Soybean single plant weight prediction method and system based on image segmentation and deep learning

Abstract

The application discloses a soybean single plant weight prediction method and system based on image segmentation and deep learning, and the system comprises a motion system, an image acquisition visual system, an image processing system and a weight prediction system, the motion system is used for controlling the system to move reasonably in a soybean field; the image acquisition visual system shoots single soybean plant samples in the field through a camera and transmits the samples to the image processing system; the image processing system performs pretreatment on soybean plant images and performs pod and main stem branch labeling, performs segmentation processing on the soybean plant images through a deep learning model to identify the number and types of pods, acquire the number of soybean grains on the main stem, and acquire the length and surface area of the main stem and branches; and the weight prediction system takes the number of pods, the number of soybean grains, the surface area of the main stem and branches as input features based on a trained integrated learning prediction model to predict the total weight of soybean grains of the soybean plant. The application provides valuable insights for soybean breeding and planting optimization through an automatic soybean phenotype feature extraction and yield prediction method, and opens up a new technical approach for crop phenomics research.

Images