Plastic cup appearance multi-defect concurrent recognition and classification system based on industrial vision

Abstract

The application provides a plastic cup appearance multi-defect concurrent identification and classification system based on industrial vision, relates to the technical field of industrial vision systems, and comprises an image acquisition preprocessing module, a multi-defect concurrent identification and classification module, an edge computing inference acceleration module, a defect feature decoupling and reorganization module, and a self-adaptive adaptation and real-time feedback linkage module; the application can identify the specifications and the placement posture of the plastic cup in real time through a self-adaptive control unit, automatically adjust the camera acquisition angle, the light source brightness and the acquisition frame rate, and avoid shooting dead angles and light reflection interference; a preprocessing unit integrates a three-step self-adaptive algorithm to realize the integration of acquisition and preprocessing linkage, wherein the improved median filter algorithm can accurately remove image noise, the algorithm combining the gray scale stretching and the histogram equalization can efficiently eliminate the light reflection area interference, and the edge enhancement algorithm can highlight the tiny defect features, thereby providing high-quality and high-consistency image data for subsequent defect identification.

Images