Family Environment To Identify And Capture, Multiple Types Of Objects Based On The Qr Code

Home service robot is one of the most important directions of the robot research. Objects handling is the basic task of the service robot in home environment. Home environment is dynamic. There are also many kinds of objects with different shapes and colors in home environment. It becomes more difficult to recognize and handle the target object for the service robot. The purpose of this paper is to develop the method with which the robot can find out the target object quickly and handle it from numerous objects in home environment.This paper introduces the current research work and the development trend of service robot in the world, and summarizes the achievements gained in the object recognition and the object handling fields. It also analyses the difficulties in home environment. This paper gives some solutions to deal with these difficulties.Handling of objects can be realized by two steps: recognition of the target object in real time, and object handling based on the poses of the object.In order to recognize various objects correctly, a recognition method based on QR Code is put forward according to the feature of home environment. The designed QR Code is composed of inside information and outside mark. QR Code has great capacity and can be recognized quickly, moreover, it can express Chinese characters efficiently etc. The object basic information and the handling information can be recorded in the inside information of the QR Code. Outside mark is designed so that the robot can find the QR Code in a long distance. The designed QR Code is attached to the object exterior respectively. The object can be located and recognized by processing images taken by the robot hand camera.In order to handle the target objects precisely, image-based vision servo technique is adopted. The relation between the image information and the control signal is studied by an artificial neural network (ANN). There are four center points in the four color regions of the outside mark. Comparing their coordinates between the current image and the expected image by image processing.Their differences are used as the inputs of the ANN. The variational values of the robot joints are used as the outputs of the ANN. Experiments show that the ANN can learn the mapping relationship between the image information and the control signal very well and the robot can execute the assigned tasks successfully and robustly.In home environment, objects are put at random. The robot sometimes collides with other objects during handling the target object in the expected pose. It is necessary to detect the obstacles and plan the robot motion. In this paper the obstacles are classified, route planning is obtained based on raster method, trajectory planning is obtained in joint space, making sure that the robot can handle the target object without any collision.In the end, by experiments, the home service robot handles object with multi-handling poses. Firstly the camera and the robot hand-eye are calibrated. Secondly the cascade of boosted classifiers is trained based on Haar-like features which can help the robot find out the QR Code. At last the robot decodes the QR Code and gets its current pose. The world coordinate is set up considering the center of the target object as the origion, and then the grasping-tasks are accomplished successfully by motion planning.