A shadow-based method to calculate the percentage of filled rice grains. (October 2016)

• Record Type: Journal Article
• Authors: Liu, Tao; Wu, Wei; Chen, Wen; Sun, Chengming; Chen, Chen; Wang, Rui; Zhu, Xinkai; Guo, Wenshan
• Is Part Of: Biosystems engineering. Volume 150(2016:Oct.)
• Page Start: 79
• Page End: 88
• Publication Date: 2016-10
• Subjects: Rice -- Percent filled grains -- Shadow features -- Digital image -- Computer vision; Bioengineering -- Periodicals; Agricultural engineering -- Periodicals; Biological systems -- Periodicals; Génie rural -- Périodiques; Systèmes biologiques -- Périodiques; 631
• DOI: 10.1016/j.biosystemseng.2016.07.011
• Languages: English
• ISSNs: 1537-5110

Abstract

Abstract : Rice grain filling is a critical factor that determines the grain yield. It is important to measure the percentage of filled grains (PFR) in rice production management and scientific research. Current methods to measure filled grain percentage are generally manual, which are all time-consuming and labour-intensive with subjective results. Here, we designed an image analysis-based method to measure the percentage of filled grains using four light sources to generate grains shadows from four different directions. The differences of grain shadows between filled and unfilled grains were found out. The ratio of shadow characteristics to grain characteristics distinguished filled and unfilled grains. The conveyor belt with a vibrating feeder and controlled variable-speed was used to measure batched grains. The maximum measuring speed of the conveyor belt was about 60–100 grains/s, and the proper measuring speed was about 40–50 grains/s. Support vector machine (SVM) identified the unfilled grains, and the percentage of the unfilled grains was calculated for 8 Indica and 8 Japonica rice cultivars. The average false positive rate for Indica rice was 3.85%, and the average false negative rate was 5.44%. The average false positive rate for Japonica rice was 5.11%, and the average false negative rate was 3.54%. All these results indicate that this method is reliable and can be used for fast and intelligent measurement of filled grain percentage. The method shows great potential in improving the efficiency of grains' trait evaluation in crop breeding and cultivation research. Highlights: We estimate percentage of filled rice grains from images. Method based on shadow features. We test indica rice and japonica rice with computer vision method. The false positive and false negatives rates were below 5%.