Research On Visual Autonomic Development Model Based On Attention Mechanism

Visual attention is a key psychological adjustment mechanism in the process of human information processing.It can enable people to find valuable information in complex situations quickly under massive resources.Researchers call this mechanism a visual selective attention mechanism.This paper focuses on the development of a visual development model with selective attention.The main research results are as follows:1)The sparse coding is introduced into the visual selective attention modelAiming at the problems of inaccurate feature extraction and large amount of computation in traditional visual models,a visual selective attention model combining sparse coding with bottom-up is proposed to process the original input image of Itti model,which should ignore and abandon,so that the feature extraction process is more accurate.2)Combine bottom-up and top-down visual developmental modelsAiming at the unity of functional principle of single visual attention model,a visual model combining top-down and bottom-up is proposed.Weight? combines the significant graph weights of two different categories of visual attention models together.Discuss the value of ? and get the correct value of ? to make the best performance of the model.3)A visual selective attention model with weight developmentAiming at the problem that the Itti model does not have the change of the weight with the change of tasks in the sub-feature graph saliency map normalization process,a weight-developable visualselective attention model is proposed.The algorithm uses three-layer self-organizing neural network and Itti model combines the decision-making optimization,through training to obtain the optimal weight update.The visual attention model proposed in this paper realizes the dynamic update of weights with a kind of development thought,which greatly improves the accuracy of the target feature extraction.