Research On Object Recognition And Localization Technology Based On NAO Robot

Robotics is a modern and widely used interdisciplinary subject which covering the control theory, multiple disciplines of computer technology, artificial intelligence, bionics etc. It integrates the development of the above disciplines, can be said to stand in the forefront of technology and application. In the robot field, it is very important to get the information of the external environment through the sensors, which Is also a necessary pre step of subsequent navigation, identification, positioning and other functions.the common way uses infrared sensing, sound wave, structured light, camera, etc.. Compared with other methods, the use of optical camera and computer vision methods in terms of cost, convenience and other aspects of a significant advantage. In this paper, based on the NAO robot platform, we use NAO camera and computer vision technology to make the application of the NAO robot object recognition and positioning method. The main work is as follows:1) a fast object recognition method based on the quasi physical property estimation. This method can realize the object recognition function by comparing the images acquired by the NAO robot camera and the image of the object to be recognized in advance. Because we plan to achieve a traversal of the environment and real-time identification of the object system and the robot's own hardware equipment constraints, the real-time requirements for the algorithm is very high. Therefore, this paper proposes a combination of the proposed algorithm and the statistical matching algorithm for fast image matching method. Because it is recommended that a false positive candidate window is removed in a short time, the method can not only improve the recognition accuracy, but also improve the speed.2) a kind of real-time object orientation method for NAO robot is proposed. In order to realize the interaction between the NAO robot and the object, the real-time positioning of the object is required. The upper and lower two cameras of the NAO robot have almost no overlap, so the method of monocular vision localization can be used, but the existing methods usually adopt the method of feature point matching, and the computation cost is large. Based on this, this paper proposes a fast locating method, which is used to calculate the distance between the front and back frame. It needs only one camera and no need to calibrate the camera.3) established prototype system, and according to the specific structure of the NAO to establish the forward and inverse kinematics model and the reach of the arm could be calculated, in their localization to object calculation is able to take to and can take the manipulation arm grasping of objects is realized.