Autonomous Robot Visual Target Identification And Positioning And Navigation Applications

Real-time image processing, object recognition , navigation as well as real-time computing is hot problem in computer vision field. This paper has summarized the current research achievement of robot vision and used existing laboratory experiments platform,in order to do in-depth study for the visual processing. By studying monocular vision three-dimensional image restoration and the calibration principle of binocular vision, and then propose use different variable focal length cameras to capture images and achieve the unknown two-dimensional plane fixed target identification, location, and on this basis, By two-dimensional plane image recovering three-dimensional image. Its main tasks are as follow:1,Beacause the visual sensor is easy to lose the precision of the image in the visual image pre-processing, causing image distortion, we introduct Camera Depth Principle, and proposes using a binocular camera system architecture to resolve the depth restore of objects in the scene, because of the two cameras of this camera system structure has different focal length and collect the near and far images of objects is different, so which provided the preconditions for the depth recovery of the image.2,About calibration of the camera,By analysising and researching the traditional camera calibration technique and the camera model in the various coordinates,Using perspective transformation matrix deal with camera calibration. Abase on above,Making a detailed study for the active camera calibration technique, In which the calibration techniques about rotating and moving parallel of the camera are discussed, and finally self-calibration technology of robot is in-depth explored, settlement the problem that the robot can in real time make camera calibration, under the changing field of view.3,About the feature extraction and target recognition, The feature extraction is affected the optical flow of movement and the moving playground has been studied, By researching different feature extraction methods, and step to step to draw the geometric space object feature extraction. Through researching Bayesian networks identification algorithm to enhance object recognition and to reduce the interference of the outside world for feature recognition. After researched recognition of the color characteristics,We introduced ray tracing in computer graphics principles to improve further the object recognition accuracy and Anti-jamming capability, ray tracing theory also has a low requirement on the advantages of light, but the disadvantage is robot's memory and computing power cost higher.4,About positioning and navigation,By researching the problems of robot localization, we proposed appropriate solutions.After compared and studied the map representation, we us the above method of the map representation to make the laboratory environment represented .5,In robot vision applications,We studied a robot model, which is different from binocular vision robot, which uses cameras that is installed in the head of robot to replace the whole camera for image acquisition, and electronic compass is used in a laboratory environment to be corrected direction of the robot.So the robot can reach the target.