Research On Grading Of Lotus Root Appearance Quality Based On Computer Vision

Currently, the segment of lotus root’s wholesale is mainly based on artificial selection. the amount of harvest of lotus root is huge, and the period of harvest is long, which both bring great trouble to artificial classification. Meanwhile, a huge amount of manual labor will lead to the workers tired, which, as a result, reducing the efficiency and precision. All of these will have a direct impact on economic benefit of lotus root’s wholesale and marketing. In this circumstance, we want to establish a system, which can classify lotus root according to their external quality automatically. At last, maximize the economic benefit of lotus root.This thesis take local lotus root of Nanjing for example. Introducing some basic theories and methods of computer vision technology used in lotus root’s external quality classification. This paper mainly focus on these fields:(1) Establish the standard of lotus root’s external quality classification according to computer vision technology. Establish a union function of lotus root’s length and tortuosity. Find the linear function between extracted length(unit pixel) and lotus root’s actual length(nuit mm) through mathematics analysis, and get appropriate coefficient.(2) Utilize the method of OTSU to segment collected images, and compare segmentation images of two methods, according to my findings, the segmentation of otsu has a better effect on segmenting. Use the technology of image projection and morphology to cutter de-noised images. The technology of image projection can obtain the shape of lotus root and calculate its tortuosity. Combine the erosion treatment and image projection technology cutter louts root to acquire the number of lotus root section and calculate the length of every section.(3) Test the precision of classification on a sample of 100 lotus roots, the precision of lotus root section is 100%, precision of comprehensive rating at level one is 90%, level two is 85.7%, level three is 80%.The system can normally satisfy the basic need.(4) Develop a system of lotus root’s external quality classification based on Matlab, and apply the software copyright for it.