As a leaf homologous organ, soybean pods are an essential factor in determining yield and quality of the grain. In this study, a recognition method of soybean pods and estimation of pods weight per plant were proposed based on improved YOLOv5 model. First, the YOLOv5 model was improved by using the coordinate attention (CA) module and the regression loss function of boundary box to detect and accurately count the pod targets on the living plants. Then, the prediction model was established to reliably estimate the yield of the whole soybean plant based on back propagation (BP) neural network with the topological structure of 5-120-1. Finally, compared with the traditional YOLOv5 model, the calculation and parameters of the proposed model were reduced by 17% and 7.6%, respectively. The results showed that the average precision (AP) value of the improved YOLOv5 model reached 91.7% with detection rate of 24.39 frames per millisecond. The mean square error (MSE) of the estimation for single pod weight was 0.00865, and the average coefficients of determination R2 between predicted and actual weight of a single pod was 0.945. The mean relative error (MRE) of the total weight estimation for all potted soybean plant was 0.122. The proposed method can provide technical support for not only the research and development of the pod’s real-time detection system, but also the intelligent breeding and yield estimation.

No abstract

The growth process of soybean plants needs a lot of water. The rapid detection of canopy wilting of soybean under drought stress is of great significance for soybean variety breeding, cultivation regulation and fine management. Aiming at the problems of cumbersome and time-consuming when the traditional chemical technology was used to determine soybean wilting index, a calculation method of wilting index for soybean canopy was proposed in this study based on multispectral images’ Fourier transform. Suinong 26, a northeast soybean variety, was taken as the object. First, four kinds of soybean multispectral images of green, red, red-edge and near-infrared channels were acquired by a Sequoia multispectral camera. Second, based on the multispectral reflection image preprocessed by median filter and mean filter, the target area of a multispectral image of the soybean canopy was extracted by the iterative threshold method and affine transformation algorithm, and the effective segmentation rate was 97.02%. In addition, Fourier transform was used to analyze the spectrum characteristics of the soybean canopy’s multispectral image. When the spectrum radius of each channel was 50, the energy reached more than 98% and was concentrated in the low-frequency region of the spectrum center. Finally, according to the difference between the low-frequency DC component and the proportion of total energy in the spectral radius of the multispectral images under normal and drought treatment, a calculation model of the soybean wilting index was constructed based on the energy spectrum of Fourier transform. The results showed that the difference of the wilting index between normal and drought treatment for the four channels (green, near-infrared, red and red-edge) was 2.38, 3.11, 3.56 and 4.11, respectively. The effectiveness of the wilting index was verified and analyzed by using the average leaf inclination angle. The determination coefficient R2 of the four channels between the calculated wilting index and the average leaf inclination angle was more than 0.85. This calculation method can provide a quantitative basis and technical support for the scientific regulation of ecological and morphological phenotypic traits of soybean plants under drought stress.

No abstract