Study On The Key Technique Of Fast And Remote Acquiring Plants Point Clouds In The Natural Environment

The shape and physiological and biochemical indexes showed in the process of the plant growth play an important role in plant protection, genetic breading, crop yield forecasting, pest control and mutant plant selecting. Moreover, capturing plant growth information speedily, efficiently, nondestructively has been the pursuit of goals. To capture the data of plant growth efficiently, this paper studied the technology that can capture the point cloud data of plant in natural environment, and built a complete system that can capture 3D point cloud. The main study contents are as follows:(1) The scheme acquiring the point cloud data of object in nature environment was proposed. This scheme mainly consists of building the hardware system and the key technology of acquiring point cloud. A simple, low cost, easy-install 3D scanner that can capture the point cloud model of object in the nature world was built according the scheme.(2) The camera calibration method based on camera-image distance calibration that can work well in nature environment was proposed. These studies mainly include the camera local length calibration method and system calibration method. By analyzing the existing calibration algorithm, this paper adopted the camera–image distance calibration algorithm, and used the self-made calibration plate to calibrate focal length of the camera. Experiments show that the focus calculated by this method within an accuracy of 0.3mm. To calculate the model parameters, the three-point calibration method is used to calibrate camera, experiments show that the calibration accuracy is 3mm.(3) Completed the research of computing 3D point cloud based on structured light strip on the image. The process mainly includes Image preprocessing, extracting structured light strip center line, stereo matching and calculating three-dimensional point. Firstly, in order to optimize image quality and improve the accuracy of calculating point cloud, the proper image processing algorithms was selected to extracting the center of the structured light stripe, in order to calculate the actual three-dimensional point data, this paper use the camera calibration parameters and the principle of binocular vision to match the image of light stripes center. The relationship between the match point pairs and the camera was used to calculate the actual three-dimensional coordinates of object.In order to verify the 3D scanner built by the author according to the method that proposed in this paper, the author make contrast experiment under different scanning environment by selecting different objects consisted of still life, potted plants and plant leaves respectively as scanning model. Experiments show that the 3D scanning system can work well in nature environment. The accuracy of point cloud can reduce to 0.56 mm, and the 3D scanner can capture the point cloud data speedily, and is not easy to be disturbed by ambient light. It can meet the need of the project capturing the 3D point cloud data of the crop.