Application And Research Of An Intelligent Algorithms In Image Saliency Detection

The rapid development of Internet technology has brought about an explosion of information and data.Vision is the most effective way for humans to interact with the outside world.It allows us to obtain the color,position,state and other properties of external objects without direct contact,and to respond to changes in the environment.With the continuous improvement of intelligent algorithms and computer performance,how to give vision to computers,so that computers can observe and perceive the world like humans,has become the goal pursued by non-mathematicians.However,in practical applications,due to the limitations of the modeling method and the calculation method,there is a certain gap between the significant area of the image detected by the algorithm model and the actual human eye gaze area.The detection accuracy and the applicability of the algorithm need to be further improve.To this end,this paper uses the intelligent optimization algorithm to provide a method for image saliency detection,and carried out the following work:1.The development status of image saliency detection at home and abroad is analyzed.Under this background,the saliency detection of image is taken as the research object,and a convolutional neural network model optimized by firefly algorithm is introduced.The related concepts and principles of firefly algorithm and convolutional neural network are introduced,and the common methods of image saliency detection are studied.2.The principle and mathematical model of firefly algorithm are studied.The chaotic sequence is used to optimize the problem of population initialization inhomogeneity in the firefly algorithm.The convergence of the algorithm is analyzed,the termination condition of the algorithm is set,and the flow of the improved algorithm is analyzed.Carding.3.The improved firefly algorithm is used for the iterative update of the connection weights and thresholds between the layers of the neural network.The firefly individuals representing the weights and thresholds are coded in real vector form,and the training accuracy of the network is measured by the error value.And analyze the parameters of the sampling layer and the fully connected layer in the network structure,and study the commonly used sampling and full connection technology.4.Establish an image saliency detection model based on the improved algorithm and experiment on the classic image recognition data set.Compared with the purely computational AC algorithm,the SR algorithm based on spatial frequency domain analysis,the FT algorithm based on image frequency domain analysis,and the SF algorithm based on filter acceleration,the actual operation efficiency of the optimization model is verified by practice.Through experiments and simulations,it can be verified that the threshold and weight between the layers of the neural network are updated by using the firefly optimization algorithm.The image saliency detection model is established according to the requirements of the recognition accuracy,compared with the commonly used algorithms.,in terms of applicability and accuracy,have certain advantages.