The Research On Human Target Detection And Tracking At Sea Based On DM8168

Marine accidents occurred frequently while marine resources were widely exploited and utilized.Maritime accidents also happened frequently with increasing maritime activities.The lives of victims can be guaranteed to the most extent if they got timely rescue when the accident happened.As an important research content of video processing,human target detection and tracking technology has been widely used in video security,human-computer interaction and other fields.After analyzing and studying the characteristics of human target in marine scene,this thesis applied human target detection and tracking algorithm to build a video processing system based on TMS320DM8168 and verified the algorithm.The background of the sea scene always changed dynamically and contained a lot of wave noise.In view of this environmental characteristic,inter-frame difference method was used to detect all moving areas in the video.In view of the character that the human target was very small relative to the sea background,this thesis realized the human target detection by detecting the human head.The Principal Component Analysis(PCA)method was used to reduce the feature dimensionality after the Histogram of Oriented Gradients(HOG)features of the training sample was extracted.And these features were used to train Support Vector Machines(SVM).HOG feature of the motion region was extracted,and human target was detected using SVM.In the process of target tracking,the scale estimation function and multi-feature fusion algorithm were used to improve the Kernel Correlation Filter(KCF)algorithm,because human target was easy to be covered and the scale of was prone to change while moving with the wave.The test sample set was used to compare the performance of the human target detection algorithm in this thesis with the traditional HOG+SVM algorithm.The experimental results showed that the proposed algorithm improves the detection speed without reducing the accuracy.The Object Tracking Benchmark(OTB)data set was used to compare the improved KCF algorithm in this thesis with the traditional KCF algorithm.The experimental results showed that the tracking accuracy and tracking speed of our algorithm were improved.In this thesis,the video processing platform based on TMS320DM8168 processor was introduced,and the software architecture and workflow were analyzed.The algorithm was transplanted after the software development environment was installed.Then the video processing system based on DM8168 was built,and the video captured by the HD camera was read and processed in real time.It had been proved in the experiment that our system can complete the detection and tracking of human target in marine scene,and it had certain engineering application value which laid a foundation for the following engineering application.