Research On Object Recognition Method Based On Fusion Of Color And Shape Features

As an information carrier with rich content and convenient transmission,image is the main object of computer vision processing.With the intelligent development of modern life,computer vision technology based on image processing has been widely used in all walks of life.Object recognition,recognizing objects in the image,is the basis for the computer to understand the image content.As an important research content in computer vision,object recognition is the premise of many applications based on computer vision.How to establish an effective object recognition method,which can identify objects in images quickly and effectively,has always been a research hotspot in computer vision.Therefore,object recognition based on computer vision has important practical significance and application value.In this paper,the object recognition method based on the fusion of color and shape features is proposed by integrating the shape recognition and color recognition of objects in two-dimensional images,and it is verified and tested on the NAO robot development platform.The effectiveness of shape,color,texture and size of the object in image is been analyzed.Finally,color features and shape features are selected to describe the image features.To extract the shape features of objects,the shape description method based on contour features is studied,and a shape feature descriptor based on multi-scale contour segments is proposed.In this method,the shape descriptors with global and local characteristics are established by feature extracting and describing from shape contour segments with different lengths.Then the shape contour is simplified by contour point sampling method.Finally,the shape recognition is realized by measuring the similarity matching distance among shape descriptors using dynamic time normalization algorithm.To extract the color features of objects,firstly,the characteristics of various color spaces are studied.HSV color space conforming to human visual perception and Lab color space with color perception uniformity are selected for color feature extracting and describing.Then,a color histogram feature vector with spatial distribution characteristics is established,which can not only count the number of colors,but also count the distribution information of colors in space.Finally,to realize color recognition,the similarity between color feature vectors is measured based on Pearson correlation coefficient.The recognition results using shape feature or color feature alone are analyzed,it is found that there are limitations in object recognition using only one feature.In order to improve the object recognition rate,a method of weighted fusion of shape similarity and color similarity for object recognition is proposed.The experimental results show that the object recognition rate is effectively improved using the shape feature and the color feature merging method.Finally,to verify the validity and practicability of the algorithm,the algorithm is implemented in the NAO robot development platform.With the image acquisition module and voice module of NAO,a complete and interactive object recognition system based on machine vision is constructed.