Quantitative evaluation of soybean (Glycine max L. MERRILL) plant shape by image analysis.

Shape of soybean plants were evaluated by image analysis. Binary soybean images were obtained by thresholding the original video images. Then, two marginal frequency distributions of pixels of plant projection (silhouette) were obtained by scanning each binary image of soybean in both perpendicular (for Y-distribution) and parallel (for X-distribution) directions to the main axis which was defined to be a regression line on the main stem of plants. Because those distributions were obtained after similar transformation of the plant projection that the height became the same among all the plants examined, the range of Y-distributions was fixed while that of X-distributions corresponded to the normalized width of the plants. Therefore, it became possible to compare directly the types of the distributions of different plants. Among 875 plants of 175 varieties, the types of two marginal distributions vary widely and features of the original plant-shape were well represented by those types.