Online-estimation of Image Jacobian based on adaptive Kalman filter

摘要

Focus on the uncalibrated visual servo, we established a filtering system based on Kalman Filter, and validated it by Matlab. Then we discussed the stability precondition of the control system with Kalman Filter in detail, and made a conclusion that if and only if the origin system is completely controllable and observable, the filtering system can converged to the expected position. Refer to the problem that we can't get the statistical information of the real system, we designed an adaptive Kalman filter and applied it in a real-time track task, and with the track error is within 5 pixel.