Research On Automatic Tracking Algorithm Of Crane Hook Based On Video

Tower crane driver in the process of the operation of the tower,there is a visual blind spot, leading to the driver’s work intensity, a little careless, it will cause the occurrence of safety accidents. At present, the intelligent monitoring equipment used in the tower crane is mainly used to measure some important performance parameters. Some of the construction site using the general network camera to solve the blind area of vision, but it is only a simple video surveillance, intelligent degree is low, some colleges and universities have also studied the crane hook tracking. But in the zoom module, lens zoom can not be achieved,the low level of intelligence.The crane driver during the operation of visual blind and visual fatigue problems, the tracking algorithm of tower crane hook based on optical flow method, including feature detection, feature tracking and camera zoom three part. In order to enlarge the recognition area of the hook, the marker is placed on the top of the hook. The first corner marker extraction of region of interest by the improved SIFT algorithm,improved SIFT algorithm greatly accelerate the speed of feature point detection and its application to improve the efficiency of the algorithm is effective in target tracking. Then, the feature points are tracked by the Pyramid optical flow algorithm, but in practical engineering applications,due to the interference of light changes, it will cause the drift of the corners,and then lead to the wrong tracking results. To solve this problem, this paper adopts a new corner classification algorithm, the algorithm can effectively filter the wrong point tracking, with unstable feature points selecting effective features, target gradually decreased and even disappeared. Aiming at the shortcoming of feature point tracking,this paper uses the method of feature point tracking and particle filter.The covariance matrix is used to describe the target, which is regarded as the observation value of particle filter. In the zoom lens module, the lens automatic zoom model. Through the automatic zoom lens model, so that the monitoring target is always displayed in the appropriate size in the screen.In order to verify the validity of the tracking algorithm, the experiments were carried out on the markers of running video and large scale rotating target video data, verify the validity of the tracking algorithm; and through marker motion simulation video hook, verify the effectiveness of automatic double variable lens model..