CN-121981950-A - Plastic master batch impurity identification method and system based on visual detection

CN121981950ACN 121981950 ACN121981950 ACN 121981950ACN-121981950-A

Abstract

The invention provides a plastic master batch impurity identification method and system based on visual detection, which are applied to the technical field of industrial visual detection; the method comprises the steps of collecting an image sequence through multi-angle illumination, processing the sequence, tracking impurity displacement, constructing a definition map, inputting a pre-training convolutional neural network to calculate a relative depth value, identifying the impurity type according to comparison of the depth value and a threshold value, and determining three-dimensional coordinates.

Inventors

ZHANG XIANJIANG
YOU KUANGZHENG
ZHAO ZHENGHONG

Assignees

深圳市华万彩实业有限公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (10)

1. The plastic master batch impurity identification method based on visual detection is characterized by comprising the following steps of: Controlling a multi-angle light source system, sequentially irradiating master batch samples from a plurality of different incident angles, and synchronously collecting and recording a first image sequence of the optical response change of the master batch samples; Processing the first image sequence to track the displacement change of the impurity region in the first image sequence in the sequence, so as to obtain an impurity displacement vector field; Extracting edge sharpness characteristics of the impurity region and fuzzy gradient distribution characteristics of the impurity region according to the impurity displacement vector field to construct a second definition map representing imaging definition states of the impurity region; inputting the second definition map into a pre-trained convolutional neural network model for processing to calculate a relative depth value, wherein the relative depth value is a depth value of the impurity region relative to the surface of the master batch sample; and identifying the impurity region as one of a surface layer type and a deep layer type according to a comparison result of the relative depth value and a preset depth threshold value, and determining three-dimensional space coordinates of the impurity region.
2. The method of claim 1, wherein said controlling a multi-angle light source system comprises: one or a group of a plurality of independent LED light source units which are annularly arranged are sequentially lightened so as to change the incident illumination angle of the master batch sample.
3. The method of claim 1, further comprising, prior to said processing said first sequence of images: And carrying out noise reduction processing on each frame of image in the first image sequence by adopting a Gaussian filter.
4. The method of claim 1, wherein the processing the first sequence of images to track the change in displacement of the impurity region in the sequence is performed using a Lucas-Kanade optical flow algorithm.
5. The method of claim 1, wherein the extracting edge sharpness features of the impurity region comprises: And carrying out convolution operation on the impurity region by adopting a Sobel operator to calculate the pixel gradient amplitude of the impurity region, and taking the pixel gradient amplitude as the edge sharpness characteristic.
6. The method of claim 1, wherein the determining three-dimensional spatial coordinates of the impurity region comprises: and converting the two-dimensional pixel coordinates of the impurity region in the image into the three-dimensional space coordinates under a world coordinate system by combining the camera internal reference matrix and the external reference matrix.
7. The method according to claim 1, further comprising, after said identifying the impurity region as one of the surface layer type and the deep layer type: If the impurity region is identified as the surface layer type, acquiring the three-dimensional space coordinates of the impurity region, and performing DBSCAN density cluster analysis on point cloud data formed by the three-dimensional space coordinates containing a plurality of surface layer type impurities to identify group boundary contours; and planning and generating a robot cleaning path capable of covering the inner area of the group boundary contour based on the group boundary contour and the preset action range parameters of the cleaning tool.
8. The method according to claim 1, further comprising, after said identifying the impurity region as one of the surface layer type and the deep layer type: If the impurity region is identified as the deep layer type, acquiring the three-dimensional space coordinates of the impurity region, and calculating the Euclidean distance between the three-dimensional space coordinates and the geometric central axis of the master batch sample to obtain an embedding depth deviation; and inputting the embedded depth deviation into a preset material mechanics risk assessment model to output a quantization index representing the structural failure risk of the master batch sample caused by the existence of the deep type impurities.
9. The method of claim 1, further comprising, after said deriving said impurity displacement vector field: and smoothing the impurity displacement vector field by adopting a Kalman filtering algorithm to eliminate abrupt noise points in the impurity displacement vector field.
10. A plastic master batch impurity identification system based on visual detection is characterized by comprising: The image acquisition module is configured to synchronously acquire a first image sequence of the master batch sample under the condition of being irradiated by the multi-angle light source system in a time sequence manner; the displacement tracking module is configured to process the first image sequence to track the displacement change of the impurity region in the first image sequence in the sequence so as to obtain an impurity displacement vector field; The characteristic construction module is configured to extract edge sharpness characteristics of the impurity region and fuzzy gradient distribution characteristics of the impurity region according to the impurity displacement vector field so as to construct a second definition map representing imaging definition states of the impurity region; the depth calculation module is configured to call a pre-trained convolutional neural network model to process the second definition map so as to calculate a relative depth value, wherein the relative depth value is a depth value of the impurity region relative to the surface of the master batch sample; And the identification and positioning module is configured to identify the impurity region as one of a surface layer type and a deep layer type according to a comparison result of the relative depth value and a preset depth threshold value, and determine three-dimensional space coordinates of the impurity region.

Description

Plastic master batch impurity identification method and system based on visual detection Technical Field The invention relates to the technical field of image processing, in particular to a plastic master batch impurity identification method and system based on visual detection. Background The purity of the plastic master batch as a raw material for producing plastic products directly affects the performance and quality of the final product. During the production, transportation and storage of the master batch, impurities such as metal scraps, organic particles and the like are often mixed. These impurities not only affect the appearance of the product, but may also reduce mechanical properties and even constitute a safety hazard in certain applications. Therefore, the plastic master batch is subjected to impurity detection. The current mainstream impurity detection method relies on two-dimensional surface visual detection technology. The technology shoots a sample surface image by an industrial camera, and identifies a color or gray scale abnormal region as an impurity by using an image processing algorithm. However, the plastic master batch generally has a certain transparent or semitransparent property, resulting in impurities being distributed at different depths. The traditional two-dimensional detection technology cannot acquire depth information of impurities, and is difficult to distinguish deep embedded impurities with different influences on product quality from surface layer attached impurities. In addition, because light can scatter and attenuate when penetrating the master batch matrix, impurities with different burial depths exhibit different sharpness characteristics on the two-dimensional image. Shallow impurity edges are sharp, and deep impurity edges are blurred. The existing detection system lacks the capability of carrying out depth analysis and quantitative analysis on the definition difference caused by depth, and cannot establish the accurate corresponding relation between the imaging characteristics of the impurities and the three-dimensional space positions of the imaging characteristics of the impurities. This lack of depth information directly affects the scientificity and economics of subsequent quality assessment and process treatment decisions. For example, misjudging surface impurities that could be removed by simple cleaning as deep impurities would result in unnecessary whole batch waste disposal. It should be noted that the information disclosed in the foregoing background section is only for enhancement of understanding of the background of the invention and thus may include information that does not form the prior art that is already known to those of ordinary skill in the art. Disclosure of Invention In view of the above, the invention provides a plastic master batch impurity identification method and system based on visual detection, which aim to solve the technical problem that the surface layer and deep layer impurities cannot be accurately distinguished due to the fact that depth information of impurities in a semitransparent object cannot be obtained in the prior art, and accurately calculate specific position coordinates of the impurities in a plastic master batch three-dimensional space by analyzing definition change rules of the impurities under specific optical imaging conditions, so that accurate classification and positioning of the surface layer and the deep layer impurities are realized, and reliable data support is provided for subsequent quality control and process decision. The embodiment of the invention provides a plastic master batch impurity identification method based on visual detection, which comprises the following steps: Controlling a multi-angle light source system, sequentially irradiating the master batch sample from a plurality of different incident angles, and synchronously collecting a first image sequence recorded with the optical response change of the master batch sample; Processing the first image sequence to track the displacement change of the impurity region in the first image sequence in the sequence, thereby obtaining an impurity displacement vector field; extracting edge sharpness characteristics of the impurity region and fuzzy gradient distribution characteristics of the impurity region according to the impurity displacement vector field to construct a second definition map representing imaging definition states of the impurity region; Inputting the second definition map into a pre-trained convolutional neural network model for processing to calculate a relative depth value, wherein the relative depth value is a depth value of an impurity region relative to the surface of a master batch sample; And identifying the impurity region as one of a surface layer type and a deep layer type according to a comparison result of the relative depth value and a preset depth threshold value, and determining three-dimensional space coordinates