Video Target Detection Method Based On Convolutional Hopfield Neural Network Model

Moving target detection in video is an important research direction in the field of computer vision.It is very important in intelligent monitoring,assisted driving,vehicle tracking,human-computer interaction and other life scenes,and has a broad application prospect.In the process of video information collection,there are often many complex situations,such as camera movement,background movement,and non-rigid deformation of moving objects,which bring great difficulties to the detection of video moving targets.Although researchers have proposed many methods to detect video moving targets,there are still many shortcomings.In order to improve the detection and the accuracy of moving target detection in video,this paper simulates the processing mechanism of brain visual cortex information and conducts research on computational models and moving target detection based on Hopfiled neural network.Firstly,video target detection method based on Hopfield neural network is presented.The Hopfield neural network neurons and the image elements of sequential images are correspondingly constructed into a network,and the Hopfield neural network peripheral neurons and its own circular feedback mechanism can be used to obtain the space-time information in video by better using the context relationship and time sequence information.According to the Hebb learning rule that is closer to the cognitive mechanism of human visual system,the weight coefficient of the network can be adjusted without the supervision of the supervisor,so that the network convergence is stable at the equilibrium point,and the moving target of video can be effectively detected.Secongly,the Hopfield neural network structure based on local connection mode is proposed,and the convolutional Hopfield neural network model is established.To simulate the brain's visual cortex information processing mechanism,is proposed based on the Hopfield network structure of local connection method,convolution Hopfield network model is established,and through theoretical analysis and the relevant certificates,the convolution Hopfiled neural network is stable and convergent,and better retain the traditional Hopfiled neural network approach to learning and related functions.This model reduces the complexity of the network,reduces the computation,and improves the efficiency of video moving target.The convolution operation improves the ability of extracting features of video sequence and improves the performance of detecting video moving targets.Finally,in order to prove the validity of the proposed model,this paper selects VIVID(Video Verification of Identity)database,ILSVRC2015 VID database for testing.And compared with the experimental results of Narayana et al.on complex Background Videos(CBV)for testing,the convolutional Hopfield neural network can achieve the detection of moving targets in video images more quickly in the case of relatively good accuracy.