Research On Detection And Tracking Algorithm Of Video Sequence Moving Target

Computer vision is now widely used in national production and life,such as: intelligent video surveillance,visual navigation,military field,and human-computer interaction.The detection and tracking of moving objects based on video sequences is an important direction of computer vision,and it has also achieved rapid development.However,due to the influence of many factors,such as shadows moving with the target,complex and varied background,sudden changes in light,changes in the attitude of the target,and occlusion,the detection and tracking effects of the moving target need to be further improved.For the commonly used ViBe algorithm,there are ghosts,stationary targets and shadow defects,and an optimized design is performed.In addition,aiming at the defects of compression tracking algorithm,a compression tracking algorithm combined with target estimation is proposed,which provides an effective solution for target detection and tracking.The main work is as follows:Firstly,aiming at the defects in the ViBe algorithm,the background updating strategy is designed to be updated in a ladder manner,and the former attractions are also given the characteristics of propagation and updating.Furthermore,the separation of high and low frequencies is performed by wavelet,the optimized ViBe algorithm is applied to the low frequency region,and the high-frequency region is used to extract detailed information by using the inter-frame difference method,and a shadow removal algorithm is designed by using spatial information in multiple scales.Then,this paper proposes a compression tracking algorithm combined with target estimation.For the original algorithm search area affected by the speed of movement,this paper introduces ORB feature point matching,and establishes a relation model by matching points to achieve the positioning of candidate targets;For the fixed scale defects,a scale adaptive method based on the variance between matching points is proposed;For the blindness of updating the original algorithm for each frame,this paper introduces the occlusion judgment factor and adjusts the update strategy of the classifier in combination with the number of matching points.Finally,in the detection of moving objects in video sequences,the comparison with ViBe algorithm shows that the algorithm can quickly absorb the interferences of ghosting and stationary objects as the background,and further through the comparison and analysis with classical detection algorithms,the effectiveness of the algorithm is demonstrated.In the detection of moving targets in the video sequence,the test sequence is chosen for experimentation,which shows that this paper not only meets the requirements of real-time,but also solves the defects of traditional compression algorithms.