No abstract

Rice (Oryza sativa L.) yield can be quantified by measuring the number of ears per square meter, grain number per ear, percentage of filled grains (PFG), and 1000‐grain weight, of which PFG is an important yield parameter. Therefore, it is pivotal to improve the measuring efficiency of PFG. The objective of the present study was to construct an automatic method for measuring PFG using the shadow features of grains. Filled grains can be distinguished from unfilled grains by their shadow characteristics under lamplight. Image segmentation algorithms were used to extract rice grains and shadows. Learning vector quantization neural networks then classified rice grains based on their shadow characteristics. To improve the flexibility of the measurement system, 19 typical rice cultivars were used to train the networks. Sixteen cultivars from two subspecies were selected to validate the new PFG measurement method, and the results showed that the PFG measured by the new method was closely related to the PFG measured by the water‐float method. The correlation coefficient between the two methods was 0.952 (P < 0.01). Shadow characteristics showed a significant difference between filled and unfilled grains. We concluded that shadow traits of rice grains could be used to discriminate filled and unfilled grains. The new PFG measurement method is a reliable method to estimate the PFG of rice and also shows great potential to improve the efficiency of trait evaluation in rice breeding and cultivation. The percentage of filled grains of rice was measured. Characteristics of grains’ shadow were used to identify filled and unfilled grains. Grain shadow characteristics values were extracted by image processing technology. Artificial neural network model is a valuable tool for grain classification.