Research On The Theory Of Visual Object Tracking Based On Particle Filter For Autonomous Robot

Visual target tracking is the cutting-edge research subject of image processing, pattern recognition, and precision-guided, and its main task is that tracking the moving target in image sequences, understanding and describing the behavior of the object. Visual target tracking has important applications in battlefield reconnaissance using drones, guidance of cruise missiles, attack helicopters at ground and fire control system of main battle tank. Therefore, the research that related to the visual target tracking technology has important academic significance and practical value.The researches of this paper included the following three aspects.First, the visual target location was estimated based on particle filter algorithm. The particle filter algorithm was created from state transition model of the target and visual features model of the target. The performance and shortcomings of visual target location estimated based on particle filter algorithm was analyzed by using experiments. In creating visual features model of the target, the independent study as the target color was applied. The degree of membership with color is used in describing color characteristics of the target. The visual characteristics of target were described using the color characteristics of the target from the independent study. While the location of the target was estimated, a threshold value was set. If the summation of all the particles were smaller than the threshold value, the target was regard as nonentity.Secondly, target motion prediction was used in image sequences. The predicted results were used in the distribution function of the establishment. In estimating target motion and predicting target motion, the calculus-like decomposition method was used to break down the nonlinear movement. After breaking down the nonlinear movement, nonlinear movement was approximately a simple linear sport, for example:uniform linear motion, uniformly accelerated linear motion, uniform deceleration linear motion. Experiments was used to prove effect of the methodFinally, target was tracked based on the mobile robot. When size of target, lens of the camera, view of the camera was knowable, measuring target distance and target position in the image using a camera. The azimuth of target was estimated based on the target position in the image. The target was located in polar coordinates based on the azimuth of target and the distance. After locating the target, the target was tracked.