Research on Multimodal Visual Saliency Detection Based on BP Neural Network Algorithm

When the number of pictures or video signals to be processed by the computer reaches hundreds of millions, the useless energy consumed will be greatly increased. Therefore, how to make the computer automatically pay attention to and only deal with the areas of human interest is of great significance to improve the efficiency of computer vision tasks. In this paper, the research of multimodal visual saliency detection based on BPNN (BP neural network) algorithm is carried out. Establish a BPNN algorithm based on GA (genetic algorithm) optimization. Using GA to roughly search out a certain weight range, and taking the weight at this time as the initial weight of BPNN can improve the shortcomings of BPNN, such as easily falling into local minimum, slow convergence speed and causing oscillation effect. The research results show that the proposed method has excellent performance in multimodal visual saliency detection, with precision, recall and F value reaching 0.8207, 0.8455 and 0.7935 respectively. The conclusion shows that this model can detect prominent targets more accurately than other methods, which proves the effectiveness of the algorithm and can adaptively fuse visible and thermal infrared information.