In order to solve the problems of low accuracy and long time of pixel matching in online three-dimension（3D） measurement， a pixel matching algorithm based on improved grid motion statistics features is presented. In this method， only one frame of sinusoidal grating is projected onto the measured object， five frames of deformed patterns caused by the moving object with equal displacement are captured by camera， and the corresponding background light field is extracted. Then， the matching pairs of feature points are extracted by fast rotation invariant feature algorithm， and then the mismatching pairs are eliminated by using the improved grid motion statistical feature algorithm. Finally， according to the characteristics of high precision in online 3D measurement， five equivalent phase-shifting deformed patterns are obtained by taking the value of the highest frequency of the statistical displacement， and then the 3D shape of online moving object is reconstructed by using equal step phase-shifting algorithm. Large numbers of experiments verify the effectiveness and feasibility of the proposed method， and compared with the traditional and online FTP algorithm， the proposed method is more robust and can reconstruct the 3D shape with higher accuracy.