A calculation method of phenotypic traits of soybean pods based on image processing technology

As a homologous organ of the leaf, the pod is an essential factor in determining the yield and quality of the grain. The physiological and ecological characteristics of pods are a critical basis for soybean breeding and yield estimation. In order to avoid time-consuming and laborious problems in the traditional manual measurement process, a fast calculation method of pod's phenotypic traits was proposed based on image processing technology. First, two varieties (Heihe 49 and Suinong 26) were taken as the research objects. The self-developed image acquisition platform was used to obtain the color images of soybean pods at different growth stages. Second, the geometric distortion of the raw image was registered by using the linear mapping equation. In addition, the edge of the pod was extracted by the Gaussian filter function and Canny detection operator. Finally, the length, width, and area of pods were calculated based on the minimum circumscribed rectangle method, the maximum inscribed circle method, and the pixel count of pod's contour area combined with the template calibration method. The shape and the number of seeds were calculated based on the interval duty cycle method, the interval duty cycle combined with the arc length of pod's convex edge. The results showed that the accuracy between the calculated values and the standard values of pod length, pod width, pod area, pod shape, and the number of seeds was 97.8%, 95.1%, 97.1%, 93.2%, and 85%, respectively. Thus, the proposed method can provide not only a rapid detection for quantitative indexes of soybean seed test and yield estimation, but also a theoretical basis and technical knowledge for soybean breeding and scientific cultivation.