Target Recognition Method Based On Difference Features Description Of Image And Its Application

Image recognition is a specific application of pattern recognition technology in the field of image. It uses the principle of pattern recognition, takes the main features of image as the basis, and then takes the use of a computer to find out the object in the image, or do classification for objects in the image. Target recognition of image belongs to the field of artificial intelligence research, which has a wide range of applications. With the rapid development of computer technology, multimedia technology and artificial intelligence technology, a large number of image target recognitions have been generated accordingly.Based on the analysis of existing variety of image recognition methods, this paper has studied on the new target recognition technology based on features description. The main theoretical and experimental researches are in the following aspects.Based on the theory of similarity measure for image matching, this paper creatively put forward a new theory system of difference measure. A difference measure method based on fuzzy membership had been established: Firstly, the concept of difference was proposed. After that, its calculating formula was defined, and its related properties were analyzed. Then, the above theory was applied to the image processing; the gray values of image pixels were converted to difference values by the conversion formula. By doing this, the sensitivity of the human eye to gray level was transformed from nonlinear change to linear change. It’s conducive to the further processing of the image. After that, this paper introduced the concept of difference value into two aspects of image feature points and feature regions. At first, the classical SIFT algorithm was improved. A large number of standard images were used for testing. The advantages and disadvantages of improved method were analysis and compared. A new region segmentation method was proposed based on the difference, named improved OTSU method. A feature description vector of a single region was defined based on the difference value, shape, angle and other features. For the target consist of multiple regions, a combine feature description was also needed by the establishment of adjacent triangular meshes. The theoretical analysis and experimental simulation verified the effectiveness of this description method. Finally, through the study and application of the actual power line inspection and traffic sign images, the procedures of image processing were integrated, and the experimental system was carried out. Thus the practical value and feasibility of this method was shown.