In the process of detecting and tracking moving human target based on visual images, in order to implement fast real-time tracking and make tracking more accurately, Tracking based on Kalman prediction is used according to Kalman filter’s recursion capacity to estimate the target location of the next time. Simulation analysis in 2-D space and indoor experiment are done. Both the simulation analysis and test result confirm that the tracking speed and accuracy has been improved effectively base on Kalman prediction which offers guarantee for the following image processing and analysis.