Aiming at the problems of labor intensity, low efficiency and poor working environment, a system of poultry slaughtering robot based on machine vision was designed. The system consisted of four parts, which included a rectangular coordinate manipulator body, poultry carcass suspension conveyor, PLC control system and machine vision system. Collecting poultry carcass and poultry muzzle images online through the visual system, the carcass image was pretreated by median filter and grayscale enhancement, and then the image was binarized by using the total fixed threshold, after setting the area threshold and obtaining the binary image which cut off the poultry wings and legs through the closed operation and hole filling. Finally, the binary image contour was calculated, the center of mass coordinates was got. In order to obtain the coordinates of center of muzzle, the poultry muzzle was positioned by the maximum internal circle method. The center of the maximum internal circle was the center of muzzle. Combining the coordinates of the center of mass with the coordinates of the center of muzzle, and then converting the coordinates information to digital information, the data of digital information would be transmitted to the PLC controller, to guide the manipulator to complete the work of the evisceration. The results of the experiment showed that the accuracy of the coordinates of poultry profile contour centroid and muzzle center was ±1.6 pixel and ±1.52 pixel. The eviscerating manipulator can accurately grasp the internal organs, the residual viscus in the body was 7.63%, and the breaking rate was 23%.