Reaearch On Image Recognition And Compression Method Of High-Speed Moving Object

Video images of moving objects contain a lot of important information data.Howeve1;due to the fast speed of high-speed moving objects,the moving images of high-speed moving objects have problems such as acquisition distortion,complex processing,and poor analysis results.In this project,we have studied image recognition and image compression methods for high-speed moving objects,and selected high-speed moving laser dot infrared images of 3D printers and high-speed moving fixed-wing UAV video images in real industry for algorithm verification and result analysis.The main research contents are as follows:(1)Analyzes the image characteristics of high-speed moving objects and the image recognition process,and proposes a high-speed moving object image recognition method based on improved convolutional neural networks.According to the image blur caused by high-speed moving objects,which makes it difficult for traditional image recognition methods to artificially extract image features,the deep learning method based on convolutional neural network(CNN)is used for image recognition research.Aiming at the problems of poor training efficiency and low accuracy of the existing CNN methods,some related improvement schemes are designed to improve the speed and accuracy of image recognition.(2)Research on image recognition of high speed moving laser point infrared image of 30 printer.Two laser point image recognition models,CNN and improved CNN,are established respectively,and the two models are compared and analyzed.The experimental results show that the improved CNN model The recognition efficiency and accuracy are greatly improved;finally,the improved CNN model is used to predict the temperature of the laser center point,and a good prediction result is obtained within the error range.(3)Studies the redundant information of high-speed moving object video images,and proposes a high-speed moving object video compression method based on the combination of intra-frame coding and inter-frame prediction.Based on static image compression,the wavelet transform method is added to the traditional BP neural network to perform image compression to reduce the spatial redundancy information in the frame.At the same time,the motion information compensation is carried out for the video sequence of the high-speed moving object to reduce the time between frames.Time redundant information,the final reconstructed video greatly reduces the amount of video data under the premise of ensuring key motion information.(4)Research on image compression of high speed fixed wing UAV video.First,perform intra-frame lossy compression based on the combination of wavelet transform and neural network on a single frame of UAV moving images,and then perform inter-frame prediction coding based on motion estimation and motion compensation on the compressed UAV sequence,and finally compress The before and after UAV motion videos are compared and analyzed in terms of target tracking and video transmission effects.The results show that the compression system can greatly reduce redundant data information and improve transmission efficiency while retaining UAV motion information.