Research On Image-Based Three-dimensional Flower Reconstruction

There are a variety of flowers in the nature. Accurate grasp of the structural features of the flowering plants plays an important role in the field of agricultural production. But the topology of the flowers is generally more complex. The conventional method for flower modeling is not only with low efficiency, but require labor-intensive. Image-based modeling technology is a hot research topic in today’s field of computer vision, it take advantage of the image modeling technology directly to the scene modeling, compared with the traditional graphics-based modeling technology, it’s easy to operate, efficient and so on. Binocular stereo vision technology is an important branch of research in the field of image modeling, it directly simulates the physical structure of the human eyes, and it’s efficient and convenient that the three-dimensional information of the object can be measured in different conditions. In this paper binocular stereo vision technology, three-dimensional reconstruction experiment on flowering plants. Five aspects of the binocular stereo vision technology in the system structure are studied. They are image pre-processing of flower, camera calibration, stereo matching algorithm and three-dimensional reconstruction technology of flower. It focuses on image pre-processing of flower and stereo matching algorithm. Target calibration method based on two-dimensional plane of the camera calibration is used in the experiments, and calibrated work is finished more accurately. About Stereo matching algorithm, this paper studied the SIFT local feature extraction algorithm, and verified in the experiment. Finally, in the experiment, I use triangulated surface modeling techniques to recover the three-dimensional model of the flowering plants, basically to achieve the desired effect.