluation information includes: Get sorting information matrix from real-time monitoring module , from the sorting information matrix Extract the qualified quality of sorting , impurity mixing rate , the number of missed detections and wrong fractions ; Get the material unit price from the scheduling control module ; Calculate the data obtained from the real-time monitoring module and the scheduling control module to obtain the batch value recovery rate , then extract the multi-spectral features and 3D contour features of the missed materials in the image acquisition module; The K-means clustering algorithm is used to group the missed detection features. That is, the multispectral features and 3D contour features of the missed detection materials in the image acquisition module are grouped, and the frequency of occurrence of each type of material is counted. Mark the top three material categories with the highest frequency of occurrence as high-frequency loss materials; For high-frequency loss materials, in the image analysis module, the material spectral feature matching weight is increased by a preset value, and the 3D point cloud segmentation sensitivity threshold of such materials is lowered by a preset value; Then diagnose the cause of misclassification, analyze the sorting logs of misclassified materials, and classify and count them. The specific statistical process is as follows: Material misjudgment: The image feature matching similarity is between the threshold range , and The material matching similarity threshold; Position offset: The Euclidean distance between the actual sorting coordinate and the target coordinate is greater than the preset value; Equipment malfunction: The pressure sensor reading of the robotic arm is lower than the standard value; Then the batch value recovery rate , integrate and classify the evaluation information of high-frequency lost materials and misclassification reasons.

8. The renewable resource sorting and management system based on image analysis according to claim 7, characterized in that: The process of diagnosing the cause of the misclassification further includes: When the material misjudgment ratio exceeds the preset value, the material matching threshold is dynamically adjusted. , the specific process is: ； in is the preset adjustment step size, and the misjudgment rate is the ratio of the number of material misjudgments to the total error score. is the original material matching similarity threshold before adjustment, Match the similarity threshold for the adjusted new material; When the position offset ratio exceeds the corresponding threshold, a coordinate calibration instruction is sent to the scheduling control module; When the proportion of device malfunctions exceeds the corresponding threshold, the device will be marked as requiring mandatory maintenance in the device management information.

Images