Acclimatization Calculation Research And Implementation Based On Multi-vision

Human-computer interaction is a complex of technical science, which investigates the interaction and communications through mutual understanding between human and computers to solve problems of information managements, services and processing in maximized degree for human needs, but robustness is not high in human-computer interaction. The principal reason for the problem is largely that people must adapt to the computer, but the computer cannot adapt to the complex environment, e.g. variable illumination, variable view and occlusion. The complicated environment has a strong impact on image visional perception. Image quality has an important effect on the recognition rate.This paper proposes the framework of acclimatization calculation, and we discuss the effects of illumination and view factors on image visional perception quality in this paper, that is to say, the computer detects the illumination and view factors and computes them, then makes judgment based on a certain method to remove interference, improve robustness of system and recognition rate. The work of this paper can be summarized as below:First, multiple-camera environment is presented to solve occlusion and view change. The vision domain of a single camera is limited, which makes the computer cannot obtain the full information of target. In order to solve occlusion and view change, multi-vision is used to obtain target images in this paper.Second, acclimatization calculation is put forward. The new method evaluates the image perception quality with illumination and view extracted from image by wavelet transform. The new method is simple and fast. The experiment results prove the method is efficient and can select the best image matching the characteristics of image perception qualityFinally, a new information fusion method based on acclimatization calculation framework is proposed. We combine SVM and Dempster Shafer (D-S) evidence theory. By improving SVM (One Against All), we convert the distance between test image and training class into probability, then use D-S evidence theory to integrate probability, and finally make a judgment on the test image.