Pedestrian Detection Re-identification Technology Based On Deep Learning

The rapid development of society has made people pay more and more attention to the safety of public society.Many large public places are gradually installing surveillance cameras on a wide range of angles to form a monitoring network.However,in the face of the massive data brought by the dramatic increase in surveillance video,how to make the analysis and use of surveillance video more efficient and intelligent is a problem to be solved.The intelligent video analysis method based on pedestrian recognition has become a hot spot in the field of computer vision.The Person Re-Identification(ReID)technology is a technique for judging whether pedestrian images appearing under different surveillance cameras belong to the same pedestrian.Applying deep learning to pedestrian recognition has become a hot research topic.Compared with the method of artificially extracting features,the deep convolutional neural network adopts the automatic learning method to obtain the characteristics of the image from the data and classify the image.,has practical significance.Firstly,the paper summarizes the algorithms,features and deep learning network framework commonly used in pedestrian recognition.The architecture of deep convolutional neural networks is analyzed in detail,and the ResNet residual network widely used as image classification is further studied.Secondly,the paper carefully analyzes the learning model needed for the expansion of the training set and discusses in detail the generation of the confrontation network,and effectively expands the data set by generating the confrontation network.Then,the paper selects the deep residual network(ResNet-50)as the basic model of the experiment,and on this basis,it improves the feature extraction and analysis of pedestrians,which effectively improves the recognition rate of specific pedestrians.At the same time,under the framework of PyTorch,the improved deep network model is fine-tuned to achieve the best recognition effect.Finally,the paper analyzes the requirements and designs and implements the pedestrian recognition re-recognition system based on deep learning.Through the simple input of the pedestrian image to be queried,the pedestrian recognition result can be intuitively feedback.The experimental results show that the model trained by the improved method can learn the features with higher robustness and effectively improve the recognition rate of pedestrian recognition.