Detection Method, Device, Equipment and Storage Medium for Plastic Waste

Abstract

The present invention relates to the field of plastic waste detection, and particularly to a detection method, device, equipment and storage medium for plastic waste. The method includes the following steps: obtaining multi-angle images of the area to be detected; performing multi-scale detail reconstruction on the multi-angle images of the area to be detected and suppressing interference of regional background features, so as to construct an optimized image with background suppression; performing visual detection of targets within the area on the optimized image with background suppression and mining the material composition of objects, so as to generate the material composition features of each object; performing deep semantic segmentation based on the material composition features of each object and performing image impurity removal processing to construct an optimized image with non-plastic impurity removal; performing visual detection of plastic waste on the optimized image with non-plastic impurity removal and performing bounding box instance division processing, so as to extract the bounding box images of each plastic waste. The present invention realizes efficient and accurate detection and identification of plastic waste.

Description

Technical Field

[0001] The present invention relates to the field of plastic waste detection, and particularly to a detection method, device, equipment and storage medium for plastic waste. Background Art

[0002] With the continuous enhancement of environmental protection awareness and the implementation of the sustainable development strategy, the recycling and treatment of plastic waste have become an urgent environmental protection issue globally. Due to their light weight, durability, and low cost, plastic materials are widely used in various industries. However, the problems in dealing with these materials after use have become increasingly prominent. The pollution problem of plastic waste to the environment, especially marine pollution, soil pollution, and the destruction of the ecosystem, has attracted extensive attention from all sectors of society. In the current recycling and treatment process, how to efficiently and accurately identify and classify different types of plastic waste has become the key to improving recycling efficiency and reducing environmental pollution.

[0003] Traditional plastic waste detection methods mainly rely on technologies such as manual classification and traditional sensors. However, these methods often face problems such as low efficiency, large classification errors, and high labor costs. Manual classification not only consumes a large amount of time but is also easily affected by human factors, resulting in inaccurate identification and sorting. Traditional sensors have limited ability to distinguish plastic types in complex environments and are easily interfered by other substances. Therefore, how to adopt a more intelligent, efficient, and accurate method to solve the detection and classification problems of plastic waste has become a technical problem that urgently needs to be broken through in the industry. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a detection method, device, equipment and storage medium for plastic waste to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides a detection method for plastic waste, including the following steps:

[0006] Step S1: Obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and suppress the interference of the regional background features, so as to construct an optimized image with background suppression.

[0007] Step S2: Perform in-region target visual detection on the optimized image with background suppression, and excavate the material composition of the object, so as to generate the material composition features of each object.

[0008] Step S3: Perform deep semantic segmentation based on the material composition features of each object, and perform image impurity removal processing to construct an optimized image with non-plastic impurity removal.

[0009] Step S4: Perform visual detection of plastic waste on the optimized image for non-plastic impurity removal, and perform bounding box instance division processing to extract the bounding box image of each plastic waste;

[0010] Step S5: Perform dynamic component learning and iterative detection optimization on the bounding box image of each plastic waste to construct an intelligent waste detection optimization model.

[0011] The present invention also provides a detection device for plastic waste, including:

[0012] An image optimization module, configured to obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and perform interference suppression on the regional background features, so as to construct an optimized image for background suppression;

[0013] A material composition recognition module, configured to perform visual detection of the target within the area on the optimized image for background suppression, and perform object material composition mining, so as to generate the material composition features of each object;

[0014] An impurity removal module, configured to perform deep semantic segmentation based on the material composition features of each object, and perform image impurity removal processing to construct an optimized image for non-plastic impurity removal;

[0015] A bounding box division module, configured to perform visual detection of plastic waste on the optimized image for non-plastic impurity removal, and perform bounding box instance division processing to extract the bounding box image of each plastic waste;

[0016] A detection optimization module, configured to perform dynamic component learning and iterative detection optimization on the bounding box image of each plastic waste to construct an intelligent waste detection optimization model.

[0017] The present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the detection method for plastic waste described in any one of the above are implemented.

[0018] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the detection method for plastic waste described in any one of the above are implemented.

[0019] The beneficial effects of the present invention are specifically as follows: Obtaining multi-angle images helps to comprehensively understand the situation of the area to be detected, providing more information for subsequent processing. Multi-scale detail reconstruction and background feature interference suppression improve the image quality and reduce interference, making subsequent processing more accurate and reliable. Target visual detection and material composition mining help to identify the specific features of each object, providing basic information for subsequent processing. Generating the material composition features of the object helps to further analyze and distinguish different objects, improving the accuracy and reliability of detection. Deep semantic segmentation helps to perform more refined segmentation of objects, improving the accuracy of object recognition. Image impurity removal processing removes interference, making subsequent processing more focused on the main target and improving the overall processing efficiency. Visual detection and bounding box instance division help to accurately locate and identify plastic waste, providing the target object for subsequent processing. Extracting the bounding box images of each plastic waste facilitates further analysis and processing of the waste. Dynamic component learning and iterative detection optimization help to continuously optimize the detection model, improving the accuracy and robustness of waste detection. Constructing an intelligent waste detection optimization model realizes the automation and intelligence of waste detection, improving the waste detection efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic flow chart of the steps of a method for detecting plastic waste according to the present invention;

[0021] Figure 2 is a schematic detailed implementation step flow chart of step S1;

[0022] Figure 3 is a schematic detailed implementation step flow chart of step S2;

[0023] Figure 4 is a schematic detailed implementation step flow chart of step S3. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0025] The embodiments of the present application provide a method, device, equipment and storage medium for detecting plastic waste. The execution subjects of the method, device, equipment and storage medium for detecting plastic waste include, but are not limited to, the following that carry this system: mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be regarded as general computing nodes of the present application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0026] Please refer to Figures 1 to 4 , the present invention provides a method for detecting plastic waste, and the method for detecting plastic waste includes the following steps:

[0027] Step S1: Obtain multi - angle images of the area to be detected; perform multi - scale detail reconstruction on the multi - angle images of the area to be detected, and suppress the interference of regional background features, so as to construct an optimized image with background suppression.

[0028] Step S2: Perform in - region target visual detection on the optimized image with background suppression, and mine the material composition of the objects, so as to generate the material composition features of each object.

[0029] Step S3: Based on the material composition features of each object, perform deep semantic segmentation, and perform image impurity removal processing to construct an optimized image with non - plastic impurity removal.

[0030] Step S4: Perform visual detection of plastic waste on the optimized image with non - plastic impurity removal, and perform bounding box instance division processing, so as to extract the bounding box images of each plastic waste.

[0031] Step S5: Perform dynamic component learning and iterative detection optimization on the bounding box images of each plastic waste to construct an intelligent waste detection optimization model.

[0032] The present invention helps to comprehensively understand the situation of the area to be detected by obtaining multi - angle images, providing more information for subsequent processing. Multi - scale detail reconstruction and background feature interference suppression improve the image quality and reduce interference, making subsequent processing more accurate and reliable. Target visual detection and material composition mining help to identify the specific features of each object, providing basic information for subsequent processing. Generating the material composition features of the objects helps to further analyze and distinguish different objects, improving the accuracy and reliability of detection. Deep semantic segmentation helps to perform more refined segmentation of the objects, improving the accuracy of object recognition. Image impurity removal processing removes interference, making subsequent processing more focused on the main target and improving the overall processing efficiency. Visual detection and bounding box instance division help to accurately locate and identify plastic waste, providing target objects for subsequent processing. Extracting the bounding box images of each plastic waste enables better further analysis and processing of the waste. Dynamic component learning and iterative detection optimization help to continuously optimize the detection model, improving the accuracy and robustness of waste detection. Constructing an intelligent waste detection optimization model realizes the automation and intelligence of waste detection, improving the efficiency and accuracy of waste detection.

[0033] In the embodiment of the present invention, refer to Figure 1 , which is a schematic diagram of the step - by - step process of a method for detecting plastic waste of the present invention. In this example, the steps of the method include:

[0034] Step S1: Obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and suppress the interference of the regional background features, so as to construct an optimized image with background suppression;

[0035] In this embodiment, a suitable imaging device is selected, such as a high-resolution camera or a multispectral camera, to ensure that high-quality multi-angle images can be obtained. These devices should have good optical performance and be able to work properly under different lighting conditions. Determine the environmental conditions of the area to be detected, ensure uniform lighting, and avoid interference from shadows and reflections on image acquisition. If necessary, use artificial lighting or blockers to optimize the lighting conditions. Set the parameters of the camera, such as focal length, shutter speed, and ISO value, to adapt to different shooting scenarios and the characteristics of the target object. Take images of the area to be detected from different angles and positions. For example, rotate the camera around the target object or adjust the pitch angle of the camera to ensure that the complete characteristics of the target object are captured from multiple perspectives. Take multiple images at each angle to increase data diversity, especially when the surface characteristics of the target object are complex. Improve the image quality through multiple shootings. Classify and store the collected images according to the shooting angle and time to ensure quick access and use during subsequent processing. Record the shooting parameters and environmental conditions of each image for subsequent analysis. Preprocess the obtained multi-angle images and apply denoising algorithms (such as Gaussian filtering, median filtering, etc.) to reduce the noise impact in the images and ensure the accuracy of detail reconstruction. According to the angle and position during shooting, register multiple images to ensure that images from different angles are processed in the same coordinate system. This step is implemented using feature matching algorithms (such as SIFT or ORB). Apply techniques such as wavelet transform or Laplacian pyramid to perform detail reconstruction on multiple aligned images. These methods can extract the detail information of the images at different scales. By separating the high-frequency information (details) from the low-frequency information (background), the details of the target object are retained during the reconstruction process, while unnecessary background information is suppressed. Synthesize the detail information at different scales to generate a reconstructed image containing multi-scale details. This process is achieved through weighted averaging or fusion techniques to ensure the natural transition of details. Use a background modeling algorithm (such as Gaussian mixture model) to analyze the background characteristics in the reconstructed image, extract the statistical characteristics of the background, and help identify which parts belong to the background and which parts are the target object. Divide the image into different regions (for example, based on superpixel segmentation technology) for interference suppression at the region level. Each region will be analyzed to determine whether it contains the target object or is interfered by the background. Apply an interference suppression algorithm to the extracted background characteristics and use background subtraction technology to subtract the background information from the reconstructed image, thereby highlighting the foreground target. Combine morphological processing (such as opening operation and closing operation) to further clean the image, remove noise and small interference regions, and ensure the clarity of the target object. The finally generated image should be an image optimized for background suppression, which can clearly display the target object and reduce the impact of background feature interference on subsequent analysis. Record all the parameters and results during the processing to ensure the traceability of subsequent analysis.

[0036] Step S2: Perform in-region target visual detection on the background suppression optimized image, and conduct object material composition mining, so as to generate the material composition features of each object;

[0037] In this embodiment, the image to be analyzed is extracted from the generated background suppression optimized image to ensure that the image maintains high quality and clarity for subsequent analysis. The image is preprocessed, such as normalization and enhancement, to improve the accuracy of visual detection. A suitable target detection algorithm is selected, such as YOLO (You Only Look Once), Faster R-CNN, or SSD (Single Shot Multi Box Detector), which are known for their fast and efficient detection capabilities and can detect multiple targets in the image simultaneously. Set the parameters of the algorithm according to specific requirements, such as the size of the input image, the confidence threshold, and the non-maximum suppression (NMS) parameters, to optimize the detection results. Input the background suppression optimized image into the selected target detection model to detect the in-region targets. The model will output the bounding box, category, and confidence score of each target. Record the detection results, including the position information (such as coordinates and dimensions), category label, and confidence of each object for subsequent analysis. Select a suitable material composition mining method, such as multispectral feature analysis, spectral reflectance measurement, or learning-based methods (such as features extracted by convolutional neural networks). Determine the types of material features to be extracted, such as spectral features, texture features, and gloss attributes, which can effectively distinguish objects of different materials. Process each detected target region, extract the corresponding image data, and crop the image of each object according to the bounding box information to ensure that only the target region is analyzed. Preprocess the cropped image (such as normalization and enhancement) to improve the effect of subsequent feature extraction. Apply the selected material composition mining technology to extract the corresponding features from the image of each object. For example, use spectral analysis technology to obtain the spectral curve of the object, or use texture analysis methods to extract surface features. Record the material composition features of each object, including spectral features, texture features, and gloss attributes, etc., which will be used as the basis for subsequent material identification and classification. Organize the detection information and corresponding material composition features of each object into a structured form and record them in a database or file for subsequent retrieval and analysis, ensuring that the detection results of each object correspond to its material features for convenient subsequent research and application.

[0038] Step S3: Perform deep semantic segmentation based on the material composition features of each object, and conduct image impurity removal processing to construct a non-plastic impurity removal optimized image;

[0039] In this embodiment, the background suppression optimized images obtained are used as inputs. These images should have removed most of the background interference to ensure the clarity of the target object, and ensure that the quality and resolution of the images are suitable for deep semantic segmentation to guarantee the accuracy of subsequent processing. A suitable deep learning semantic segmentation model is selected, such as U-Net, DeepLabv3+ or MaskR-CNN. These models perform excellently in image segmentation tasks and can effectively identify and separate the target object from the background. According to the specific task requirements, the hyperparameters of the model are set, such as the learning rate, batch size and loss function, to optimize the training effect. For the input images, manual annotation is carried out to generate segmentation labels for training. These labels should clearly identify each target object and its boundary to ensure the accuracy of the training data. The annotated dataset is divided into a training set, a validation set and a test set for model training and evaluation. The selected deep learning model is trained using the annotated dataset. Through the backpropagation algorithm, the model parameters are continuously optimized to improve the segmentation accuracy. The loss value and accuracy during training are recorded to ensure that the model converges during training, and the training strategy is adjusted according to the performance of the validation set. The trained semantic segmentation model is applied to the background suppression optimized images for deep semantic segmentation. The model will output the class label of each pixel to identify the object and the background. The generated segmentation result will include the boundary information and class label of each object, providing a basis for subsequent impurity removal. According to the segmentation result, the non-plastic impurity parts in the image are identified. These impurities include fragments of other materials, dirt or noise. A threshold-based segmentation method or a region-growing algorithm is used to determine which regions should be regarded as impurities. For the identified non-plastic impurity regions, removal processing is carried out. Using a masking operation, these regions are deleted from the image or replaced with background information. Combining morphological processing techniques (such as opening and closing operations), the image after removal is further cleaned to ensure the integrity and clarity of the target object. The image after impurity removal is saved as a non-plastic impurity removal optimized image. This image should clearly display the target object and be free from impurity interference. The parameters and results during the processing are recorded to ensure the traceability of the removal process to support subsequent analysis and application.

[0040] Step S4: Perform visual detection of plastic waste on the non-plastic impurity removal optimized image and carry out bounding box instance division processing to extract the bounding box images of each plastic waste.

[0041] In this embodiment, the obtained optimized images of non-plastic impurities are used as input. These images should clearly show the characteristics of the target objects and have removed the redundant background interference to ensure that the quality and resolution of the images are suitable for visual inspection, so as to improve the accuracy of subsequent processing. Select a suitable object detection algorithm, such as YOLO, FasterR-CNN or SSD. These algorithms can efficiently detect and locate multiple targets in the image. Adjust the hyperparameters of the model according to the specific task requirements, such as the size of the input image, the confidence threshold and the non-maximum suppression (NMS) threshold, to optimize the detection effect. For the optimized images of non-plastic impurities removal, perform manual annotation to generate object detection labels for training. These labels should accurately identify the bounding boxes and categories of each plastic waste to ensure the accuracy of the training data. Divide the annotated dataset into a training set, a validation set and a test set for the training and evaluation of the model. Use the annotated dataset to train the selected object detection model, and continuously optimize the model parameters through the backpropagation algorithm to improve the detection accuracy. Record the loss value and accuracy rate during the training process to ensure that the model converges during training and adjust the training strategy according to the performance of the validation set. Apply the trained object detection model to the optimized images of non-plastic impurities removal for the visual detection of plastic waste. The model will output the bounding boxes, category labels and confidence scores of each detected plastic waste. Record the detection results, including the position information (such as coordinates and dimensions), category labels and confidence of each plastic waste for subsequent analysis. According to the bounding box information of the detection results, perform instance segmentation on each detected plastic waste to ensure that each bounding box accurately contains the corresponding plastic waste and does not overlap. Combine the non-maximum suppression (NMS) algorithm to remove the bounding boxes with an overlap degree higher than the set threshold to ensure the accuracy of the final detection results. Extract the bounding box images of each plastic waste from the optimized images of non-plastic impurities removal. According to the detected bounding box coordinates, crop the images of each object to ensure that only the target area is analyzed. Record the position information and dimensions of each bounding box image for subsequent application and analysis. Organize the detection information of each plastic waste and the corresponding bounding box images into a structured form and record them in a database or file for subsequent retrieval and analysis to ensure that the detection results of each plastic waste correspond to its bounding box image for convenient subsequent research and application.

[0042] Step S5: Perform dynamic component learning and iterative detection optimization on the bounding box images of each plastic waste to construct an intelligent waste detection optimization model.

[0043] In this embodiment, the bounding box images of each plastic waste are obtained to ensure that these images cover different types of plastic waste, so that the model can learn diverse features. The images are preprocessed, including normalization and enhancement, to improve the learning effect, including techniques such as random rotation, scaling, and color adjustment to increase the diversity of the data. A suitable dynamic component learning algorithm is selected. It is recommended to use transfer learning or incremental learning methods, which can quickly update the model when new data arrives, improving adaptability and real-time performance. The key parameters of learning are determined, such as the learning rate, update frequency, and optimization algorithm (such as Adam or SGD), to adapt to the dynamic data stream. A pre-trained object detection model (such as YOLO, FasterR-CNN) is selected and fine-tuned according to the collected bounding box images. Fine-tuning enables the model to better adapt to the specific characteristics of plastic waste. The previously annotated bounding box data is used for initial training to ensure that the model can effectively identify and distinguish different types of plastic waste. During the model training process, new bounding box image data is continuously introduced and used for dynamic learning. The model will update its parameters in real time to adapt to the feature changes in the new data. The performance of the model after each update is recorded, including metrics such as accuracy and recall, to evaluate the learning effect. The detection performance of the model is evaluated regularly, using the validation set for testing, and various metrics (such as precision, recall, and F1-score) are calculated to ensure the robustness of the model under different conditions. The performance of the model on specific types of plastic waste is analyzed to identify areas or categories where the model performs poorly. Based on the evaluation results, optimization strategies are implemented. For example, the hyperparameters of the model are adjusted, the diversity of the training data is increased, or a more complex model architecture is introduced. The method of ensemble learning is used to combine the prediction results of multiple models to improve the overall detection performance. The iteratively optimized model is integrated into an intelligent waste detection optimization model to ensure that the model can respond quickly and perform accurate detection in practical applications. Final model validation is carried out to ensure that it can operate stably under different environments and conditions. The architecture, training process, performance evaluation, and optimization strategies of the model are recorded for subsequent maintenance and iterative updates. The model is updated regularly to adapt to new data changes and detection requirements to ensure its continuous effectiveness.

[0044] In this embodiment, refer to Figure 2 , which is a schematic diagram of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0045] Step S11: Obtain multi-angle images of the area to be detected;

[0046] Step S12: Perform multi-scale detail reconstruction on the multi-angle images of the area to be detected to construct multi-scale detail reconstruction images;

[0047] Step S13: Perform environmental background visual analysis on the multi-scale detail reconstruction image to generate the environmental background visual features in the image;

[0048] Step S14: Capture multi-angle features based on the environmental background visual features in the image, so as to generate multi-angle environmental background visual features;

[0049] Step S15: Suppress the interference of regional background features on the multi-scale detail reconstruction image based on the multi-angle environmental background visual features, so as to construct an optimized image with background suppression.

[0050] In this embodiment, a high-resolution camera or camera is selected to ensure that the device can capture the details of the area to be detected. If necessary, multiple cameras are used to achieve multi-angle shooting to improve the comprehensiveness of the image. The camera parameters (such as exposure time, ISO, white balance, etc.) are configured to adapt to different lighting conditions to ensure the image quality. A detailed shooting plan is formulated to determine the position and shooting angle of each camera. Generally, all sides of the area to be detected should be covered to ensure there are no blind spots. A site survey is carried out to ensure the stability and safety of the shooting position, and to avoid occlusion during the shooting process. At the set shooting time, the imaging device is started to perform real-time acquisition of multi-angle images, ensuring that each camera shoots according to the predetermined angle and position.

[0051] Use synchronous triggers to ensure that all cameras capture images simultaneously, so as to ensure the temporal consistency of images during subsequent processing. Convert the multi-angle images collected into a unified format (such as JPEG, PNG) for subsequent processing, ensure that the resolution and color mode of the images are consistent, perform image denoising and enhancement processing, use filters (such as Gaussian filtering) to remove image noise, and improve the quality of subsequent detail reconstruction. Select appropriate multi-scale reconstruction algorithms, such as wavelet transform, Laplacian pyramid, or deep learning models (such as convolutional neural networks), etc. These methods can effectively extract and reconstruct image details at different scales. Determine the reconstruction scale parameters to retain sufficient detail information during the reconstruction process. Apply the selected reconstruction algorithm to process the multi-angle images and generate multi-scale detail reconstruction images. Through a multi-level reconstruction method, enhance the details and clarity of the images. Fuse the reconstructed images to form a comprehensive image containing rich details, ensuring the effective integration of information from different angles. Select appropriate visual parsing methods, such as image segmentation, edge detection, or deep learning-based feature extraction models (such as Mask R-CNN). These methods can extract the feature information of the environmental background from the images. Determine the parameters and models for feature extraction for subsequent processing. Perform visual parsing of the environmental background on the multi-scale detail reconstruction images, extract the main background features (such as color, texture, shape, etc.) in the images, and generate a feature map of the environmental background. Record the spatial position and importance of each feature for subsequent analysis and application. Select appropriate feature matching and fusion techniques, such as SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), or ORB, to extract multi-angle features of the environmental background. Determine the parameters for feature matching to adapt to the changes in features at different angles. Based on the extracted visual features of the environmental background, capture multi-angle features of the images. Through the feature matching algorithm, compare the images at different angles, extract the common background features, and generate a multi-angle environmental background visual feature dataset. Record the performance and spatial distribution of each feature at different angles. Select appropriate interference suppression algorithms, such as background modeling (such as Gaussian mixture model), image fusion technology, or deep learning models (such as autoencoders) to suppress background interference. Determine the parameters and algorithms for suppression to adapt to different types of interference. Input the multi-angle environmental background visual features into the suppression algorithm to perform interference suppression on the regional background features. By comparing the background features with the interference features, remove unnecessary interference information and generate an optimized image with background suppression, ensuring that important target features are retained while effectively eliminating the impact of background interference. Store the generated optimized image with background suppression in the database to ensure the traceability and integrity of the images. Record the key parameters during the processing for subsequent analysis and improvement. Evaluate the optimization results, analyze the suppression effect, and provide a reliable basis for subsequent detection and application.

[0052] In this embodiment, the specific steps of step S12 are as follows:

[0053] Perform multi-region equal-ratio segmentation on multi-angle images of the area to be detected to obtain multiple image regions;

[0054] Calculate the local contrast between regions for multiple image regions to obtain multiple local contrasts between regions;

[0055] Perform adaptive contrast enhancement based on multiple local contrasts between regions to generate a contrast-enhanced image;

[0056] Analyze the brightness distribution of the contrast-enhanced image to generate image brightness distribution data;

[0057] Perform non-linear brightness equalization based on the image brightness distribution data to construct a brightness-equalized enhanced image;

[0058] Perform discrete wavelet transform on the brightness-equalized enhanced image to obtain multiple images of different scales;

[0059] Calculate each frequency component for multiple images of different scales to generate frequency component features for each scale;

[0060] Perform edge detail sharpening for the corresponding scale based on the frequency component features for each scale to generate edge detail sharpening features for each scale;

[0061] Perform multi-scale feature fusion and reconstruction on multiple images of different scales based on the edge detail sharpening features for each scale to construct a multi-scale detail reconstruction image.

[0062] In this embodiment, an image to be processed is obtained from a multi-angle image set to ensure the clarity and resolution of the image for subsequent processing. The image is preprocessed, such as denoising and enhancement, to ensure that the image quality meets the segmentation requirements. The uniform grid segmentation algorithm is used to divide the image into multiple small regions. Image processing libraries such as OpenCV are used to set the same width and height for segmentation, and the segmentation ratio and the number of regions are determined and adjusted according to the characteristics of the image and the analysis requirements. The selected segmentation algorithm is used to process the image of the region to be detected, generating multiple equi-ratio segmented image regions. The size and number of each region should be within the preset range. The position information and features of each region are recorded for subsequent processing and analysis. A local contrast calculation method is selected, such as Local Contrast Enhancement or standard deviation calculation, to evaluate the contrast by calculating the degree of change of pixel values within each region. The size of the calculation window is determined to facilitate the analysis of contrast changes within the local region. The local contrast of each segmented image region is calculated, generating multiple contrast values between regions. The mean and standard deviation of pixel intensity are used to calculate the contrast, and the calculation results are recorded in a data structure for subsequent contrast enhancement processing. Methods such as Contrast Limited Adaptive Histogram Equalization (CLAHE) or gamma correction are used for contrast enhancement to adaptively adjust the contrast in different regions. The parameters for enhancement, such as the block size and contrast limit, are determined to optimize the enhancement effect. The calculated local contrast values are applied to the adaptive contrast enhancement algorithm to generate a contrast-enhanced image, ensuring that image details are retained during the enhancement process. The parameters and effects of the enhanced image are recorded for subsequent analysis and optimization. The histogram method is used to calculate the brightness distribution of the image, recording the number of pixels with different brightness values to generate brightness distribution data. Tools such as OpenCV or MATLAB are used to generate the histogram and analyze its shape and features. The brightness distribution of the contrast-enhanced image is analyzed to generate brightness distribution data, including the mean, variance, etc. of the brightness values, and the analysis results are recorded for subsequent non-linear brightness equalization processing. A non-linear brightness equalization algorithm is selected, such as adaptive histogram equalization, or gamma correction is used to achieve brightness equalization. The parameters for equalization are determined to optimize the brightness distribution of the image. The brightness distribution data is input into the non-linear equalization algorithm to process the contrast-enhanced image, generating a brightness equalization-enhanced image. The parameters and effects of the equalization process are recorded and visually displayed to ensure that users can intuitively understand the processing results. The Discrete Wavelet Transform (DWT) method is selected to decompose the image into frequency components of different scales through wavelet functions. Appropriate wavelet bases (such as Haar, Daubechies) and the number of decomposition levels are selected to ensure the processing effect. The brightness equalization-enhanced image is subjected to discrete wavelet transform, generating multiple images of different scales. The details and approximation components of each scale are recorded, and the transformation results are saved for subsequent frequency component calculation and feature extraction.Select a frequency component calculation method, such as amplitude calculation after Fourier transform or wavelet transform, to extract the frequency features of each scale, determine the calculation parameters to optimize the extraction effect of frequency features, calculate each frequency component for the wavelet transform result of each scale, generate the frequency component features of each scale, record the calculated features in a data structure for convenient subsequent sharpening processing, select a suitable edge detail sharpening method, such as Laplacian operator, Sobel operator or high-pass filter, to enhance the edge details of the image, determine the sharpening parameters, such as the size of the filter and the enhancement factor, to ensure the processing effect, perform edge detail sharpening processing for the corresponding scale based on the frequency component features of each scale, generate the edge detail sharpening features of each scale, record the parameters and effects of the sharpening results for subsequent analysis and display, select a multi-scale feature fusion method, such as weighted average, max pooling or deep learning method for feature fusion, to generate a comprehensive multi-scale reconstructed image, determine the fusion strategy and weights to adapt to the importance of different scale features, perform multi-scale feature fusion reconstruction on multiple images of different scales based on the edge detail sharpening features of each scale, and generate the final multi-scale detail reconstructed image to ensure rich image details and good visual effects.

[0063] In this embodiment, refer to Figure 3 , which is a schematic diagram of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0064] Step S21: Perform visual detection of the target within the region on the background suppression optimized image, and mark the object nodes within multiple regions;

[0065] Step S22: Mine the multi-spectral features of the target nodes within multiple regions to generate the spectral features of each object;

[0066] Step S23: Analyze the surface texture features of each object spectral feature to obtain the surface texture features of each object;

[0067] Step S24: Analyze the visual gloss attribute of each object spectral feature and extract the visual gloss attribute of each object;

[0068] Step S25: Mine the material composition of each object based on the visual gloss attribute of each object and the surface texture features of each object, thereby generating the material composition features of each object.

[0069] In this embodiment, the region to be detected is extracted from the background-suppressed and optimized image to ensure the quality and clarity of the image for subsequent object detection. The image is preprocessed, such as denoising and enhancement, to improve the accuracy of object detection. A suitable object detection algorithm is selected, such as YOLO, Faster R-CNN, or SSD, which can quickly and accurately detect multiple objects in the image. The configuration parameters of the algorithm, such as the input image size, detection threshold, etc., are determined to optimize the detection effect. The background-suppressed and optimized image is input into the selected object detection model for object detection within the region. The model will identify each object and generate its bounding box. The detected object nodes are marked, and the position information of each object (such as coordinates and dimensions) is recorded to generate the object detection result data. The multispectral imaging technology is selected. By collecting multispectral images of each target node, spectral information in different bands is obtained. A multispectral camera or sensor is used for data collection, and the band range to be collected is determined to cover the characteristic spectrum of the target material. For each marked target node, the multispectral features are extracted. By analyzing the reflectivity in different bands, the spectral feature data of each object is generated. The spectral data is recorded in the database, including band information, reflectivity values, etc., for subsequent analysis and comparison. Methods such as gray-level co-occurrence matrix (GLCM), local binary pattern (LBP), or wavelet transform are used for surface texture analysis. By analyzing the details of the texture, the texture features of the object surface are extracted. The parameters for analysis, such as the window size and the number of features, are determined to ensure the comprehensiveness and accuracy of the analysis. The surface texture feature analysis is performed on the spectral feature data of each object to generate the feature data related to the surface texture of the object (such as contrast, entropy, uniformity, etc.). The extracted texture features are recorded in the data structure for subsequent material composition mining. The visual gloss attribute analysis method is selected, such as reflectivity analysis, gloss measurement, or a deep learning-based gloss feature extraction model. These methods can accurately evaluate the gloss characteristics of the object surface. The parameters and models for analysis are determined to adapt to the gloss characteristics of different materials. The visual gloss attribute analysis is performed on the spectral features of each object to extract the relevant attribute features such as glossiness and reflectivity. The data of the gloss attributes, including the reflection characteristics and glossiness index of the object surface, are recorded for subsequent material composition analysis. A suitable material composition analysis method is selected, such as support vector machine (SVM), random forest, or a deep learning-based classification model. These models can identify the composition of materials based on spectral and texture features. The input features of the model are determined, including spectral features, texture features, and gloss attributes, for accurate classification. The visual gloss attributes and surface texture features of each object are input into the material composition analysis model for object material identification and classification, and the material composition feature data of each object, including the identified material type and related feature values, are generated and recorded in the database for subsequent use.

[0070] In this embodiment, refer to Figure 4 , which is a schematic diagram of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of the step S3 include:

[0071] Step S31: Based on the material composition characteristics of each object, identify the non-plastic impurity characteristics of the object nodes in multiple regions, and extract multiple non-plastic impurities;

[0072] Step S32: Perform depth semantic segmentation on the background suppression optimized image to obtain multiple object category region images;

[0073] Step S33: Perform image impurity removal processing on the multiple object category region images according to the multiple non-plastic impurities to construct a non-plastic impurity removal optimized image.

[0074] In this embodiment, the material composition features of each object extracted in the previous steps are collected and sorted out, including spectral features, texture features, and gloss attributes. These features will be used to distinguish plastics from non-plastic impurities, ensuring the accuracy and consistency of the data for subsequent impurity identification. A suitable machine learning algorithm, such as random forest, support vector machine (SVM), or a deep learning model (such as a convolutional neural network), is selected to identify non-plastic impurities. These algorithms can classify based on material features. The hyperparameter configuration and training strategy of the identification model are determined so that the model can effectively distinguish different materials. The collected material composition features are input into the selected identification model for the feature identification of non-plastic impurities. The model will output the impurity type and its confidence level of each object. The identification results are recorded, including the types and quantities of the identified non-plastic impurities and their positions in the image, forming an impurity feature report. The background-suppressed optimized image is used as the input to ensure the quality and clarity of the image for semantic segmentation. A suitable deep learning semantic segmentation model, such as U-Net, DeepLab, or MaskR-CNN, is selected. These models can effectively identify and segment different object categories in the image. If a pre-trained model is used, it is fine-tuned according to the dataset. If training from scratch, a labeled training dataset needs to be prepared to ensure that there are sufficient samples for each object category in the image. The training parameters, such as the learning rate, batch size, and number of training epochs, are determined to optimize the model performance. The background-suppressed optimized image is input into the deep semantic segmentation model to perform the segmentation operation and generate multiple object category region images. The model will output the category label of each pixel. The segmentation results are recorded, including the region boundaries and pixel information of each object category for subsequent processing and analysis. According to the identified non-plastic impurities, corresponding image processing strategies are formulated. A pixel-based mask processing or region elimination method is selected, and the elimination conditions, such as the size, shape, and position of the impurities, are determined to ensure the effectiveness of the elimination process. According to the identified non-plastic impurity information, the object category region images generated after deep semantic segmentation are processed. The image processing library (such as OpenCV) is used to implement the elimination of impurities, generating an optimized image with non-plastic impurities removed. The eliminated regions are filled or smoothed to ensure the naturalness and integrity of the image. The generated optimized image with non-plastic impurities removed is saved in the database to ensure the traceability of the processing results. The key parameters and effects during the elimination process are recorded, and the elimination results are evaluated to analyze the effectiveness of the elimination and its impact on the overall image quality, providing a basis for subsequent optimization.

[0075] In this embodiment, step S4 includes the following steps:

[0076] Step S41: Conduct visual detection of plastic waste on the optimized image with non-plastic impurities removed, and mark each plastic waste node;

[0077] Step S42: Perform regional precise positioning calculation on each plastic waste node to obtain the positioning coordinates of each plastic waste node;

[0078] Step S43: Based on the positioning coordinates of each plastic waste node, perform three-dimensional shape analysis on each waste item one by one to generate the three-dimensional shape characteristics of each plastic waste item;

[0079] Step S44: According to the three-dimensional shape characteristics of each plastic waste item, perform bounding box instance division processing on the non-plastic impurity removal and optimized image, so as to extract the bounding box image of each plastic waste item.

[0080] In this embodiment, the image to be processed is extracted from the optimized image of non-plastic impurity removal to ensure its quality and clarity for subsequent plastic waste detection. The image is denoised and preprocessed to improve the accuracy of subsequent detection. A suitable object detection algorithm is selected, such as YOLO (You Only Look Once), Faster R-CNN, or EfficientDet. These algorithms effectively identify and label the plastic waste nodes. The hyperparameters of the algorithm are set, such as the detection threshold and confidence level, to optimize the detection effect. The optimized image after removal is input into the selected object detection model for visual detection of plastic waste. The model will output the bounding box and class label of each detected plastic waste. The detection results are recorded, including the position information (such as coordinates and dimensions) of each plastic waste node, and marked on the image to form a detection result image. A suitable positioning algorithm is selected, such as center point positioning or bounding box calculation, to ensure that the coordinate information of each plastic waste node can be accurately obtained. The calculation parameters are determined, such as positioning accuracy and coordinate system setting, for subsequent analysis. For each detected plastic waste node, regional precise positioning calculation is performed. According to the bounding box information in the detection results, the center coordinates of each node are calculated. The positioning results are recorded in a data structure, including the precise coordinates (x, y) of each waste node and the corresponding bounding box dimensions for subsequent processing. A suitable three-dimensional shape analysis method is selected, such as point cloud-based analysis, three-dimensional reconstruction technology, or shape descriptor extraction (such as Hu moments). These methods can effectively analyze the three-dimensional characteristics of plastic waste. The three-dimensional characteristic indexes to be analyzed are determined, such as volume, surface area, contour characteristics, etc. Based on the positioning coordinates of each plastic waste node, the three-dimensional point cloud data or image information of the corresponding area is extracted for three-dimensional shape analysis of each waste one by one, generating the three-dimensional shape characteristic data of each plastic waste, including volume, surface shape, contour, and other geometric characteristics, and recording them in the database. A suitable bounding box division algorithm is selected. Common methods include pixel-based mask processing or region growing algorithm, to ensure that the accurate bounding box can be extracted according to the three-dimensional shape characteristics. The division criteria and parameters are determined to ensure that different plastic waste instances can be effectively distinguished. According to the three-dimensional shape characteristics of each plastic waste, the optimized image after non-plastic impurity removal is subjected to bounding box instance division processing. Using the previously calculated positioning coordinates and three-dimensional characteristic data, the bounding box image of each plastic waste is accurately extracted, generating the final bounding box image of each plastic waste, and recording the position information and dimensions of the bounding box for subsequent application and analysis.

[0081] In this embodiment, step S5 includes the following steps:

[0082] Step S51: Perform dynamic component learning on the bounding box image of each plastic waste to generate plastic waste detection learning data;

[0083] Step S52: Calculate the detection accuracy of the plastic waste detection learning data to generate waste detection accuracy parameters;

[0084] Step S53: Based on the waste detection accuracy parameters, perform iterative detection optimization to construct an intelligent waste detection optimization model.

[0085] In this embodiment, collect the bounding box images of each plastic waste to ensure good image quality for subsequent dynamic component learning. These images should contain plastic wastes under different angles and different background conditions. Perform appropriate preprocessing on the images, such as normalization and enhancement, to improve the effect of the learning process. Select a suitable dynamic learning algorithm, such as transfer learning, online learning, or incremental learning. These methods can quickly update the model when new data arrives and continuously improve the detection performance. Determine the parameter settings of learning, such as the learning rate, batch size, and update frequency, to adapt to the dynamic data stream. Input the bounding box images of each plastic waste into the dynamic learning model for feature extraction and learning. The model will update its internal weights according to the input data and gradually improve the recognition ability of plastic wastes. Record the loss function and accuracy rate during the learning process to ensure that the model is continuously optimized during the learning process. The generated learning data will be used for subsequent detection tasks. Select a suitable detection accuracy calculation method, such as Precision, Recall, and F1-score. These metrics can comprehensively reflect the detection ability of the model. Determine the evaluation threshold and criteria to facilitate the effective classification and evaluation of the detection results. Use the plastic waste detection learning data to evaluate the detection results of the model. According to the true labels and prediction results, calculate various accuracy metrics and record the evaluation results, including the values of each metric and the performance of the model under different conditions, providing a basis for subsequent optimization. Select a suitable model optimization method, such as hyperparameter tuning, model ensemble, or using a more complex network architecture. The goal of optimization is to improve the detection accuracy and robustness. Determine the optimization strategy and process to facilitate systematic iterative optimization under the guidance of the detection accuracy parameters. Based on the waste detection accuracy parameters, perform iterative detection optimization. For the aspects with poor performance in the evaluation results, adjust the model structure or training process, and conduct a new round of training and evaluation. Record the optimized detection accuracy parameters. Through multiple iterations, continuously improve the model performance, and finally construct an intelligent waste detection optimization model.

[0086] In this embodiment, the present invention also provides a detection device for plastic waste, including:

[0087] An image optimization module, configured to obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and suppress interference of regional background features, so as to construct a background-suppressed optimized image;

[0088] A material composition recognition module, configured to perform visual detection of targets within the area on the background-suppressed optimized image, and mine the material composition of objects, so as to generate the material composition features of each object;

[0089] An impurity removal module, configured to perform deep semantic segmentation based on the material composition features of each object, and perform image impurity removal processing to construct a non-plastic impurity removal optimized image;

[0090] A bounding box division module, configured to perform visual detection of plastic waste on the non-plastic impurity removal optimized image, and perform bounding box instance division processing, so as to extract the bounding box images of each plastic waste;

[0091] A detection optimization module, configured to perform dynamic component learning and iterative detection optimization on the bounding box images of each plastic waste, and construct an intelligent waste detection optimization model.

[0092] The present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the detection method of plastic waste described in any one of the above is implemented.

[0093] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the detection method of plastic waste described in any one of the above are implemented.

[0094] Those skilled in the art clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0095] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it is stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, is embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: various media for storing program codes such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0096] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0097] As described above, these are only specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A detection method for plastic waste, characterized in that It includes the following steps: Step S1: Obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and suppress the interference of regional background features, so as to construct an optimized image with background suppression; Step S2: Perform in-region target visual detection on the optimized image with background suppression, and mine the material composition of objects, so as to generate the material composition features of each object; Step S3: Perform deep semantic segmentation based on the material composition features of each object, and perform image impurity removal processing to construct an optimized image with non-plastic impurity removal; Step S4: Perform visual detection of plastic waste on the optimized image with non-plastic impurity removal, and perform bounding box instance division processing, so as to extract the bounding box images of each plastic waste; Step S5: Perform dynamic component learning and iterative detection optimization on the bounding box images of each plastic waste to construct an intelligent waste detection optimization model; Among them, the specific steps of Step S1 are: Step S11: Obtain multi-angle images of the area to be detected; Step S12: Perform multi-scale detail reconstruction on the multi-angle images of the area to be detected to construct a multi-scale detail reconstruction image; Step S13: Perform environmental background visual analysis on the multi-scale detail reconstruction image to generate the environmental background visual features in the image; Step S14: Capture multi-angle features based on the environmental background visual features in the image, so as to generate multi-angle environmental background visual features; Step S15: Suppress the interference of regional background features on the multi-scale detail reconstruction image based on the multi-angle environmental background visual features, so as to construct an optimized image with background suppression; Among them, the specific steps of Step S12 are: Perform multi-region equal-ratio segmentation on the multi-angle images of the area to be detected to obtain multiple image regions; Calculate the local contrast between regions for the multiple image regions to obtain the local contrast between multiple regions; Perform adaptive contrast enhancement according to the local contrast between multiple regions to generate a contrast-enhanced image; Analyze the brightness distribution of the contrast-enhanced image to generate image brightness distribution data; Perform non-linear brightness equalization based on the image brightness distribution data to construct a brightness equalization enhanced image; Perform discrete wavelet transform on the brightness equalization enhanced image to obtain multiple images of different scales; Calculate the frequency component features of each scale for the multiple images of different scales to generate the frequency component features of each scale; Perform corresponding scale edge detail sharpening according to the frequency component features of each scale to generate the edge detail sharpening features of each scale; Perform multi-scale feature fusion reconstruction on the multiple images of different scales based on the edge detail sharpening features of each scale to construct a multi-scale detail reconstruction image.

2. The detection method of plastic waste according to claim 1, wherein The specific steps of Step S2 are: Step S21: Perform in-region target visual detection on the optimized image with background suppression, and mark the object nodes in multiple regions; Step S22: Mine the multi-spectral features of the target nodes in multiple regions to generate the spectral features of each object; Step S23: Analyze the surface texture features of each object spectral feature to obtain the surface texture features of each object; Step S24: Perform visual gloss property analysis on the spectral features of each object, and extract the visual gloss property of each object; Step S25: Perform object material composition mining on the visual gloss property of each object and the surface texture feature of each object, so as to generate the material composition feature of each object.

3. The detection method of plastic waste according to claim 1, characterized in that, The specific steps of Step S3 are as follows: Step S31: Based on the material composition feature of each object, identify the non-plastic impurity features of the object nodes in multiple regions, and extract multiple non-plastic impurities; Step S32: Perform deep semantic segmentation on the background suppression optimized image to obtain multiple object category region images; Step S33: Perform image impurity removal processing on the multiple object category region images according to the multiple non-plastic impurities, and construct a non-plastic impurity removal optimized image.

4. The detection method of plastic waste according to claim 1, wherein The specific steps of Step S4 are as follows: Step S41: Perform visual detection of plastic waste on the non-plastic impurity removal optimized image, and mark each plastic waste node; Step S42: Perform regional precise positioning calculation on each plastic waste node to obtain the positioning coordinates of each plastic waste node; Step S43: Perform three-dimensional shape analysis on each plastic waste node based on the positioning coordinates of each plastic waste node to generate the three-dimensional shape feature of each plastic waste; Step S44: Perform bounding box instance division processing on the non-plastic impurity removal optimized image according to the three-dimensional shape feature of each plastic waste, so as to extract the bounding box image of each plastic waste.

5. The detection method of plastic waste according to claim 1, characterized in that The specific steps of Step S5 are as follows: Step S51: Perform dynamic component learning on the bounding box image of each plastic waste to generate plastic waste detection learning data; Step S52: Perform detection accuracy calculation on the plastic waste detection learning data to generate waste detection accuracy parameters; Step S53: Perform iterative detection optimization based on the waste detection accuracy parameters to construct an intelligent waste detection optimization model.

6. A detection device for plastic waste, characterized in that, Used to execute the detection method of plastic waste as described in claim 1, including: An image optimization module, configured to obtain multi-angle images of the area to be detected; perform multi-scale detail reconstruction on the multi-angle images of the area to be detected, and perform interference suppression on the regional background feature, so as to construct a background suppression optimized image; A material composition recognition module, configured to perform in-region target visual detection on the background suppression optimized image, and perform object material composition mining, so as to generate the material composition feature of each object; An impurity removal module, configured to perform deep semantic segmentation based on the material composition feature of each object, and perform image impurity removal processing to construct a non-plastic impurity removal optimized image; A bounding box division module, configured to perform visual detection of plastic waste on the non-plastic impurity removal optimized image, and perform bounding box instance division processing, so as to extract the bounding box image of each plastic waste; A detection optimization module, configured to perform dynamic component learning and iterative detection optimization on the bounding box image of each plastic waste to construct an intelligent waste detection optimization model.

7. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that, When the processor executes the computer program, it implements the steps of the detection method of plastic waste as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method for detecting plastic waste according to any one of claims 1 to 5.

Images