An intelligent identification method for plastic particles based on image analysis

Abstract

The application discloses a kind of based on image analysis's plastic particle intelligent identification method, it is related to plastic detection technical field, including the following steps: obtaining and standardizing the normal temperature particle image data of batch of plastic particles to be identified;Extract normal temperature appearance feature and generate normal temperature particle identification result;Plastic particle sample is obtained and is fused into film processing;Obtain and standardize the continuous image data of fusion film sample in the process of fusion film;The time sequence positioning of developing area and film area boundary in different film forming stages of fusion film sample is carried out, and the evolution sequence of fusion film developing is generated;Determine the evolution correlation between different developing areas, and correlation check is carried out to evolution correlation;Normal temperature particle identification result is revised, and plastic particle intelligent identification result is generated.The application utilizes normal temperature image analysis and fusion film developing check method, realizes plastic particle material identification and quality detection, with the advantages of high identification accuracy, strong abnormality finding ability.

Description

Technical Field

[0001] This invention relates to the field of plastic detection technology, and in particular to an intelligent identification method for plastic particles based on image analysis. Background Technology

[0002] Plastic granules, as an important raw material in the production of plastic products, are widely used in packaging, electronics, automotive, and building materials industries. The material type, purity, and quality of plastic granules directly affect subsequent processing performance and product quality. Existing methods for detecting plastic granules mainly employ manual visual inspection, laboratory physicochemical testing, and image analysis-based automatic identification technology. These methods acquire appearance information such as color, shape, transparency, and surface condition of the plastic granules to classify them and assess their quality.

[0003] Existing image analysis-based methods for identifying plastic particles primarily rely on their appearance at room temperature. This approach is prone to misidentification for particles with significant internal structural differences but similar appearances, such as recycled materials, mixed particles, internally contaminated particles, and particles containing fillers. Furthermore, current technologies typically lack utilization of the developmental changes during the melting process of plastic particles. This makes it difficult to cross-validate room-temperature identification results with the internal developmental characteristics after melting, leading to insufficient accuracy in material classification and an inability to effectively detect hidden impurities, internal contamination, and material mixing issues.

[0004] Therefore, how to provide an intelligent identification method for plastic particles based on image analysis is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an intelligent identification method for plastic particles based on image analysis. This invention utilizes room temperature image analysis and melt film formation and development verification methods to achieve identification and quality inspection of plastic particle materials, and has the advantages of high identification accuracy and strong anomaly detection capability.

[0006] A method for intelligent identification of plastic particles based on image analysis according to an embodiment of the present invention includes the following steps: Acquire room temperature particle image data of the batch of plastic particles to be identified and perform standardization processing to generate standardized room temperature particle image data. Based on standardized room temperature particle image data, room temperature appearance features are extracted, and room temperature particle recognition results are generated based on these features. Plastic particle samples are obtained from the batch of plastic particles to be identified, and the plastic particle samples are melt-filmed to generate melt-film samples. Continuous image data of the melt-forming sample during the melt-forming process are acquired, and the continuous image data are standardized to generate standardized melt-forming image data. Based on standardized melt-film image data, the development region and film boundary of the melt-film sample in different film-forming stages are located temporally to generate a melt-film development evolution sequence. The evolutionary relationships between different development regions are determined based on the melt-film formation and development evolution sequence, and the evolutionary relationships are verified by combining the room temperature particle identification results to generate material consistency verification results. The results of room temperature particle identification are corrected based on the material consistency verification results to generate intelligent identification results for plastic particles.

[0007] Optionally, the room temperature particle image data includes reflection images, transmission images, and locally magnified images of the batch of plastic particles to be identified at room temperature. The standardization processing includes image denoising, illumination equalization, background separation, particle region segmentation, particle scale normalization, and particle boundary enhancement processing.

[0008] Optionally, the generation of the room-temperature particle identification result specifically includes: Based on the color distribution of each plastic particle region in the standardized room temperature particle image data, the dominant color, color dispersion, and color transition continuity are extracted to generate particle color features. Based on the contour boundaries, area, length and width variations and boundary smoothness of each plastic particle region in the standardized room temperature particle image data, particle morphology parameters are extracted to generate particle morphology features. By combining the brightness distribution of each plastic particle region in the reflection and transmission images, the uniformity of light transmission, the degree of light transmission attenuation, and the degree of light transmission continuity are extracted to generate light transmission characterization features. Based on the grayscale changes, texture changes, and reflectivity changes of each plastic particle region in the magnified image, abnormal texture regions, abnormal reflectivity regions, color change regions, and visible impurity regions are identified to generate surface abnormal features. Correlation analysis is performed on particle color characteristics, particle morphology characteristics, light transmission characteristics and surface anomaly characteristics to determine the material category corresponding to each plastic particle region and generate room temperature particle identification results, which include material category identification results and material category confidence results.

[0009] Optionally, the melt film formation process includes heating and softening treatment, melt spreading treatment, tableting and stretching treatment, and cooling and shaping treatment.

[0010] Optionally, the continuous image data includes a sequence of transmission images and a sequence of reflection images acquired during the melting and film formation process of the molten film sample, and the normalization processing includes image registration, film region boundary localization, brightness correction, stage frame marking, and development region enhancement processing.

[0011] Optionally, the generation of the melt-film formation and development evolution sequence specifically includes: Based on the stage frame marker, the standardized melt film formation image data is divided into stages to obtain image data of the heating and softening stage, the melting and spreading stage, the tableting and stretching stage, and the cooling and shaping stage. Based on image data from the heating and softening stage, the bubble region, black spot region, and impurity region in the melted film sample were initially developed and located to obtain the time of bubble appearance, the location of black spot development, and the period of impurity exposure. Based on the image data of the melting and spreading stage, continuous frame tracking is performed on the bubble region, the color spot region, and the light-transmitting region to obtain the bubble migration path, the color spot diffusion direction, and the light transmission change trajectory. Based on the image data of the tablet compression and stretching stage, the spreading boundary of the molten film sample is extracted and the inter-frame position is correlated to obtain the spreading boundary change trajectory; Based on image data from the cooling and shaping stage, texture location and texture extension direction are identified in the shrinkage texture region to obtain the formation trajectory of the cooling shrinkage texture. Based on the time sequence of image frames, the timing of bubble appearance, bubble migration path, direction of color spot diffusion, location of black spot development, period of impurity exposure, trajectory of light transmission change, trajectory of spreading boundary change, and trajectory of cooling and shrinkage texture formation are sorted, connected in stages, and correlated in trajectory to obtain the melting film formation and development evolution sequence.

[0012] Optionally, the generation of the material consistency verification result specifically includes: Based on the melt-film formation evolution sequence, the appearance time, exposure period, migration path and formation trajectory of different development regions in the melt-film formation process are arranged in chronological order to obtain the formation sequence relationship between different development regions; Based on the direction of color spot diffusion, the trajectory of light transmission change and the migration path of bubbles, the positional changes of color spot area, light transmission area and bubble area in consecutive image frames are continuously compared to obtain the diffusion continuity between different development areas. Based on the temporal correspondence between the spreading boundary change trajectory and the bubble migration path, the spot diffusion direction and the impurity exposure period, the positional traction state of the spreading boundary change on the bubble area, the spot area and the impurity area is extracted to obtain the boundary traction relationship between different development areas. Based on the spatial correspondence between the cooling shrinkage texture formation trajectory and the black spot development location, impurity exposure time and light transmission change trajectory, the residual distribution state of the black spot area, impurity area and light transmission area in the cooling and setting stage is extracted to obtain the cooling residue relationship and spatial association relationship between different development areas. The relationships of formation sequence, diffusion continuity, boundary traction, cooling residue, and spatial accompaniment are correlated and verified to generate material consistency verification results.

[0013] Optionally, the generation of the intelligent identification result of the plastic particles specifically includes: Based on the material category consistency results, the material category identification results in the room temperature particle identification results are corrected to generate corrected material category identification results; Based on the abnormal material status results in the material consistency verification results, the corrected material category identification results are marked with status to generate material status identification results. Based on the quality anomalies in the material consistency verification results, the abnormal development areas in the plastic granule batch are classified into risks, and anomaly risk warning results are generated. The corrected material category identification results, material state identification results, and abnormal risk warning results are correlated to generate intelligent identification results for plastic particles.

[0014] The beneficial effects of this invention are: This invention constructs an intelligent plastic particle identification mechanism that combines room-temperature particle recognition with melt-film formation and development analysis. Based on acquiring the room-temperature appearance characteristics of plastic particles, it further acquires development evolution information during the melt-film formation process and uses the evolutionary correlation between development regions to verify the consistency of the room-temperature particle identification results. Compared to existing technologies that rely solely on appearance information such as color, shape, or transparency for identification, this invention not only utilizes the appearance characteristics at room temperature for preliminary identification but also uses the internal development characteristics of the melted plastic particles to further verify the identification results. This improves the reliability and accuracy of plastic particle material category identification results and reduces the risk of misjudgment due to the similarity in appearance between different materials.

[0015] This invention continuously tracks the generation, migration, diffusion, and residue processes of bubble regions, discolored areas, black spot regions, impurity regions, translucent regions, and shrinkage texture regions at different film-forming stages, constructing a melt-forming and developing evolution sequence. Furthermore, it establishes relationships between developing regions regarding their formation sequence, diffusion continuity, boundary traction, cooling residue, and spatial association. Analyzing the internal state of plastic particles using these evolutionary relationships enables the identification of material mixing phenomena, internal contamination phenomena, and recycled material characteristics that are difficult to detect directly at room temperature. This expands plastic particle quality inspection from a single visual assessment to internal developing behavior analysis, improving the ability to identify hidden defects and potential quality problems.

[0016] This invention achieves the collaborative generation of material category identification results, material state identification results, and abnormal risk warning results by feeding back material consistency verification results to room-temperature particle identification results. Utilizing the development correlation features during the melt film formation process to dynamically correct the room-temperature identification results, it can promptly detect inconsistencies between room-temperature appearance characteristics and melt development characteristics, improving the consistency between the final identification result and the actual material state of the plastic particles. Simultaneously, this invention can output abnormal risk warning results such as internal contamination risk, impurity residue risk, abnormal light transmission risk, and abnormal cooling shrinkage risk, providing a more comprehensive and accurate quality evaluation basis for plastic particle production, sorting, quality inspection, and subsequent processing. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an image analysis-based intelligent identification method for plastic particles proposed in this invention; Figure 2 This is a flowchart illustrating the generation of the melt-film formation and development evolution sequence of an image-based intelligent identification method for plastic particles proposed in this invention. Figure 3 This is a flowchart illustrating the process of generating material consistency verification results for an image-based intelligent identification method for plastic particles proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-3 A method for intelligent identification of plastic particles based on image analysis includes the following steps: Acquire room temperature particle image data of the batch of plastic particles to be identified and perform standardization processing to generate standardized room temperature particle image data. Based on standardized room temperature particle image data, room temperature appearance features are extracted, and room temperature particle recognition results are generated based on these features. Plastic particle samples are obtained from the batch of plastic particles to be identified, and the plastic particle samples are melt-filmed to generate melt-film samples. Continuous image data of the melt-forming sample during the melt-forming process are acquired, and the continuous image data are standardized to generate standardized melt-forming image data. Based on standardized melt-film image data, the development region and film boundary of the melt-film sample in different film-forming stages are located temporally to generate a melt-film development evolution sequence. The evolutionary relationships between different development regions are determined based on the melt-film formation and development evolution sequence, and the evolutionary relationships are verified by combining the room temperature particle identification results to generate material consistency verification results. The results of room temperature particle identification are corrected based on the material consistency verification results to generate intelligent identification results for plastic particles.

[0020] In this embodiment, the room temperature particle image data includes the reflected image, transmitted image and locally magnified image of the batch of plastic particles to be identified under room temperature conditions. The standardization processing includes image denoising, illumination equalization, background separation, particle region segmentation, particle scale normalization and particle boundary enhancement processing.

[0021] In this embodiment, the generation of room-temperature particle identification results specifically includes: Based on the color distribution of each plastic particle region in the standardized room temperature particle image data, the dominant color, color dispersion, and color transition continuity are extracted to generate particle color features. The color dispersion is the average color distance between all pixel color values ​​and the dominant color value in the plastic particle region. The color transition continuity is the variance of the color difference between all adjacent pixels in the plastic particle region. The dominant color value is the pixel color value that appears most frequently in the plastic particle region. Based on the outline boundary, area, length and width variation and boundary smoothness of each plastic particle region in the standardized room temperature particle image data, particle morphology parameters are extracted to generate particle morphology features; the boundary smoothness is the average value of the turning angle of adjacent boundary line segments on the boundary outline of the plastic particle region. By combining the brightness distribution of each plastic particle region in the reflection and transmission images, the uniformity of light transmission, the degree of light transmission attenuation, and the degree of light transmission continuity are extracted to generate light transmission characterization features. The uniformity of light transmission is the variance of the transmitted brightness value of each pixel in the plastic particle region; the degree of light transmission attenuation is the difference between the average transmitted brightness value of the plastic particle region and the average brightness value of the corresponding background region; and the degree of light transmission continuity is the variance of the difference in transmitted brightness values ​​between adjacent pixels in the plastic particle region. Based on the grayscale, texture, and reflectivity changes of each plastic particle region in the magnified image, abnormal texture regions, abnormal reflectivity regions, color abrupt change regions, and visible impurity regions are identified to generate surface anomaly features. Abnormal texture regions are areas in the magnified image where the grayscale co-occurrence matrix contrast is higher than the average grayscale co-occurrence matrix contrast of the plastic particle region. Abnormal reflectivity regions are areas in the magnified image where the pixel brightness value is higher than the average brightness value of the plastic particle region. Color abrupt change regions are areas in the magnified image where the color difference between adjacent pixels is higher than the average color difference between adjacent pixels in the plastic particle region. Visible impurity regions are the overlapping areas of abnormal texture regions, abnormal reflectivity regions, and color abrupt change regions. Correlation analysis was performed on particle color characteristics, particle morphology characteristics, light transmittance characteristics, and surface anomaly characteristics to determine the material category corresponding to each plastic particle region and generate room temperature particle identification results. The room temperature particle identification results include material category identification results and material category confidence results. Specifically, the differences in color characteristics, morphology characteristics, light transmittance characteristics, and surface anomaly characteristics of each plastic particle region were calculated. The differences in color characteristics, morphology characteristics, light transmittance characteristics, and anomaly characteristics were added together to obtain the comprehensive difference value between regions. The plastic particle regions were merged into clusters according to the comprehensive difference value in ascending order to form cluster categories. The cluster category with the most clusters was taken as the material category identification result. The material category confidence result is the proportion of the number of plastic particle regions corresponding to the material category identification result to the total number of plastic particle regions in the batch of plastic particles to be identified.

[0022] In this embodiment, the melt-film forming process includes heating and softening treatment, melt spreading treatment, sheeting and stretching treatment, and cooling and shaping treatment. Heating and softening treatment refers to applying heat to the plastic particle sample, causing the plastic particle sample to change from a solid particle state to a softened state. Melt spreading treatment refers to continuing to heat the plastic particle sample after it has softened, causing the plastic particle sample to melt and spread along the bearing surface to form a molten layer. Sheeting and stretching treatment refers to applying a sheeting action to the molten layer, reducing the thickness of the molten layer and extending it circumferentially to form a film structure. Cooling and shaping treatment refers to cooling and solidifying the film structure, causing the film structure to form a melt-film sample.

[0023] In this embodiment, the continuous image data includes a sequence of transmission images and a sequence of reflection images of the molten film sample acquired during the molten film formation process. The normalization processing includes image registration, film area boundary localization, brightness correction, stage frame marking, and development area enhancement processing. The development area enhancement processing includes local contrast enhancement, edge sharpening, color difference enhancement, and texture detail enhancement of the transmission image sequence and reflection image sequence of the molten film sample to enhance the image difference between the light-transmitting area, bubble area, color spot area, black spot area, impurity area, and shrinkage texture area and the film area background.

[0024] In this embodiment, the generation of the melt-film formation and development evolution sequence specifically includes: Based on the stage frame marker, the standardized melt film formation image data is divided into stages to obtain image data of the heating and softening stage, the melting and spreading stage, the tableting and stretching stage, and the cooling and shaping stage. Based on image data from the heating and softening stage, initial development and localization of bubble regions, black spot regions, and impurity regions in the molten film sample are performed to obtain the bubble appearance time, black spot development location, and impurity exposure period. Specifically, the film region in the image data of the heating and softening stage is extracted frame by frame, and the transmission brightness value distribution, color value distribution, and texture distribution of the film region are calculated. Connected regions with transmission brightness values ​​lower than the average transmission brightness value of the film region are identified as bubble regions, connected regions with color values ​​deviating from the dominant color value of the film region are identified as black spot regions, and connected regions with color values ​​deviating from the dominant color value of the film region and texture distribution different from the surrounding film regions are identified as impurity regions. The overlapping area and center distance between similar developed regions in adjacent image frames are calculated, and the developed regions with the largest overlapping area and the smallest center distance are correlated accordingly. The image frame position corresponding to the first appearance of the bubble region is identified as the bubble appearance time, the spatial coordinates of the black spot region in the film region are identified as the black spot development location, and the image frame interval corresponding to the first and last appearance of the impurity region is identified as the impurity exposure period. Based on the image data of the melting and spreading stage, continuous frame tracking of bubble regions, color spot regions, and translucent regions is performed to obtain the bubble migration path, color spot diffusion direction, and translucent change trajectory. Specifically, the bubble regions, color spot regions, and translucent regions corresponding to each image frame in the melting and spreading stage image data are extracted, and the spatial position of each region in the film region is recorded. For bubble regions in adjacent image frames, the overlapping area, center distance, and area difference are calculated, and the bubble region with the largest overlapping area, smallest center distance, and smallest area difference is taken as the corresponding region for consecutive frames of the same bubble region. The bubble migration path is formed by connecting the bubble regions according to the spatial position change order of the same bubble region in consecutive image frames. The boundary expansion position of the color spot region in consecutive image frames is counted, and the direction with the largest expansion distance of the color spot region is taken as the color spot diffusion direction. The distribution change of the translucent region's transmission brightness value in consecutive image frames is calculated, and the translucent change trajectory is formed by connecting the corresponding time order of the transmission brightness value distribution. Based on image data from the tablet compression and stretching stage, the spreading boundary of the molten film sample is extracted and its position is correlated with the frame position to obtain the spreading boundary change trajectory. Specifically, the following steps are taken: extract the film region boundary contour corresponding to each image frame in the tablet compression and stretching stage image data, and record the spatial coordinates of each boundary point on the film region boundary contour; select the minimum, maximum, minimum, and maximum values ​​of the horizontal coordinates in each image frame to form the outer boundary range of the film region; calculate the difference in the horizontal and vertical coordinates of the outer boundary range of the film region in adjacent image frames to obtain the boundary expansion distance between adjacent image frames; obtain the boundary expansion direction between adjacent image frames based on the positive and negative directions corresponding to the difference in the horizontal and vertical coordinates; and arrange the boundary expansion distance and boundary expansion direction according to the time sequence of the image frames to generate the spreading boundary change trajectory. Based on image data from the cooling and shaping stage, texture location identification and texture extension direction tracking are performed on the shrinking texture region to obtain the cooling shrinking texture formation trajectory. Specifically, the shrinking texture region corresponding to each image frame in the cooling and shaping stage image data is extracted, and the spatial coordinates of each texture line segment in the shrinking texture region are recorded; the length, direction angle, and center position of each texture line segment are calculated to obtain the texture position data corresponding to each image frame; the center position distance, length difference, and direction angle difference of texture line segments in adjacent image frames are calculated, and the texture line segment with the smallest center position distance, smallest length difference, and smallest direction angle difference is taken as the corresponding line segment of consecutive frames in the same shrinking texture region; the texture movement path is formed by connecting the corresponding line segments of consecutive frames in the same shrinking texture region according to the displacement order of the center position of the corresponding line segments, and the texture extension path is formed by connecting them according to the change order of the direction angle of the corresponding line segments of consecutive frames in the same shrinking texture region; the texture movement path and the texture extension path are temporally correlated to generate the cooling shrinking texture formation trajectory. Based on the time sequence of image frames, the timing of bubble appearance, bubble migration path, direction of color spot diffusion, location of black spot development, period of impurity exposure, trajectory of light transmission change, trajectory of spreading boundary change, and trajectory of cooling and shrinkage texture formation are sorted, connected in stages, and correlated in trajectory to obtain the melting film formation and development evolution sequence.

[0025] In this embodiment, the generation of material consistency verification results specifically includes: Based on the melt-film formation evolution sequence, the appearance time, exposure period, migration path and formation trajectory of different development regions in the melt-film formation process are arranged in chronological order to obtain the formation sequence relationship between different development regions; Based on the diffusion direction of color spots, the trajectory of light transmission changes, and the migration path of bubbles, the positional changes of color spot regions, light-transmitting regions, and bubble regions in consecutive image frames are continuously compared to obtain the diffusion continuity between different developing regions. Specifically, the diffusion endpoint positions of color spot regions, the brightness change center positions of light-transmitting regions, and the migration center positions of bubble regions in consecutive image frames are extracted; the positional differences of the diffusion endpoint positions of color spot regions, the brightness change center positions of light-transmitting regions, and the migration center positions of bubble regions in adjacent image frames are calculated respectively; the positional differences are arranged in the time sequence of image frames to generate a sequence of positional changes of color spot regions, light-transmitting regions, and bubble regions; the direction and magnitude of the changes in adjacent positional differences in the positional change sequence are compared to generate the diffusion continuity between different developing regions. Based on the temporal correspondence between the spreading boundary change trajectory and the bubble migration path, the spot diffusion direction, and the impurity exposure period, the positional traction state of the spreading boundary change on the bubble region, spot region, and impurity region is extracted to obtain the boundary traction relationship between different developing regions. Specifically, the spreading boundary change trajectory, bubble migration path, spot diffusion direction, and impurity exposure period are time-aligned according to the image frame time sequence; the boundary expansion direction and boundary expansion distance of the spreading boundary in the same image frame are extracted, and the offset direction and offset distance of the bubble region, the diffusion direction and diffusion distance of the spot region, and the exposure position change of the impurity region are extracted respectively; the boundary expansion direction and boundary expansion distance are respectively mapped to the offset direction and offset distance of the bubble region to obtain the offset state of the bubble region moving with the spreading boundary; the boundary expansion direction and boundary expansion distance are respectively mapped to the diffusion direction and diffusion distance of the spot region to obtain the diffusion state of the spot region extending with the spreading boundary; the boundary expansion direction and boundary expansion distance are respectively mapped to the exposure position change of the impurity region to obtain the exposure state of the impurity region caused by the spreading boundary change; the boundary traction relationship between different developing regions is generated based on the offset state, diffusion state, and exposure state. Based on the spatial correspondence between the formation trajectory of cooling shrinkage texture and the development position of black spots, the exposure period of impurities, and the trajectory of light transmission changes, the residual distribution state of black spot areas, impurity areas, and light-transmitting areas in the cooling and setting stage is extracted to obtain the cooling residue relationship and spatial association relationship between different development areas. Specifically, the shrinkage texture area, black spot area, impurity area, and light-transmitting area in the last frame image of the cooling and setting stage are extracted, and the spatial coordinate range corresponding to each area is recorded; the development position of black spots is mapped to the spatial coordinate range corresponding to the last frame image to obtain the residual position of black spot areas; the spatial coordinate range of the impurity area corresponding to the impurity exposure period in the last frame image is taken as the impurity area. Residual position; the spatial coordinate range corresponding to the termination position of the light transmission change trajectory is taken as the residual position of the light transmission area; calculate the spatial distance between the residual positions of the black spot area, the impurity area, and the light transmission area and the shrinkage texture area to generate the residual distribution state of the black spot area, the impurity area, and the light transmission area; statistically analyze the spatial distance between the residual positions of the black spot area, the impurity area, and the light transmission area and the shrinkage texture area to generate the cooling residual relationship between different development areas; count the number of times the black spot area, the impurity area, the light transmission area, and the shrinkage texture area are located in the same film area or adjacent film area to generate the spatial association relationship between different development areas; Cooling residue relationship is used to characterize the retention state of the developed area in the film structure after the molten film sample has cooled and solidified; spatial association relationship is used to characterize the co-occurrence of black spot area, impurity area, light-transmitting area and shrinkage texture area in the same film area or adjacent film area locations. The formation sequence, diffusion continuity, boundary traction, cooling residue, and spatial association relationships are correlated and verified to generate material consistency verification results. Specifically, the frequency of sequential occurrence between different developing regions is counted based on the formation sequence; the frequency of continuous diffusion between different developing regions is counted based on diffusion continuity; the frequency of positional traction between different developing regions is counted based on boundary traction; the frequency of residual co-occurrence between different developing regions is counted based on cooling residue; and the frequency of spatial association between different developing regions is counted based on spatial association. The frequencies of sequential occurrence, continuous diffusion, positional traction, residual co-occurrence, and spatial association for corresponding developing region combinations are added together, and the sum is calculated. The proportion of the total frequency of all developing region combinations is used to obtain the correlation frequency corresponding to different developing region combinations. A developing region correlation distribution is constructed based on the correlation frequency of different developing region combinations and their spatial distribution within the film region. The developing region combination with the highest correlation frequency in the developing region correlation distribution is selected, and the average distance between each spatial distribution position in the developing region combination with the highest correlation frequency is calculated to obtain the spatial distribution concentration. The material category identification result in the room temperature particle identification result is correlated with the developing region combination with the highest correlation frequency, and the material category confidence result in the room temperature particle identification result is correlated with the spatial distribution concentration to generate a material consistency verification result.

[0026] In this embodiment, the generation of the intelligent identification result of plastic particles specifically includes: The material category identification results in the room temperature particle identification results are corrected based on the material category consistency results to generate corrected material category identification results. Specifically, the category consistency identifier is extracted from the material category consistency results, and the material category identification results in the room temperature particle identification results are also extracted. When the category consistency identifier corresponds to a material category identification result, the material category identification result is used as the corrected material category identification result. When the category consistency identifier does not correspond to a material category identification result, the material category identification result is replaced with the corrected category from the material category consistency results to generate corrected material category identification results. Based on the abnormal material status results in the material consistency verification results, the corrected material category identification results are marked with status to generate material status identification results. Specifically, the abnormal status categories in the abnormal material status results are extracted, and the abnormal status categories are mapped to the corrected material category identification results. When the abnormal status category corresponds to a single material development-related feature, the corrected material category identification results are marked as material consistent. When the abnormal status category corresponds to multiple development-related features existing simultaneously, the corrected material category identification results are marked as material mixed. When the abnormal status category corresponds to an abnormal recycled material status, the corrected material category identification results are marked as suspected recycled material status. When the abnormal status category is material status abnormal, the corrected material category identification results are marked as material status abnormal. The material status identification results are generated based on the status marking results. Based on the quality anomaly results in the material consistency verification, the abnormal development areas in the plastic granule batch are classified into risks, and anomaly risk warning results are generated. Specifically, the abnormal development areas and anomaly types corresponding to the quality anomaly results are extracted; abnormal development areas where black spot areas and impurity areas appear together are classified as internal contamination risk; abnormal development areas with residual impurity areas are classified as impurity residue risk; abnormal development areas with abnormal changes in light transmittance areas are classified as light transmittance anomaly risk; abnormal development areas with abnormal distribution of shrinkage texture areas are classified as cooling shrinkage anomaly risk; and anomaly risk warning results are generated based on the risk classification results. The corrected material category identification results, material state identification results, and abnormal risk warning results are correlated to generate intelligent identification results for plastic particles.

[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to the quality inspection scenario of plastic granules in a plastic modified material manufacturing company. This company has long purchased plastic granules from various sources as raw materials, and different batches of plastic granules exhibit similar colors, transparency, and morphology. Although some plastic granules show similar appearance characteristics at room temperature, they may contain issues such as blending of recycled materials, material mixing, residual impurities, and internal contamination. Using conventional visual inspection methods can easily lead to inaccurate material classification and difficulty in detecting hidden quality anomalies, thus affecting subsequent product quality control.

[0028] In the application process, firstly, reflective, transmissive, and magnified images of the batch of plastic particles to be identified are acquired. These images are then standardized, and particle color, morphological, translucent characteristics, and surface anomaly features are extracted to generate room-temperature particle identification results. Subsequently, plastic particle samples from the same batch are selected for melt-film formation. During the heating, softening, melting, spreading, pressing, and cooling processes, transmissive and reflective image sequences are continuously acquired. Bubble areas, discolored areas, black spot areas, impurity areas, translucent areas, and shrinkage texture areas are identified and tracked, generating a melt-film development evolution sequence. Further analysis is conducted on the formation sequence, diffusion continuity, boundary traction, cooling residue, and spatial association relationships among different development areas. This analysis is combined with the room-temperature particle identification results for correlation verification, generating material consistency verification results. Finally, the room-temperature particle identification results are corrected based on the material consistency verification results, and the intelligent plastic particle identification results are output.

[0029] Application results show that this invention can detect plastic particles with highly similar appearance characteristics at room temperature but significant differences in internal development behavior. It can identify material mixing states, suspected recycled material states, and internal contamination risks—problems that are difficult to detect with conventional appearance inspections. By jointly analyzing the appearance identification results at room temperature with the development evolution characteristics during the melt film formation process, the consistency between material category identification results and the actual material state is improved. This enhances the ability to identify hidden quality anomalies and the reliability of plastic particle quality detection, providing more accurate technical support for raw material screening and quality control in the production process of modified plastic materials.

[0030] Table 1. Performance Comparison of the Invention and Traditional Plastic Particle Identification Methods

[0031] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0032] Regarding the accuracy of material category identification, the traditional room-temperature image recognition method achieves 92.3%, while this invention reaches 96.8%, an improvement of 4.5 percentage points. This improvement is due to the fact that this invention not only utilizes the particle color, morphology, and light transmission characteristics at room temperature for identification, but also introduces the development evolution characteristics during the melting and film formation process for cross-validation. This further distinguishes plastic particles with similar appearances but different internal structures, thereby improving the accuracy of material category identification.

[0033] In terms of the accuracy of identifying recycled materials and mixed materials, traditional methods achieve 88.7% and 86.9% respectively, while this invention achieves 93.5% and 92.1% respectively. This is because recycled and mixed materials often have similar colors and appearances to virgin materials at room temperature. This invention, by analyzing the evolution of bubble regions, discolored areas, impurity regions, and shrinkage texture regions during the melting and film formation process, can discover the differences in the development of different materials during the heating and melting process, thus providing a higher ability to identify the state of recycled and mixed materials.

[0034] Regarding the detection rates of internal contamination and impurity anomalies, traditional methods achieve 81.5% and 84.2% respectively, while this invention improves these rates to 89.4% and 90.3%. This is because internal contaminants and some minute impurities are easily obscured by the surface structure of plastic particles in images at room temperature. This invention, however, uses a melt-film forming process to gradually reveal the abnormal structures hidden inside the particles, and analyzes the evolution of black spot areas, impurity areas, and transparent areas, thus enhancing the ability to detect internal contamination and impurity anomalies.

[0035] Regarding the false negative rate and false positive rate, the traditional method had rates of 8.1% and 6.8%, respectively, while this invention reduced them to 4.9% and 4.2%, respectively. Simultaneously, the testing time per batch increased from 12.4 minutes to 13.6 minutes. The slight increase in testing time is due to the addition of melt-film formation and development evolution analysis processes in this invention. However, the accuracy rates for material category identification, recycled material identification, mixed material identification, and internal contamination detection are all significantly improved, indicating that the increased testing time results in higher identification reliability and quality detection capabilities, demonstrating significant engineering application value.

[0036] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent identification of plastic particles based on image analysis, characterized in that, Includes the following steps: Acquire room temperature particle image data of the batch of plastic particles to be identified and perform standardization processing to generate standardized room temperature particle image data. Based on standardized room temperature particle image data, room temperature appearance features are extracted, and room temperature particle recognition results are generated based on these features. Plastic particle samples are obtained from the batch of plastic particles to be identified, and the plastic particle samples are melt-filmed to generate melt-film samples. Continuous image data of the melt-forming sample during the melt-forming process are acquired, and the continuous image data are standardized to generate standardized melt-forming image data. Based on standardized melt-film image data, the development region and film boundary of the melt-film sample in different film-forming stages are located temporally to generate a melt-film development evolution sequence. The evolutionary relationships between different development regions are determined based on the melt-film formation and development evolution sequence, and the evolutionary relationships are verified by combining the room temperature particle identification results to generate material consistency verification results. The results of room temperature particle identification are corrected based on the material consistency verification results to generate intelligent identification results for plastic particles.

2. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The ambient temperature particle image data includes reflection images, transmission images, and magnified images of the batch of plastic particles to be identified at ambient temperature. The standardization processing includes image denoising, illumination equalization, background separation, particle region segmentation, particle scale normalization, and particle boundary enhancement processing.

3. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The generation of the room-temperature particle identification result specifically includes: Based on the color distribution of each plastic particle region in the standardized room temperature particle image data, the dominant color, color dispersion, and color transition continuity are extracted to generate particle color features. Based on the contour boundaries, area, length and width variations and boundary smoothness of each plastic particle region in the standardized room temperature particle image data, particle morphology parameters are extracted to generate particle morphology features. By combining the brightness distribution of each plastic particle region in the reflection and transmission images, the uniformity of light transmission, the degree of light transmission attenuation, and the degree of light transmission continuity are extracted to generate light transmission characterization features. Based on the grayscale changes, texture changes, and reflectivity changes of each plastic particle region in the magnified image, abnormal texture regions, abnormal reflectivity regions, color change regions, and visible impurity regions are identified to generate surface abnormal features. Correlation analysis is performed on particle color characteristics, particle morphology characteristics, light transmission characteristics and surface anomaly characteristics to determine the material category corresponding to each plastic particle region and generate room temperature particle identification results, which include material category identification results and material category confidence results.

4. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The melt film formation process includes heating and softening treatment, melt spreading treatment, tableting and stretching treatment, and cooling and shaping treatment.

5. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The continuous image data includes a sequence of transmission images and a sequence of reflection images acquired during the melting and film formation process of the molten film sample. The normalization processing includes image registration, film region boundary localization, brightness correction, stage frame marking, and development region enhancement processing.

6. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The generation of the melt-film formation and development evolution sequence specifically includes: Based on the stage frame marker, the standardized melt film formation image data is divided into stages to obtain image data of the heating and softening stage, the melting and spreading stage, the tableting and stretching stage, and the cooling and shaping stage. Based on image data from the heating and softening stage, the bubble region, black spot region, and impurity region in the melted film sample were initially developed and located to obtain the time of bubble appearance, the location of black spot development, and the period of impurity exposure. Based on the image data of the melting and spreading stage, continuous frame tracking is performed on the bubble region, the color spot region, and the light-transmitting region to obtain the bubble migration path, the color spot diffusion direction, and the light transmission change trajectory. Based on the image data of the tablet compression and stretching stage, the spreading boundary of the molten film sample is extracted and the inter-frame position is correlated to obtain the spreading boundary change trajectory; Based on image data from the cooling and shaping stage, texture location and texture extension direction are identified in the shrinkage texture region to obtain the formation trajectory of the cooling shrinkage texture. Based on the time sequence of image frames, the timing of bubble appearance, bubble migration path, direction of color spot diffusion, location of black spot development, period of impurity exposure, trajectory of light transmission change, trajectory of spreading boundary change, and trajectory of cooling and shrinkage texture formation are sorted, connected in stages, and correlated in trajectory to obtain the melting film formation and development evolution sequence.

7. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The generation of the material consistency verification result specifically includes: Based on the melt-film formation evolution sequence, the appearance time, exposure period, migration path and formation trajectory of different development regions in the melt-film formation process are arranged in chronological order to obtain the formation sequence relationship between different development regions; Based on the direction of color spot diffusion, the trajectory of light transmission change and the migration path of bubbles, the positional changes of color spot area, light transmission area and bubble area in consecutive image frames are continuously compared to obtain the diffusion continuity between different development areas. Based on the temporal correspondence between the spreading boundary change trajectory and the bubble migration path, the spot diffusion direction and the impurity exposure period, the positional traction state of the spreading boundary change on the bubble area, the spot area and the impurity area is extracted to obtain the boundary traction relationship between different development areas. Based on the spatial correspondence between the cooling shrinkage texture formation trajectory and the black spot development location, impurity exposure time and light transmission change trajectory, the residual distribution state of the black spot area, impurity area and light transmission area in the cooling and setting stage is extracted to obtain the cooling residue relationship and spatial association relationship between different development areas. The relationships of formation sequence, diffusion continuity, boundary traction, cooling residue, and spatial accompaniment are correlated and verified to generate material consistency verification results.

8. The intelligent identification method for plastic particles based on image analysis according to claim 1, characterized in that, The generation of the intelligent identification result for the plastic particles specifically includes: Based on the material category consistency results, the material category identification results in the room temperature particle identification results are corrected to generate corrected material category identification results; Based on the abnormal material status results in the material consistency verification results, the corrected material category identification results are marked with status to generate material status identification results. Based on the quality anomalies in the material consistency verification results, the abnormal development areas in the plastic granule batch are classified into risks, and anomaly risk warning results are generated. The corrected material category identification results, material state identification results, and abnormal risk warning results are correlated to generate intelligent identification results for plastic particles.

Images