Research On Target Recognition And Tracking Algorithm Based On Micro Vision

Micromanipulation robot is a robot system which often works in a very tiny space,and can do very precise operation, which owns bright application prospect in many areas,such as biomedicine, new material preparation, integrated circuit manufacture, and so on.As a main control method in micromanipulation robot system, micro visual servo nowbecomes a hot issue during related research areas.In this paper, we’ll concentrate in research about micro visual servo, including targetrecognition, and visual servo tracking algorithm.In research on target recognition from micro visual image, this paper firstly ensurecavity as the main point in detection based on the specific characteristic about the targets inmicromanipulation system. After overall consideration about some nowadays popularimage object detection algorithms, we selected the method based on feature matching. Andthen, we employed a quietly new approach named ORB to extract the features of theimages, and we demonstrated the method that we adopted owned an excellent accuracy andreal-time property.After that, we developed the study on visual servo tracking algorithm. At first, weanalysed the advantage and weakness of standard Kalman filter and conventional adaptiveKalman filter, and aimed at their weakness that they were sensitive to process andobservation noise, we put the emphasis on the research on the robust Kalman filteralgorithm. Based on the works above, we proposed an improved adaptive Kalman filter,combined it with current statistic model, and the property of micro visual system, weselected the control method, and built a testing system on visual tracking, through thesimulation, we demonstrated the effectiveness of our method.