Application Of Visual Attention Model With Automatic Scale Selection In Target Recognition

Human can quickly locate the most interest area in the scene and pay the highly attention to it through the processing of a mechanism after receiving a lot of visual information,it is visual attention mechanism.The mechanism can greatly improve the efficiency of visual information processing for human,thus it has been widely applied in the field of computer image processing.Scale space theory is a framework of multi-scale image representation,objects will take the different form in different observation scale,so it is very important to describe the existence of the object.The edge detection technique under the multi-scale has played a vital in getting information from image and the the target recognition task in computer vision and image analysis.In this paper,we study some of the traditional classical visual attention models and edge detection algorithms,after learning the basic knowledge in the relevant aspects of the theory,expound the advantages and disadvantages of traditional methods,the biological plausibility and improvement ways of the coherent visual attention model,the important role of the automatic scale selection edge detection has played in target recognition,they all provide research ideas for this topic.This paper uses three typical visual attention phenomenon on the basis of neurophysiological research,they are: the continuous edge and the internal consistency of the significant target,the attraction of the center of the scene to form the final visual attention model.Firstly,utilize the automatic scale selection edge detection algorithm and the coherent visual attention model to obtain the continuous edge and the coherent area.The secondly,due to the two disadvantages of the coherent visual attention model: do not consider the edge feature and the existence of the cross-channel union,so we add the two methods together in one framework to make the continuous edge and coherence complementary advantages and improve recognition accuracy of the results.Thirdly,it uses the center attraction as a supplementary means to simulate the phenomenon that the human eyes are more willing to pay attention to the center of the scene to optimize the result of significant target recognition and make the new method has higher biologicalplausibility.Finally,the algorithm is applied to a particular environment,that is the target recognition task under the complex background,and embodies its practical efficiency in the target recognition task through the significance extraction of the target objects.