Key Research On Rice Yield Traits Scorer For Multi-varieties

Rice is one of the major crops in the world, and has proved to be the most important food in china. Rice phenotypic traits are required to be detected, which are significant in rice breeding and functional gene analysis. In rice breeding research, the rice phenotypic traits are used to screen out high-yield, high quality, stress-tolerated rice varieties gained by either traditional hybridization or modern molecular breeding. Meanwhile a large number of rice phenotypic traits need to be detected to establish a functional relationship between genotype and phenotype in rice gene analysis.Rice phenotypic traits can be divided into growth traits and yield traits, this study focuses on obtaining rice yield traits acqusition, including panicle length measurement, rice threshing and grain traits extraction. However, traditional measurement method is manual by ruler and steelyard, which is slow and inconsistent. Therefore the traditional method is far from meeting the requirement of mordern rice research. To solve the problem, an automatic yield traits scorer is developed based on mechanical design, machine vision, image processing, and automatic control technology in this study. This research could provide a novel method for extracting the panicle length, threshing the rice, and detecting the grain parameters with high accuracy and high efficiency.The panicle length automatic extraction system was based on a dual camera imaging. Panicle neck was accurately located by the specific algorithm from detailed image. The curved path of the panicle and the accurate panicle length was obtained by skeleton extraction, main path analysis and image registration algorithm from whole panicle image. The panicle length measurement system was tested by approximately3000panicle samples. And the results showed that the mean absolute percentage error (MAPE) of the system was about1.86%, and the system efficiency was about900panilces per hour. The automatic rice threshing device was designed according to manually threshing procedure. The system separated spikelets from panicles with scratch method and programmable logic controller was applied for automation. The test results showed that average threshing loss rate was0.87%and grain-broken rate0.05%in this threshing unit. The grain trait automatic extraction system was able to aquire the total spikelets number, filled spikelets number, grain length, grain width, and thousand grain weight by image analysis and automatic control.2132rice samples were tested by the system and skilled workers, and the results proved that the MAPE for total spikelets number, filled spikelets number, grain length, and grain width was1.52%,1.81%,1.98%and2.80%respectively. In addition, the average measurement efficiency of system was61.23seconds per plant.The results demonstrated that there was a good consistency between the automatic measurement and manual measuremnet. Therefore the systems developed in the study could be applied to replace manual measurements. The rice mature phenotypic traits measured by the systems could be not only used to characterize rice growth situation directly, but also applied for genetic association analysis study (GWAS).This study proposed a new the rice yield traits extraction method, which overcomed the issues of existing measurment such as low efficiency, unreliability and limited parameters. In this paper, the rice yield traits scorer was developed by using the technology of mechanical design, electrical control, machine vision, image processing and data analysis. And the main innovations included:(1) proposed a fast, automatic approach for rice panicle length measurement, which solved the problem of the panicle neck identification, and bent panicle measurement;(2) presented a high-precision method for rice automatic threshing, which solved the key issues of threshing residues and damage;(3) invented a effective method to obtain multiple grain traits, which solved the problems of filled/unfilled spikelets identification, impurity recognition and parameters extraction. This research was an important part of crop phenomics, and the application of this research was able to contribute to the rice breeding and functional gene analysis. This study would provide a useful scientific tool for the rice research.