Research On Keytechniques Of On-line Surface Inspection System For The Bullet

In recent years, machine vision inspection technology has applied to in manyaspects of product quality testing, defect detection, however, detection of dim targetobject and low-contrast objects or auto-differentiation, such as the bullet surface defectsdetection, has no well-established methods and related products yet. In the modernizedbullet industrial production, there’s a crying need for a bullet surface online defectdetection system for military units.To face the objective demand of market, in this dissertation, the research was madeto solute these problems, which is surface defect image acquisition, original imageprocessing and feature extracting discriminant.First the key technical problems of online bullet surface defect detection isanalyzed, then an in-depth research on the detection principle is done, then choosingthe devices module by module and setting up the experimental platform, and thisplatform implements the image capture and vision information acquisition.In the image preprocessing stage, several methods for removing the system noise isstudied, and then the method of low contrast image enhancement is studied for thebullet image with low contrast, then, a multistage adaptive thresholding method forbinarization processing of grey image, whose effect is relatively ideal, was proposed,through the experiment contrasts several different approaches for image segmentation.In the feature extraction and recognition stage, analysising the geometric feature,gray feature, texture feature and invariant moment features. According to thesecharacteristics, defect classification has been doing, and analyzed the experiment data.Finally introduced the design and implementation of the software.The focus of this dissertation is solution design of a bullet surface online defectdetection system and the core algorithm such as the bullet surface defect imageprocessing, feature extraction, judging and the corresponding software code.In this dissertation the characteristics: algorithm for bullet surface characteristicsbased on Matlab verification，then transplantation of C++development mode, rely onthe development of OpenCV is more targeted, the code clear, easy to modify and refactor; In addition, based on linear array CCD (Charge Coupled Device, a chargecoupled device) bullet surface online defect detection, as far as we know, there is stillunreported in china.