Study On The Visual Recognition And Calibration Of Blueberry

Blueberry picker is an intelligent robot, which can run autonomously and continuously on outside and hills. Its research, which involves theories and technologies of many aspects, and embodies the latest achievements of information and artificial intelligence, is of great value in research and application. In recent years, all fruit pickers and harvest robots in agriculture and forestry have received great attention all over the world. Among all the blueberry harvesting technologies, vision-based blueberry detection and positioning are the main challenges. The key problems of the blueberry picker are the rapid detection and positioning of the blueberry. Compared with other approaches, stereo vision system is a passive sensor, which does not emit light and radiation. And stereo vision system has many advantages such as cheap, accurate, and can acquire rich information.The thesis aims at the study of blueberry detection and positioning based on the stereo vision system. The basic problem for stereo vision system is the accurate calibration of the cameras. Different from that in the laboratory environment, the calibration for blueberry picker must be done quickly and accurately as the cameras are mounted on the robot, and the work space of the robot is the outside and many should be in the forestry, therefore the calibration procedure should be flexible enough. In the system that we have implemented, a flexible technique has been used to easily calibrate a camera. It is well suited for use without specialized knowledge of computer geometry. The calibration technique only requires the camera to observe some ticked points on the arm. While the arm is moving, a serial of points can be acquired. The test results show that the proposed method is valid.The stereo matching problem and hand-eye calibration system are discussed in detail in this thesis. And this thesis has done deep research in stereo matching. An inverse stereo matching method is proposed to realize the stereo matching, and the method is better on the effects of the qualitative and quantitative analysis than the traditional template matching method. Through the experimental study of three-dimensional localization and verification, the onboard calibration system and stereotactic hand-eye coordination unity is realized.According to the demands of the amount of calculation, intelligent level, real time requirement, which asked by the stereo vision system of the blueberry picker, a new blueberry detection method s are proposed based on vision information, including the blueberry segmentation method of the artificial neural network model based on the color information and the method based on the shape information, and they were validated by experiments. A grading method according to the size and shape of blueberry is discussed and experiment based on visual technology and torque description method.The comprehensive experiments show the blueberries detection, positioning and classification methods based on visual system have high practical and theoretical value.