Research Of Detection On Moving Target Under Moving Base

In recent years,with the rapid development of computer-related technology,computer-related technology has been deep into all aspects of people's lives.Moving target detection technology is one of the key underlying technologies in the field of computer vision,and it is also a hotspot in the field of computer vision.It has important practical value in real-time investigation,real-time monitoring and real-time control.The moving target detection technology under moving base is the key and difficult point in the moving target detection technology and the research and improvement of the algorithm has always been paid attention to by a large number of scholars at home and abroad.However,due to the complexity of the dynamic scene itself,the variability of the moving target scale and the occlusion of the moving target,the problem of moving target detection under the moving base is still very challenging.Obtaining higher precision and better real-time moving base moving target detection algorithm must be the future development trend of this field.Based on this,this paper has carried on the exploration and research to the moving target detection technology under the moving base,among which the main research achievement and the innovation work are as follows:In order to overcome the influence of background noise on the accuracy of moving target detection,enhance the real-time performance and improve traditional moving target detection method's inadaptation to the change of illumination.A moving target detection algorithm based on Fourier transform and kernel function gray statistic is proposed.The algorithm first introduces the evaluation function into the selection of the feature matching block,and performs the block matching of the video image background.Then a Fourier transform phase correlation algorithm is used to estimate the global motion compensation parameters.Then we calculate the Gaussian kernel function value of each image sub-block,establish the gray function graph of the kernel function and judge the region of the moving target through the change of the adjacent frame Gaussian kernel function value.Finally,the image sub-block containing the moving target is segmented,and the moving target detection under moving base is finally completed.The experimental results show that the proposed algorithm can effectively suppress the background interference and has good detection accuracy and good real-time performance.It also has some adaptability to the interference of illumination changes during the moving target detection process.A novel moving target detection algorithm based on improved CLG optical flow model and Gaussian pyramid is proposed to solve the problem that the classical optical flow algorithm can not adapt to the change of illumination and the large amount of computation.Firstly,the gradient hypothesis conservation CLG introduced to traditional optical flow model,the combined structure texture decomposition treatment to reduce light flow estimate the impact of changes in light.Then the Gaussian pyramid is used to calculate the optical flow and the optical flow of the other points is obtained through the optical flow iterations of the larger gray scale points,thus reducing the computational complexity and improving the detection efficiency.Finally,according to the above-mentioned displacement and amplitude of the optical flow to,the background motion vector is calculated and the moving target is detected.The experimental results show that compared with the classical optical flow method,the algorithm can adapt to the influence of illuminate changes on the detection accuracy of the moving target,and the computational complexity of the algorithm is reduced to a great extent,the detection efficiency of the moving target under moving base is increased.Aiming at the problem that the moving target detection algorithm is difficult to detect the moving target accurately under the moving base condition,a moving target detection algorithm based on the cellular automata is proposed.Firstly,the video image is segmented according to the SLIC algorithm,and the background modeling of the video image is carried out by using the multimodal hybrid dynamic texture model.Secondly,obtain optimized saliency map by combination of the space-time saliency detection and the automatic update mechanism of cellular automata.Finally,the moving target in the video image is obtained by appropriate threshold segmentation of the optimized graph.The experimental results show that the algorithm can effectively suppress the influence of the non-moving object in the video image on the detection result under the moving base condition,and detect the moving target with high precision.