Research And Implementation Of Plant Root Image Edge Detection Method

Root is one of the vital organs of plants. It grows in the soil, and its function is absorbing nutrient and water from the soil, and then transporting them to plant parts above the earth’s surface. So root plays an important role in the process of the growth of plants. In addition, root can make the fastest response to soil environment, and its growth is not only affected by soil properties, but also indirectly affected by the atmospheric environment. So, on forestry, agriculture and ecology, etc, studying growth of plant roots has important significance.This thesis is originated from the project of ministry of science and technology that research the dynamic in-situ nondestructive monitoring and analysis system about the growth of root of plants. The previous achievement has designed and implemented the system of root image processing and quantitative analysis, and this system can process simply root image of plants(such as image enhancement, image segmentation, etc) and measure root morphology parameters. But this system has some flaws, for example, complex root image can be seriously interfered with noise in the process and this can reduce the precision of measurements of root parameters; when extracting the skeleton of root image, the noise, such as burr, holes, etc, can result in errors; there is big errors about the number of root tips, etc. The work of this paper is to study problems of the system, and selecting appropriate algorithm is to reduce the errors, further improving the system, in order to make it more reliable to the actual research. In this paper, the main jobs are as below:① Traditional edge detection algorithm based on Canny operator has be selected in the previous achievement of this study, and then we research the edge detection algorithm based on traditional Canny operator, and find that this algorithm exists two problems: one is that, the algorithm is sensitive to the local noise, and detect the false edge easily, and lose edge detail; another is that, the adaptability of this algorithm is poor. The algorithm uses Gaussian filter in the process of smooth. Gaussian filter parameter determines the filtering effect. The parameter is not too big or too small, otherwise it is not conducive to edge detection. So according to the complexity of the root image, we respectively select improved Canny edge detection operator and improved anti-noise morphological edge detection algorithm to detect the edge of plant roots. Then two edge images are obtained by above two improved methods. Finally, the two improved methods is integrated by using wavelet transform to obtain the final edge image. Several edge detection algorithms are analyzed through the experiment in the paper, finally compared with the fusion algorithm, and find that the fusion algorithm not only has stronger ability to resist the noise, but also the root edge integrity is better, and the edge detail is relatively abundant.② In this article, the morphological preprocessing is applied to root edge image, and sag, burr, Isolated points are processed by technique of mathematical morphology, such as opening operation, closing operation, etc. Then we select thinning arithmetic of mathematical morphology to refine root image. In this paper, the algorithm of counting root tips has the judgment of protrusions(burr) on the skeleton to further filter protrusions, improving the accuracy of counting root tips. The above process can provide more reliable data source for subsequent measurement of root morphology parameters.