Research On Computer Object Recognition Based On Image

With the rapid development of multimedia technology, object recognition technology has received more and more attention, and has become a hot area in machine vision. The features of one specific object from different perspectives are different. On the basis of the phenomenon, the thesis selects the shape feature of the object in image as the criterion to distinguish different objects. The object images from different perspectives are chosen as templates, and then the binary features of the tested images and the template images from different perspectives are obtained. If the size of the images is same and there is only one object without shelter in the images, this thesis recognizes objects, according to the method of template matching. By analyzing the results of the calculation, the template image with highest coefficient of correlation is the recognition result.In order to improve the calculation speed of the system, the features of the shape are classified by the geometrical features in the process of recognition to divide the template images into different categories. The number of the maximum distance of each point of the outline to the center determines the geometrical features. According to the contour of the shape, the thesis determines which category the tested image belongs to and then do matching in the same category to reduce the computation time.The relevant experiments verify whether the system can recognize the obvious features or not. Analyze the experimental results of images of different objects and from different perspectives of the same object. A series of images from a certain range of perspective are obtained to match the images in the template library. Then this thesis compares the similarity coefficient of the tested images with the images in the template library to validate the sensitivity of the algorithm to the angle. Experimental results show that the recognition system is feasible and effective and the method proposed in this thesis meets the demand for adaptability of computer object recognition. In addition, this method can be applied in the study of the general object recognition.