The Study On Bullet Apparent Defect Detection Based On AOI System

The bullets, which is one of the m odern basic weapon, is very im portant significance to the war. So a safe and efficient production of bullets becom e an important measure of a country’ s military-industrial levels. However, although the bullets safety is so important, but in our country, production testing after completion of the bullet is still a major part dependent on manual detection methods. this approach not only test results unstable, while human cons umption, detect poor results. After working long hours, the product is very eas y to fatigue, it is easy to lead to false positives and false negatives, leaving a very serious safety hazard.According to the shortcom ings of the manual inspection method, this dissertation studies the development of a set that is fo r detecting the apparent bullet based on AOI(Automatic Optic Inspection) detection system, and accomplish automation, high-speed, high-precision requirements. To adapt to m odern intelligent produc tion worldwide, it has an important significance.This dissertation describes the status of the dom estic bullet detec tion, and then introduces the developm ent of ma chine vision and AOI technology to detect bullets brought new opportunities. Then the overall fr amework for the characteristics of the bullet itself, the system design, the use of multi-angle lighting the way to ensure that the effect of optical im aging, using linear CCD cam eras mining maps to ensure the accuracy and completeness of the information bullets apparent.After acquiring the im age into the im age pre-processing stage, pres entation and discussion of several de-noising m ethod, and then the contrast obvious problem for acquiring an im age, we us a linear gray-s cale transformation of the way the image enhancement, finally introduces several bullets im age segmentation methods available, after testing and comparing we select Otsu algorithm.After he image pre-processing finished, using a fast connected component labeling algorithm for image connected component labeling. In the image feature extraction and recognition phase, acco rding to th e characteristics of the b ullet itself, introdu ced the geometry which is usef ul to defect det ection and classification, gray feature and invariant moments, and the corresponding defect classification indicator.Finally, introduce the design and implem entation of software, including interface design, image acquisition, image processing, database use, upper and lower com puter communication control, and given the realization of results.Image processing in this di ssertation are the conclus ions obtained after Matlab validation before perform ing porting C + + language, it has its own way of independence, extended and reusability addition. In this paper, using linear CCD camera apparent of bullet m ining map, then defect de tection, as far as we know at present, is first reported in China. The final average speed detection we get is 0.7 seconds per bullet, meet the desired requirement. The study on this system can also be used on other conical objects.