A new line detection algorithm - Automatic measurement of character parameter of rapeseed plant by LSD

Abstract

Molecular biology shows that phenotype of the crop is the ultimate concern in the research of crop yield, and to evaluate quality of results of the selective breeding, and it's also needed to acquire ideal plants by measuring each key parameter of phenotype of the crop, so as to popularize them. This article takes general rapeseed plant as researching example. Through adopting the new automatic measurement method-LSD (Line Segment Detector), it extracts line segments of image of the rapeseed plant rapidly; Then we joint and combine the extracted line segments that belong to the same stem, and adopt reasonable threshold value restriction respectively at different stages to remove the line segments having few contributions; finally, on the basis of the endpoint information of the whole jointed line segment and proportional scale relationship of the reference substance, we calculate the overall plant height, height of the first effective branch and length of the branch. At the end of the experiment, the measuring result of this article is compared with the parameter value acquired by measurement of high precision laser scanner. The result shows that: measuring error of plant height is within 3.9% and that of other parameters is less than 1.3%. Thus it proves that the method proposed in this article is effective, it can automatically extract and measure parameters of rapeseed plant phenotype and lay the foundation of researching relationship between phenotype parameters and yield and mechanized production, so as to better guide field work; Moreover, due to the high speed, low price and large quantity of memory space of measurement based on two-dimensional space, it has more benefits in the practical application.