At present, assessment based on traditional scale and manual measure are the main methods for evaluation of hand function for stroke patient. However, due to the disadvantages of large errors, poor efficiency and discontinuous evaluation results, it’s difficult for these methods to meet rehabilitation needs. This paper presents a quantitative evaluation method of wrist extension based on multidimensional vision evaluation system, the system captures action parameter using optical motion capture device and computer vision technology, it can get the motion data of patient’s both hands through multidimensional vision acquisition and intelligent analysis and presents the analysis result. This paper introduces hardware and software structure of the system, data acquisition process and analysis of evaluation result in detail and the result is compared with the result of traditional scale and manual measurement. In this paper, we use the ‘wrist back’ as an example for experiment, the result can meet the requirements of clinical accuracy and can be more accurate and flexible. This method also can make up for the shortcomings of discontinuity and nonprecision.