Pedestrian Re-identification Based On Deep Learning

In recent years,in the research of computer' vision problems,pedestrian re-identification has been widely concerned by scholars as a hot topic.Pedestrian re-recognition still has problems such as changes in camera angle of view,changes in light,changes in pedestrian posture,and partial occlusion of pedestrian images.How to extract more discriminative pedestrian information for pedestrian re-identification is still a problem facing pedestrian re-identification.In order to solve the above problems,this paper studies pedestrian re-recognition in terms of models and algorithms based on deep learning methods,and proposes the following model.The traditional pedestrian re-identification method relies on artificially constructed visual features,which is easily affected by other external factors and has low recognition accuracy.The deep learning model can extract features autonomously,but as the number of network layers deepens,the gradient disappears.The residual network can alleviate the gradient disappearance problem,but the extracted feature information is not used rationally.Partial occlusion of pedestrian images is another important factor that affects the accuracy of pedestrian re-identification.Aiming at the above problems,this paper proposes a pedestrian re-recognition algorithm combining random erasure and residual attention network.The algorithm:Firstly,On the basis of the residual network,the attention mechanism module is introduced,and the discriminative ability of the network is improved by strengthening the useful features and the features with little inhibition.Secondly,Introduce random erasure data enhancement method in order to reduce over-fitting phenomenon,at the same time improve network generalization ability,and solve the occlusion problem in pedestrian re-identification.Thirdly,Using triplet loss to supervise and train the fusion network to achieve better clustering effect of samples in the feature space and improve the accuracy of pedestrian re-recognition.The recognition accuracy of the algorithm on the Market-1501 data set and DukeMTMC-reID data set is verified through experiments.According to the traditional pedestrian re-identification method,the distinguishing ability of pedestrian characteristic information extracted is weak,which makes it difficult for the model to achieve better recognition results.In the metric feature space,the recognition accuracy is low due to the small distance between similar samples.Initial matching failure is another important factor that leads to low model recognition accuracy.To solve the above problems,this paper proposes a multi-loss pedestrian re-recognition algorithm based on the attention mechanism.The algorithm:Firstly,Use the twin attention mechanism network to extract features,and improve the discriminative ability of the network by emphasizing useful channel features that have less inhibitory effect;Secondly,Introduce center loss and use a multi-loss training method to achieve similar samples in feature space Better gathering,making heterogeneous samples farther away.The recognition accuracy of the algorithm on the Market-1501 data set and DukeMTMC-reID data set is verified through experiments.