The Recognition And Location Of Dense Complex Objects Based On Machine Vision

The research of recognition and location based on machine vision is a frontiersubject of science and technology which is developing rapidly recent years. It hasmany applications in many fields, and it also has very wide development prospects.And as the21st century information digitalization, visualization, automation trendgoes, more and more industrial production lines need this kind of automatic visualidentification system for updating. The identification and location based on machinevision is using a camera image acquisition device to convert3D objects from realspace into machine language, using the locating software to locate and recognizeaccording to the features, then taking further control and operation.The research of recognition and location based on machine vision is a verybroad field and an emerging issue. This article mainly aimed at the research of denseregular arrayed objects’ recognition and location, and using the cellular image withdense honeycomb holes as the representative, carries on the analysis and practice.In the chapter of visual locating system’s design and calibration, discussedabout the system operation, the selection of the system hardware, the selection oflight source, the software process and calibration problems. In the chapter of imagepre-processing，discussed about the problems such as image fuzzy, excess noise,insufficient lighting, uneven lighting, then gave the solutions of each defectaccording to the problems of different images. At last, this article used the cellularimage with dense honeycomb holes as the representative, carried on the analysis andpractice. In the next chapter, this article discussed the recognition and location of thecellular image with dense honeycomb holes, with the use of the locating softwarenamed Hexsight, developed a module for locating the cellular image with densehoneycomb holes. And this article also carried on the analysis and practice. In theimage post-processing chapter, this article proposed the two innovational fittingmethods, which are Y type template fitting method and the baseline fitting method.The Y type template fitting method uses Y to replace the honeycomb holeimage as the feature template, and uses the geometric properties of hexagon to makea program for honeycomb hole center location coordinates for fitting processing, itspurpose is to increase the recognition rate, reduce the honeycomb hole contourdistortion for the effects of positioning. Baseline fitting method is to use the samethe cellular image characteristic, according to the arrangement characteristics of itsrules to repeat， putting forward the method of drawing baselines to fit thehoneycomb hole center which is not recognized, it can have very good treatmenteffect to the uneven arrangement of honeycomb holes, the contour distortion etc. This chapter puts forward two kinds fitting method, they are main innovation pointsof this article. With further development, these methods will play a large role in theactual industrial production line, they will help to improve the overall efficiency andaccuracy of visual recognition system.This paper realizes the construction of the system which is based on themachine vision. Many samples are given through all aspects of the research resultsof the analysis. For the object of "compact complex objects", this article uses thecellular image with dense honeycomb holes as the example, with the imagepre-processing and image post-processing, it accomplishes its goals.