For motion error abduction in CNC machine tools, complicated mathematical models are often needed to detect specific motion error for a specific CNC machine tool in the literature. A convenient method to simplify motion error abduction was proposed. The corner distribution on a divided error circle image generated by numerical control system was detected and the RBF neural network was combined to recognize motion error. Firstly, a new corner which indicated the distance from circle curve to circle center saltation of 16-piece divided circle was defined. The average radius and corner number of every piece of the divided circle were put into characteristic matrix to represent error circle images. In order to verify the performance of the characteristic matrix, SVM was employed to test five types of motion error circle images, which turned out to be an excellent solution with high recognition accuracy. Lastly, a RBF neural network was applied to recognize the motion errors by using the characteristic matrix as input and every motion error as output. Results show that the proposed method is of high efficiency and good accuracy for motion error abduction.