Abstract: Aiming at the medical service robot used in the indoor environment, this study proposes a method to test the pose accuracy of its working surface based on the binocular vision system. This method uses a binocular vision coordinate system to measure targets fixed on the working surface of the service robot, aligns the measurement system with the robot base coordinate system through the nonlinear least squares method, and integrates the multi-eye image data to achieve the accuracy test of the working surface. Finally, the vision test program was tested and verified on a mobile service robot model ABIR X8. According to the accuracy index given in GB/T 38124 Service Robot Performance Test Method, the test results of its pose accuracy were obtained.