PTZ Camera Control Based On Dynamic Target Detection

PTZ(Pan Tilt Zoom)camera refers to a camera with omni-directional pan-tilt rotation and lens zoom function.In recent years,it has been widely used in the surveillance field.In order to further improve the intelligence level of the surveillance system,image processing algorithms are currently used to automatically detect dynamic targets which we control the rotation of the PTZ camera to track in real time.However,the actual surveillance scene is complex and diverse,and the rotation of the camera makes the background area of the image not fixed,which increase the difficulty of dynamic target detection and positioning.At the same time,there is no definite rule for the movement of the target,and it is difficult to establish an accurate model,which also puts forward higher requirements on the control strategy of the camera.Therefore,the researches of efficient and accurate dynamic target detection algorithms and PTZ camera control strategies are of great significance to the realization of intelligent surveillance.In order to enhance the performance of the Gaussian Mixture Model detection algorithm,this paper adopts a new initialization method on the basis of the traditional algorithm and adds the global illumination change detection link.And based on the image information entropy,an adaptive Gaussian Mixture Model is established,so that the algorithm has the ability to adapt to complex scenes.Experiments show that this algorithm can effectively locate dynamic targets and extract their complete contours.Secondly,this paper improves the online tracking algorithm of the Kernelized Correlation Filters,and fuses HOG(Histogram of Oriented Gradient)features with color features for target feature extraction.And this paper adds the occlusion detection and relocation mechanism,which make the algorithm more robust.The improved algorithm has been evaluated both qualitatively and quantitatively on the OTB data set.The results show that the algorithm has better accuracy and real-time performance than traditional algorithms.Finally,the detection algorithm is combined with the online tracking algorithm,and a fuzzy control strategy is designed to control the PTZ camera.This paper constructs an active tracking system and verifies the effectiveness of the tracking system in a variety of realistic scenarios.The verification results show that the system can automatically detect dynamic targets in the scene.It can overcome the influence of background changing,and achieve continuous positioning of dynamic targets.At the same time,the PTZ camera can realize autonomous rotation with the change of the target's position,so that the target is in the center of the field of view.The statistical analyses of the PTZ camera rotation error rate and system operation time show that the system has good accuracy and real-time performance.Therefore,the system is of engineering application value.