Arabidopsis Phenotype Detection Based On Computer Vision System

Arabidopsis is an important model plant in the field of botany, genetics and genetics. The study of Arabidopsis phenotype can extend the knowledge of its physiological functions, particularly the relationship between its gene and phenotype, as well as the effects of different environmental conditions to its growth.This paper uses computer vision technology to nondestructively detect the phenotypes of Arabidopsis, including the total area, contour features, and number of leave. The area and the number of leaves reflect the size and growth of the plant. The overall contour reflects the overall plant morphology and growth orientation and help to better understand the morphological characteristics of the plant. However, since the number of contour points are too large, in this paper we used Fourier transform to analyse the contour features. The number of Fourier descriptors is much smaller than the original contour points, which will save storage space. These phenotypic parameters not only can describe the growth of Arabidopsis quantitatively, but also can be used to analyze genome function of Arabidopsis. The main work of this thesis is as follows:First, the image preprocessing was conducted, including automatic black and white checkerboard location and corner detection, thus to correct and calibrate image and restore the measure characteristics of the Arabidopsis image. Then according to the features of the Arabidopsis and the background pixels, a new segmentation model was derived with principal component analysis and support vector machine. Finally, the total area of Arabidopsis was calculated. In addition, the contour was represented with elliptic Fourier descriptors. Also, the number of leaves was detected and the distance of each leaf apex to the center was computed. Experiment result shows that the proposed segmentation method and feature extraction approach can fast separate Arabidopsis from image and compute the phenotypes. Compared with the traditional manual observation and measurement, using computer vision technology to extract the phenotype of Arabidopsis greatly improves the efficiency, and achieves the goal of non-destructively detect the growth of Arabidopsis.