Detection of low signal noise ratio small even point moving target in the condition of complicated background is discussed. A new approach based on Kalman filter, mathematical morphology, motion analysis and two stage hypothesis testing to detect and track point target in such condition is presented. The experiment results with real IR image sequence show that the proposed method can greatly improve the SNR, and effectively detect and track the moving point targets. It can especially dissolve the problems that the targets miss and new targets occur during the detecting and tracking.