Research On Image-based Defect Detection Method Of Parts

The development of intelligent manufacturing visual computing,image perception and video-aware are increasingly integrated into daily life.Image technology can help people identify and screen objects,reduce errors in manual detection,realize surface flaw detection non-contact with parts,and reduce damage to parts,so it is widely used in industry.This article parts defects of non-contact detection method for the study,in view of the traditional algorithm to detect problems such as low precision,slow speed is improved,the research content is introduced from the directions of image preprocessing,image threshold segmentation,defect area recognition:Firstly,the algorithm implementation and process of image preprocessing are introduced,the analysis focuses on image size transformation,image gray processing,image de-noising,image segmentation and image morphology processing and introduced the specific implementation process of the algorithm.Secondly,the traditional adaptive median filtering algorithm is optimized to eliminate incomplete or excessive filtering of fuzzy image details.The adaptive median filtering algorithm is used to identify the noise,and fractional integral is used to construct the filtering mask to replace the median value of the traditional algorithm.The concept of image entropy is introduced to make the order of the mask operator adaptively change according to the amount of information in the image area,achieving the purpose of preserving image details and filtering out the noise in the image.Then,fractional calculus optimization firefly algorithm to improve the two-dimensional Otsu.Genetic characteristics of fractional calculus may be such that multiple thresholds fireflies,and further by using an improved algorithm to select the optimal threshold dimensional Otsu algorithm to identify the outline of the connected area to identify the small defects.Use edge detection and Hough transform to identify parts angle deviation defects and use the three color parts values of the color image to mark the red area to identify the surface coating peeling defects.Finally,experiments to verify the feasibility and accuracy of the algorithm.Experiments show that image-based parts defect detection can well identify the above-mentioned defects and can be applied to actual detection.