Research Of Algorithm For Objects Extraction And Identification In Machine Visual Image

Images are the main carrier of information. How to fetch sufficient and precise information in bulk and multifarious ones is a frequently asked question. In order to extract much more accurate information in intelligent machine design, we need strengthen the processing and analysis of machine vision image. So the processing of machine vision image has became an important content of message processing. Machine vision has developed vigorously since the 1980s and the extraction and identification of objects as an important research direction and method has great potential in development and application.Firstly, this article introduces some basic knowledge such as the concept and method of image processing, image digitization, etc. We also put forward some methods of extraction and identification systematically to the objects of different characteristics in the machine vision image and experiment them with crops image. The methods include image pretreatment, classification and analysis to different objects, the extraction of dominant color and general color, image matching, objects identification test and so on.Secondly, we use the contrast of RGB and HSV patterns to deal with the color-dominant object's extraction and identification and take every color component grayscale severally to contrast. According to the particular case of different object, we choose the optimum component grayscale and process it by threshold method so as to improve the accuracy of the result. At last, we identify the color-dominant object and get its position in image by matrix operating in the form of binary image. In the processing of non-color-dominant object, this article designed two matching algorithm to extract and identify object's information and contrast in comparative experiment. Then get the best matching effects in the end.Moreover, in this article we try to realize the de-noising processing based on wavelet transform in the process of image pretreatment and it's important to images of bad quality. This system selects MATLAB as development platform of the processing. All process is given the MATLAB code and results contrast.The main conclusions of this subject are as follows.(1) We should use optimal methods to deal with the objects with different characteristics during the extraction and identification of different objects in machine vision image. In view of the noticeable features of some objects, algorithms of great pertinence should be adopted.(2) During the identification of flower object, in view of its application in intelligent agricultural machine, we proposed the central position identification algorithm. So Procedures are simplified and accurate information that we need is retained as accurate parameter of intelligent agricultural machine.(3) For the general object, we adopted the extraction and identification algorithm based on image matching technology.