A Method Of 3D Vector Modeling And Analyzing For Plant Root

The root system architecture(RSA) is the result of root's growing and branching,it determines the plant's ability to absorb and transmit moisture and nutrients, and affects the efficiency of absorbing and utilizing nutrients and water significantly. RSA can be divided into 2D RSA(or plane RSA) and 3D RSA(or stereo RSA). 3D RSA refers to the spatial distribution of different types of roots in the growth medium, including connection relations(topology) and spatial distribution characteristics. Relative to 2D RSA, 3D RSA can reflect root's geometry and distribution characteristics under the ground more accurately and comprehensively. Therefore, 3D RSA's measurement and analysis has gradually and internationally become an important research direction and hotspot in recent years.The main technical bottleneck of 3D RSA's quantitative research is that soil and other medium prevents people to observing root in situ directly, and the complexity and overlaps of root itself make it more difficult to be measured precisely. Currently, the technologies and methods of acquiring 3D RSA in situ mainly include magnetic resonance imaging(MRI), X-ray computed tomography(XCT), ground penetrating radar(GPR) and so on.With the support of Natural Science Foundation of China(NSFC) and National High-tech R&D Program(863 Program), South China Agricultural University(SCAU)'s agricultural imaging detection technology laboratory(AIDTL) has studied methods for imaging root in situ based on X-ray CT, processing and analyzing spatial sequence images, 3D visualizing root in situ, measuring basic parameters of RSA based on visual model and so on. The AIDTL has also constructed and developed CT system for imaging root in situ and visualization system for visualizing root in 3D space. In order to ulteriorly measure and analyze parameters of 3D RSA, on the basis of the work mentioned above, this paper research the method of 3D vector modeling and analyzing for plant root based on spatial sequence images. The vector model(VM) would be more convenient to describe root's spatial geometry in situ quantitatively, and more suitable for measuring, calculating andanalyzing 3D RSA parameters. The VM would aslo reduce the amount of data, make it easier to measure and calculate 3D RSA parameters, and improve the precision and efficiency significantly. In this paper, the main research work is as follows:(1) On basis of understanding the technologies of detecting 3D RSA at home and abroad by acquainting with the previous work of research group and reading and analyzing related literatures, the overall scheme and technical route of research work were determined by comparing and analyzing the existing methods of calculating and analyzing RSA parameters.(2) Plant root X-ray CT images acquiring and processing. Plant root CT images were acquired with CBCT; After being filtered by median method and segmented by region growing method, CT images were packed into 3D volume data, which was the original data of 3D vector modeling for plant root.(3) Method of 3D vector modeling for plant root. Firstly, virtual slices were acquired by reslicing volume data constructed above, they contained root cross sections. Initial root nodes were built with the informations of root cross sections, such as centroid, area,equivalent spherical radius, equivalent spherical perimeter and so on. Secondly, root banches were created by grouping root nodes according to the morphological characteristics of plant root. Thirdly, fitted curves that reflected the skeleton of root were created by fitting root branches, and the fitted curves were sampled at fixed interval to produce new root centroid where the volume data was resliced again(reslice plane was vertical to fitted curve and went through the new root centroid). New root nodes were created from the new virtual slices and replaced with old ones. In the end, root branches' topological relationships were set up by judging the connectivity of each root branch.(4) Algorithms for calculating 3D RSA parameters. Algorithms were designed to calculate the number of root branches, the length of root, the angle between child root branch and parent root branch, the volume of root and the surface area of root based on the plant root 3D VM proposed above. The number of root branches equaled to the number of branches in plant root 3D VM; The length of root equaled to the sum of the distance between the adjacent root nodes within each root branch; The angle between child root branch and parent root branch equaled to the angle bewteen child root branch's first rootnode normal and parent root branch's fork root node normal; The volume of root equaled to the sum of the volume of root segments made by neighboring root nodes within each root branch; The surface area of root equaled to to the sum of the surface area of root segments made by neighboring root nodes within each root branch. The volume and surface area of root segment was calculated by the equivalent for the frustum of a cone.(5) The system of 3D vector modeling for root plant. A pirmary system were developed by programming with C++, Qt, ITK(Insight Segmentation and Registration Toolkit) and VTK(Visualization Toolkit) on the Windows. The pirmary system was used to build and analyze 3D VM for plant root.(6) Experimental verified and analyzed the pirmary system. Mining method was used to obtain plant root, and the root were measured manually to get root length, root number,root angle, root volume and root surface area. The primary system was verified by comparing with mining method. The result showed that error of root branches number was4.55%, error of total root length was 2.1%, error of total surface area was 6.26, error of total volume was 1.02%, and mean error of angle was 2.58%.In summary, the following contents were completed in this paper: Researched the method of 3D vector modeling for plant root based on X-ray CT images; Designed algorithms for calculating 3D RSA paramenters based on 3D VM; Developed a relevant system which could be used to measure the basic 3D RSA parameters automatically. The work of this paper could improve the precision and efficiency of the measurement of 3D RSA.