Design And Implementation Of Multi-objective Detection And Recognition System Based On Deep Learning

With the rapid development of the information society,target classification recognition and detection are widely used in various fields such as image retrieval.Target recognition refers to searching for objects that need to be identified in known non-dynamic images and performing classification detection,mainly solving the problem of location and classification of multiple targets on the image.The traditional algorithm is to separate the extraction feature and the decision classification,but it is difficult to get better results in a more complex environment,and the traditional algorithm can not simultaneously target multiple different kinds of targets on the same image.Test and identify.In order to solve this problem,more and more researchers have applied methods such as deep learning and convolutional neural networks to the algorithm of target detection and recognition.Multi-objective recognition and detection based on deep learning is improved on the basis of the existing mature target recognition algorithm and deep learning network model.It can be adapted to more interference situations in practical applications.Compared with the traditional target detection and recognition algorithm,by introducing the convolutional neural network and the deep learning method,the extraction feature and the target classification can be simultaneously performed.Through a large number of experiments,the increase of the number of layers in the convolutional neural network can more effectively deal with object detection and classification and recognition under complex scenarios.In this paper,the traditional target detection and recognition algorithm and various multi-objective detection and recognition algorithms based on deep learning are deeply compared and studied.At the same time,this paper applies the new deep learning algorithm YOLO(You Only Look Once),and improves on this algorithm.By adjusting the image mesh division,the number of regional meshes is increased,and the network model is improved.The mAP on the object.At the same time,the paper also adds functions such as image acquisition,image annotation,image display,and timely warning,and transplants the trained model files to the high-performance embedded platform,thus realizing the multi-target detection and recognition system.The convolutional neural network algorithm has great potential in image feature extraction.Through a large number of samples for training,the network can automatically update the weight and complete the recognition and detection of the target object.