Research Of Obstacle Recognition Algorithm Based On Machine Vision

With researches in the field of machine vision and digital image processing deepening,obstacle recognition technology is widely used in many fields.But most of the existing visual image recognition system is very vulnerable to background factors,so that the recognition system cannot be popularized in real complex scenes.This paper introduces the process and base theories of recognition system.Then focuses on the above problem,and proposes an optimized GrabCut algorithm to separate the subject from background effectively in an image to reduce the influence of different backgrounds on the recognition system.This method uses the resolution reduction and image superposition techniques when separating the image which pretreated to guarantee the separation accuracy and ensure details of the edge in image.In this paper,several algorithms of image feature extraction are studied and compared in terms of operational efficiency.The paper mainly use HOG to extract the characters from the image without background.In the implementation of obstacle recognition,this paper constructs the recognition classifier by using support vector machine(SVM)and radial basis function(RBF).In the process of selecting optimal parameters of SVM classifier,an optimized grid optimization algorithm is proposed to optimize the parameter C and ? in the model.The experiment proves that the optimized algorithm is similar to the traditional one in terms of the accuracy of K-CV,but take much less time.Finally,this paper uses the above proposed optimization algorithms to design and implement a blinds obstacle avoiding system with the Opencv2.4.9 platform,and then verify the effect of the optimization algorithms proposed above in this paper with the recognition system.