The Binoculaur Visual Object Recognize And Location On SIFT Algorithm

With the use of complicated and sophisticated technological trends in robots, the robots are becoming more intelligent and it would eventually lead to a standard computer operating system which is dedicatedly designed for robots. Machine vision is one of the key technologies to improve the intelligent level of robots. Binocular vision could simulate human visual capabilities, perceiving and comprehending the three-dimensional world, as a result, it is widely used in modern industries and our daily life. So, the research of binocular vision has a lot of theory meanings and practicality value. With the image information obtained by the binocular vision, this dissertation recognizes and locates the object using image features detected by SIFT algorithm, and controls the robot to grab objects. Our main work includes four aspects:object recognition, locating, camera calibration and manipulator control.The task of our work is to recognize and locate the non-specific objects. Therefore, we first describe the image features, analyze the advantages and disadvantages of these features. Introduction of feature extraction process of the SIFT operator is detailed relatively, when comparisons are making between several commonly used feature detection operators, and a comparative analysis of the strengths and weaknesses of these operators are conducted. Secondly, an improved algorithm is proposed for the drawbacks of SIFT, and a series experiments are done to verify the improved method. Finally, we use the improved SIFT algorithm to recognize object and calculate the three-dimensional position of object, Which send position of the object to robot control system, and control the manipulator to grab object.The experimental results show that:the improved SIFT algorithm can accurately identify targets and robustness by processing lots of images. The improved SIFT algorithm has keep not only the traditional SIFT algorithm’s invariance to rotation, scaling, partial occlusion, but also full invariance to affine, and enhancing resistance to the illumination change.