Urban garbage automatic classification method, system and equipment integrating intelligent identification
By combining a pre-trained waste identification model with spectrometer analysis, the problem of inaccurate waste identification and classification was solved, achieving high-precision automated waste processing and classification.
Patent Information
- Application Number
- CN202511118552.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-11
- Publication Date
- 2025-11-25
AI Technical Summary
Existing urban waste image recognition technologies struggle to accurately distinguish between waste that looks similar or has mixed materials, leading to inaccurate identification and classification.
A pre-trained waste identification model is used for initial identification to screen out the identification error areas. The spectral characteristics of the identification error areas are analyzed by a spectrometer, and the waste material type is determined by combining reputation evaluation and spectral database matching.
It improves the accuracy of waste identification and classification precision, reduces the possibility of misclassification, and realizes automated waste processing and efficient classification.
Smart Images
Figure CN121010818A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of urban waste treatment technology, specifically to an automatic urban waste sorting method, system, and equipment with integrated intelligent identification. Background Technology
[0002] With the acceleration of urbanization, the amount of urban waste generated is increasing day by day, and waste disposal has become an important issue in urban environmental management. In recent years, although some automatic waste sorting methods have emerged, most of them use neural networks to train large amounts of image data to build waste recognition models to handle different types of waste.
[0003] Due to the diverse materials of urban waste, some plastic products (such as plastic film) and paper can be very similar in appearance, especially in color and texture. Image recognition may misidentify plastic as paper, or vice versa. Some metal products (such as aluminum foil) and plastic products (such as plastic film) may be difficult to distinguish in appearance, especially in highly reflective environments. Transparent or translucent plastic products and glass products may also be confused in image recognition, especially in low light or highly reflective conditions. Furthermore, some waste may be made of a mixture of plastic and metal, and image recognition may misidentify these mixed-material wastes as waste of a single material. Therefore, relying on a single image recognition method may lead to errors in material identification, resulting in inaccurate waste classification. Summary of the Invention
[0004] To address the aforementioned technical problems, the purpose of this application is to provide an integrated intelligent identification method, system, and equipment for automatic urban waste sorting. The specific technical solution adopted is as follows:
[0005] In a first aspect, embodiments of this application provide an automatic urban waste sorting method integrating intelligent recognition, the method comprising the following steps:
[0006] Waste images are collected and preprocessed through a waste treatment terminal;
[0007] The preprocessed garbage image is input into a pre-trained garbage recognition model for preliminary garbage material type identification, and the score of each garbage region in the garbage image is output to filter out the recognition error region;
[0008] The spectrometer collects spectral curves at all preset sampling points within the identification error area; the absorption peaks and their intensities are obtained from the spectral curves; the differences in the number and intensity of absorption peaks in the spectral curves of all sampling points within the identification error area are analyzed to determine the credibility evaluation index for each sampling point.
[0009] The slope of the spectral curve at each sampling point in different wavelength ranges is matched with the standard spectral curve in the spectral database based on similarity. The standard spectral curve with the highest similarity match is taken as the standard spectral curve for that sampling point.
[0010] The credibility evaluation indicators are sorted from largest to smallest. The material type of the waste in the identification error area is determined by the standard spectral curves of all sampling points corresponding to the top preset percentage of credibility evaluation indicators.
[0011] Based on the material type identification results of all waste in the waste image, the waste is assigned to different classification areas for processing.
[0012] Preferably, the preprocessing operations include image grayscale conversion, noise removal, size normalization, and edge enhancement.
[0013] Preferably, the pre-trained garbage identification model is obtained by training multiple historical garbage images using a convolutional neural network (CNN).
[0014] Preferably, the method for filtering the identification error region is as follows: when the score is less than a preset score threshold, the garbage region corresponding to the score is recorded as the identification error region.
[0015] Preferably, when the waste area is not an identification error area, the waste type identified by the waste area is used to send the waste in the waste area to the corresponding waste type area.
[0016] Preferably, the intensity is determined by the spectral reflectance value at the lowest point of the absorption peak.
[0017] Preferably, the method for determining the reputation evaluation index for each sampling point is as follows:
[0018] For each sampling point within the identification error region and the remaining sampling points, the average value of the difference in the number of absorption peaks on the spectral curves of each sampling point and the remaining sampling points is calculated.
[0019] Calculate the absolute difference between the number of absorption peaks on the spectral curves of each sampling point and the remaining sampling points and the average value;
[0020] Calculate the product of the difference in intensity of the absorption peak on the spectral curve of each sampling point and the remaining sampling points and the absolute difference;
[0021] The reciprocal of the average of the products calculated for each sampling point and all remaining sampling points is used as the reputation evaluation index for each sampling point.
[0022] Preferably, the method for matching the slope of the spectral curve of each sampling point in different wavelength ranges with the standard spectral curves in the spectral database based on similarity is as follows:
[0023] Obtain the DTW distance of the slope calculated for all corresponding wavelength ranges between the spectral curve of each sampling point and any standard spectral curve in the spectral database; use the reciprocal of the DTW distance as the similarity between the spectral curve of each sampling point and any standard spectral curve in the spectral database.
[0024] Secondly, embodiments of this application provide an integrated intelligent identification urban waste automatic sorting system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the integrated intelligent identification urban waste automatic sorting method as described above.
[0025] Thirdly, embodiments of this application also provide an automatic urban waste sorting device with integrated intelligent identification, wherein the device has a built-in automatic urban waste sorting system with integrated intelligent identification as described above.
[0026] As can be seen from the above embodiments, the integrated intelligent identification method, system, and equipment for automatic urban waste sorting provided in this application have at least the following beneficial effects:
[0027] This application inputs pre-processed waste images into a pre-trained waste recognition model for preliminary waste material type identification, and outputs a score for each waste region in the waste image to filter out recognition error areas. Its beneficial effects are that the pre-trained waste recognition model can quickly perform preliminary classification of waste images and provide a preliminary recognition result, greatly reducing the number of regions that need further analysis; by outputting a score for each waste region, recognition error areas, i.e. those regions with inaccurate recognition results, can be effectively filtered out, thereby avoiding unnecessary spectral analysis of all regions and improving the overall efficiency of the system.
[0028] This application uses a spectrometer to analyze the spectral characteristics of waste in the identification error area to determine the type of waste material. Its beneficial effect is that spectral analysis can accurately measure the spectral reflectance of waste at different wavelengths, thereby determining the chemical composition of waste and effectively distinguishing waste that looks similar but has different compositions, such as plastic and paper, metal and plastic, glass and plastic, etc., significantly improving the accuracy of waste identification. Through spectral analysis, the type of waste material in the identification error area can be further confirmed, improving the accuracy of classification and reducing the possibility of misclassification.
[0029] This application assigns waste to different classification areas for processing based on the material type identification results of all waste in the waste image. It automatically assigns waste to different classification areas according to the identification results, thereby automating waste processing, reducing manual intervention, and improving work efficiency. Attached Figure Description
[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0031] Figure 1 A flowchart illustrating the steps of an integrated intelligent identification method for automatic urban waste sorting, as provided in one embodiment of this application;
[0032] Figure 2 This is a flowchart illustrating a method for analyzing and identifying the spectral characteristics of waste in an error region using a spectrometer, as provided in one embodiment of this application. Detailed Implementation
[0033] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation methods, structures, features, and effects of the integrated intelligent identification urban waste automatic sorting method, system, and equipment proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0034] Unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.
[0035] The following description, in conjunction with the accompanying drawings, details the specific solutions for the integrated intelligent identification method, system, and equipment for automatic urban waste sorting provided in this application.
[0036] Please see Figure 1 The diagram illustrates a flowchart of an integrated intelligent identification method for automatic urban waste sorting according to an embodiment of this application. The method includes the following steps:
[0037] Step 1: Collect waste images through the waste processing terminal and perform preprocessing.
[0038] Automated waste collection vehicles gather municipal waste and transport it to waste treatment centers. At the treatment centers, the waste is first shredded by a crusher to facilitate subsequent sorting. The crusher employs a dual-shaft shearing design, effectively handling various types of waste, including large debris and hard-to-grind items. The shredded waste is kept to a size of less than 10 centimeters to ensure smooth processing.
[0039] The crushed waste is transported to the pre-processing area via a conveyor belt. The conveyor belt speed is automatically adjusted according to the amount of waste to ensure that the waste passes evenly through the subsequent identification and sorting equipment. The pre-processing area is equipped with a spray system to remove dust and impurities from the surface of the waste, improving the accuracy of image recognition.
[0040] The pre-processing area contains a waste collection module responsible for capturing images of the waste on the conveyor belt. Equipped with a high-definition camera, it captures images of the broken-up waste from a top-down view, positioned 1.5 meters from the centerline of the conveyor belt. The camera has a resolution of at least 4K, capable of capturing minute features of the waste. The image data is transmitted via a high-speed network to the pre-processing module at the waste treatment center.
[0041] The collected garbage images are preprocessed, including grayscale conversion, noise removal, size normalization, and edge enhancement. These preprocessing operations are common knowledge and will not be elaborated further.
[0042] Step 2: Input the pre-processed garbage image into the pre-trained garbage recognition model to perform preliminary garbage material type identification, and output a score for each garbage region in the garbage image to filter out the recognition error areas.
[0043] The image recognition module employs deep learning algorithms; in this embodiment, a convolutional neural network (CNN) is used to identify captured garbage images. The recognition model, trained on a large number of garbage images, can accurately identify various types of garbage materials, including plastics, paper, metals, batteries, and kitchen waste, which are then specifically categorized into four main types: recyclable waste, hazardous waste, kitchen waste, and other waste.
[0044] The specific steps for building a trained garbage identification model using CNN are as follows:
[0045] Data input: 10,000 pre-processed garbage images from the past are input into the CNN model;
[0046] Convolutional layers: These layers extract features from garbage images using multiple convolutional layers. Each convolutional layer contains multiple kernels to extract features at different levels.
[0047] Pooling layers: Pooling layers reduce the size of feature maps, thereby reducing computational complexity, while preserving important features;
[0048] Fully connected layer: The extracted features are classified through a fully connected layer, and the probability of each type of garbage is output.
[0049] Output layer: Outputs the probability of each type of waste material and determines the final recognition result based on the set threshold.
[0050] For each garbage region in the garbage image, the CNN model outputs a score to evaluate the accuracy of identifying the type of garbage in that region. The score ranges from 0 to 1, with a higher score indicating higher accuracy. For example, if a region has a score of 0.95, it means that the identification result for that region is very reliable.
[0051] In this embodiment, a scoring threshold T = 0.8 is set. When the score is less than T, the waste area corresponding to that score is recorded as an identification error area, indicating that the identification result in that area may have an error and needs to be further confirmed through spectral analysis. Conversely, when the waste area is not an identification error area, the waste type identified by that waste area is used to send the waste in that area to the corresponding waste type area, completing the automatic waste sorting.
[0052] Accordingly, this application inputs the preprocessed garbage image into a pre-trained garbage recognition model for preliminary image recognition in order to filter out recognition error areas.
[0053] Step 3: Analyze and identify the spectral characteristics of the waste in the error area using a spectrometer to determine the type of waste material.
[0054] To further improve the accuracy of waste identification, especially when image recognition results are inaccurate, spectral analysis technology is used to analyze the spectral characteristics of the identification error region. In this embodiment, a flowchart of the method for analyzing the spectral characteristics of waste in the identification error region using a spectrometer is attached. Figure 2 As shown, the detailed analysis is as follows:
[0055] The spectrometer employs high-resolution grating spectrophotometry to accurately measure the spectral reflectance of waste in the wavelength range of 200 nanometers to 2500 nanometers. By analyzing specific absorption peaks in the spectrum, it is possible to distinguish between different types of waste materials, such as plastics and metals, whose appearances can be easily confused.
[0056] In this application, the spectral analysis module uses a high-resolution spectrometer to measure the spectral reflectance of waste in the identification error region at different wavelengths, thereby determining the chemical composition of the waste. The specific steps of the spectral analysis are as follows:
[0057] A spectrometer is installed above the camera on the conveyor belt to collect spectral data at several pre-defined, evenly distributed sampling points at the location of the waste image while simultaneously capturing the waste image. The spectral data at each sampling point is a spectral curve, composed of the reflectance of the sampling points at different wavelengths. In this embodiment, the spectrometer collects reflected light from the waste surface through a mirror, ensuring the accuracy and reliability of the spectral data. In this embodiment, 100 sampling points are evenly set for the waste image; the specific number can be set by the implementer.
[0058] This application performs spectral feature analysis on the spectral curves collected from all sampling points within the identification error region selected in the previous step.
[0059] First, for the spectral curve collected at any sampling point within the identification error region, the location of the absorption peak is found by calculating the first and second derivatives of the spectral curve. Points with a first derivative of zero are likely the location of absorption peaks, while points with a negative second derivative can further confirm the location of the absorption peak. For each absorption peak, the spectral reflectance value at its lowest point is calculated as the intensity of the absorption peak, used to characterize the absorption capacity of a substance at a specific wavelength. The intensities of different absorption peaks can be used to distinguish different substances.
[0060] To distinguish whether the spectral curves collected from multiple sampling points in the identification error region are accurate, a credibility evaluation is performed on all absorption peaks on the spectral curve collected from each sampling point in the identification error region. The spectral curves of sampling points with higher credibility evaluations are more realistic and can more accurately characterize the spectral characteristics of the waste material in the identification error region where the sampling point is located.
[0061] The credibility evaluation method for each sampling point is as follows: For each sampling point within the identification error region and all remaining sampling points, the average value of the difference in the number of absorption peaks on the spectral curves of each sampling point and all remaining sampling points is calculated; the absolute difference between the difference in the number of absorption peaks on the spectral curves of each sampling point and all remaining sampling points and the average value is calculated; the product of the difference in the intensity of the absorption peaks on the spectral curves of each sampling point and all remaining sampling points and the absolute difference value is calculated; the reciprocal of the average value of the products calculated for each sampling point and all remaining sampling points is used as the credibility evaluation index for each sampling point.
[0062] It should be understood that the higher the credibility rating index, the more likely the absorption peak in the sampling point is to appear in the spectral curves of more sampling points in the identification error region, meaning that the sampling point is more likely to be used to characterize the identification error region.
[0063] Furthermore, features are extracted by calculating the slope of the spectral curve across different wavelength ranges. By dividing the spectral curve into a predetermined number of wavelength ranges, a linear fit is performed on the reflectance values within each wavelength range, and the slope of the fitted line is obtained. This slope can be used to characterize the spectral variation trend of the material within that wavelength range. If the slope is negative, it indicates that the material's ability to absorb light increases with increasing wavelength within that wavelength range; if the slope is positive, it indicates that the material's ability to reflect light increases with increasing wavelength within that wavelength range. In this embodiment, the width of each wavelength range is set to 100 nm. For example, multiple wavelength ranges are defined, such as 400-500 nm, 500-600 nm, 600-700 nm, etc.
[0064] Based on the known spectral database, the slope of the spectral curve of each sampling point in the corresponding wavelength range is matched with the standard spectral curve in the spectral database. The standard spectral curve with the highest similarity is taken as the standard spectral curve of that sampling point.
[0065] The similarity calculation method between the spectral curve of each sampling point and any standard spectral curve in the spectral database is as follows: Obtain the DTW distance (slope calculated for all corresponding wavelength ranges) between the spectral curve of each sampling point and any standard spectral curve in the spectral database; use the reciprocal of the DTW distance as the similarity between the spectral curve of each sampling point and any standard spectral curve in the spectral database. The calculation of the DTW distance is a well-known technique and will not be elaborated further.
[0066] Furthermore, the credibility evaluation indicators are sorted from largest to smallest. The material type of the standard spectral curve of all sampling points corresponding to the credibility evaluation indicators of the top preset percentage in the identification error area is taken as the material type of the garbage in the identification error area. In this embodiment, the preset percentage is set to 5%. When the preset percentage is not an integer, the smallest integer closest to it is taken.
[0067] Step 4: Based on the material type identification results of all the garbage in the garbage image, the garbage is assigned to different classification areas for processing.
[0068] Based on the identification results of the waste material types in the waste images from the above steps, the waste is assigned to different classification areas.
[0069] Recyclable waste is sent to the recycling line for sorting and packaging; hazardous waste is sent to the safe processing area for professional treatment; kitchen waste is sent to the composting area for composting; and other waste is sent to the landfill or incineration area for landfill or incineration.
[0070] Thus, this application optimizes the automatic waste sorting process by adding spectral feature analysis, thereby improving the accuracy and efficiency of waste sorting.
[0071] Based on the same inventive concept as the above method, this application provides an integrated intelligent identification automatic urban waste sorting system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described integrated intelligent identification automatic urban waste sorting method.
[0072] Based on the same inventive concept as the above method, this application also provides an integrated intelligent identification urban waste automatic sorting device, which has a built-in integrated intelligent identification urban waste automatic sorting system as described above.
[0073] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0074] It should be noted that, unless otherwise specified and limited, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a circuit structure, article, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such article or device. Without further limitations, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the article or device that includes said element. Furthermore, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.
[0075] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not invented in this application.
[0076] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.
Claims
1. An automatic urban waste sorting method integrating intelligent recognition, characterized in that, The method includes the following steps: Waste images are collected and preprocessed through a waste treatment terminal; The preprocessed garbage image is input into a pre-trained garbage recognition model for preliminary garbage material type identification, and the score of each garbage region in the garbage image is output to filter out the recognition error region; The spectrometer collects spectral curves at all preset sampling points within the identification error area; the absorption peaks and their intensities are obtained from the spectral curves; the differences in the number and intensity of absorption peaks in the spectral curves of all sampling points within the identification error area are analyzed to determine the credibility evaluation index for each sampling point. The slope of the spectral curve at each sampling point in different wavelength ranges is matched with the standard spectral curve in the spectral database based on similarity. The standard spectral curve with the highest similarity match is taken as the standard spectral curve for that sampling point. The credibility evaluation indicators are sorted from largest to smallest. The material type of the waste in the identification error area is determined by the standard spectral curves of all sampling points corresponding to the top preset percentage of credibility evaluation indicators. Based on the material type identification results of all waste in the waste image, the waste is assigned to different classification areas for processing.
2. The integrated intelligent identification method for automatic urban waste sorting as described in claim 1, characterized in that, The preprocessing operations include image grayscale conversion, noise removal, size normalization, and edge enhancement.
3. The integrated intelligent identification method for automatic urban waste sorting as described in claim 1, characterized in that, The pre-trained garbage identification model is obtained by training multiple historical garbage images using a convolutional neural network (CNN).
4. The integrated intelligent identification method for automatic urban waste sorting as described in claim 3, characterized in that, The method for filtering the identification error region is as follows: when the score is less than the preset score threshold, the garbage region corresponding to the score is recorded as the identification error region.
5. The integrated intelligent identification method for automatic urban waste sorting as described in claim 4, characterized in that, When the waste area is not an identification error area, the waste type identified by the waste area is used to send the waste in the waste area to the corresponding waste type area.
6. The integrated intelligent identification method for automatic urban waste sorting as described in claim 1, characterized in that, The intensity is determined by the spectral reflectance value at the lowest point of the absorption peak.
7. The integrated intelligent identification method for automatic urban waste sorting as described in claim 6, characterized in that, The method for determining the reputation evaluation index for each sampling point is as follows: For each sampling point within the identification error region and the remaining sampling points, the average value of the difference in the number of absorption peaks on the spectral curves of each sampling point and the remaining sampling points is calculated. Calculate the absolute difference between the number of absorption peaks on the spectral curves of each sampling point and the remaining sampling points and the average value; Calculate the product of the difference in intensity of the absorption peak on the spectral curve of each sampling point and the remaining sampling points and the absolute difference; The reciprocal of the average of the products calculated for each sampling point and all remaining sampling points is used as the reputation evaluation index for each sampling point.
8. The integrated intelligent identification method for automatic urban waste sorting as described in claim 1, characterized in that, The method for matching the slope of the spectral curve of each sampling point in different wavelength ranges with the standard spectral curves in the spectral database based on similarity is as follows: Obtain the DTW distance of the slope calculated for all corresponding wavelength ranges between the spectral curve of each sampling point and any standard spectral curve in the spectral database; use the reciprocal of the DTW distance as the similarity between the spectral curve of each sampling point and any standard spectral curve in the spectral database.
9. An integrated intelligent identification-based automatic urban waste sorting system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the integrated intelligent identification method for automatic urban waste sorting as described in any one of claims 1-8 above.
10. An automatic urban waste sorting device integrating intelligent recognition, characterized in that, The device has a built-in integrated intelligent identification automatic urban waste sorting system as described in claim 9.