Research On Single Color Image Object Detection Method Based On Convolutional Neural Network

Object detection is an extremely important branch in the field of computer vision,and its role in the field of vision is irreplaceable.With the continuous development of artificial intelligence and computer vision,object detection is applied to many fields such as visual navigation,military public security,and virtual reality.However,the detection effect does not really meet the actual needs,so how to achieve better goals.The detection effect has become an urgent problem to be solved in the field of object detection and even in the whole computer vision field.Based on a comprehensive analysis of the research status at home and abroad,combined with deep learning,computer vision and other related knowledge,we have conducted in-depth research on how to obtain better object detection results.The specific research contents are as follows:Firstly,the basic concepts and related knowledge of object detection are introduced.The convolutional neural network used in this topic and the convolutional layer,pooling layer and fully connected layer of convolutional neural network are briefly introduced.Secondly,according to the characteristics of each level feature map extracted by convolutional neural network,an object detection method based on feature map weighted fusion is proposed.This method uses the convolutional neural network to extract color image features.And according to the feature map weighted fusion idea,the feature map weighted fusion method is proposed and the new feature map is obtained by the method.Then enter the new feature map into the improved RPN network to get the region proposals.Finally the ROI Pooling layer is used to extract the region proposals features and classify the features and simultaneously perform the object position bounding box regression operation to achieve the object detection.Thirdly,an object detection method based on full convolutional neural network is proposed.The method uses the residual network to extract the color image features and performs feature map weighted fusion to obtain new feature maps of color images.And the convolutional layer that increases the number of output channels is used to obtain the position sensitive feature map and the improved RPN network is used to extract the region proposals.Then through the operation of pooling and other operations to obtain the feature vector of the region proposals and classify the vector to achieve object detection.Finally,the experimental verification of the proposed two object detection methods is carried out.The proposed method is compared with the experimental results of the existing object detection methods.