Research On Image Segmentation And Separation Of Mechanical Parts

Machine vision is an extremely important part of modern manufacturing,involving interdisciplinary disciplines such as artificial intelligence,neurobiology,psychophysics,computer science,image processing and pattern recognition.Machine vision detection and intelligent control of mechanical parts in the manufacturing process has become a research hotspot in the field of modern manufacturing,such as automatic line production and assembly monitoring,robot and robot guidance,visual servo system,automatic understanding and identification of images of mechanical parts.In order to effectively obtain the information in the images of mechanical parts to improve the level of automation in modern manufacturing,the research on image segmentation and separation of mechanical parts is of great significance.In the aspect of machine vision system construction,this article combines the characteristics of mechanical parts and builds a visual system that highlights the features of mechanical parts,including: digital cameras,light source equipment,image capture cards,PC,conveyors,and system software;light intensity for mechanical parts for comparison experiments,select the most common light intensity to highlight the feature information of the mechanical part.The image pre-processing of mechanical parts is the guarantee of image segmentation and separation of mechanical parts.This paper adopts bilateral filtering and segmented linear enhancement based on gray histogram.Based on the image signal-to-noise ratio and peak signal-to-noise ratio,the average filter,Gaussian filter and bilateral filter were evaluated.Bilateral filtering was used to remove noise while still effectively retaining edge information of mechanical parts;linear transformation was performed on the filtered image.The piecewise linear transform and histogram transform based on gray histogram,combined with the image gray mean value,gray standard deviation and subjective judgment of the processed image,the experiment shows that based on gray histogram piecewise linear transform in image enhancement of mechanical parts has good results.The image segmentation and separation algorithm of mechanical parts is the core technology of automatic analysis of mechanical parts.The effectiveness of image segmentation and separation algorithms directly affects the effectiveness of the extraction of the morphological features of mechanical parts.In this paper,the segmentation method based on threshold segmentation,contour detection and region detection is studied.An improved watershed algorithm based on distance image tag is designed to solve the problem of watershed algorithm over-segmentation.It also facilitates the extraction of segmented region information.There are multiple mechanical parts stacking in the area.This paper proposes a single unit determination criterion for mechanical parts and a watershed separation algorithm that gradually changes the threshold based on the cross-point number of area and skeleton lines.After a single mechanical part is separated from a single machine part based on the criteria for determining the mechanical parts.Extracted from the image,and then use the separation algorithm proposed in this paper to separate the remaining areas,and finally achieve complete separation of mechanical parts.The mechanical part feature model and the BP neural network mechanical part identification model are designed.This is based on the application of the mechanical part image segmentation and separation,and it is also the verification of the mechanical part’s segmentation and separation effect.On the basis of image segmentation and separation,the method of image feature extraction is further studied.The features of mechanical parts are analyzed from the aspects of appearance,structure,texture,frequency domain,and area,and the BP neural network is used to pattern recognition of mechanical parts.The experimental test has good classification effect and proves the effectiveness of the mechanical part separation and separation algorithm.