Research On Characterization Method Of Rice Panicle Characters Based On X-ray CT

The characterization of rice panicle spikelet number,seeding rate and other traits is of great significance for rice yield calculation and genetic trait analysis.In the past,manual measurement methods were labor-intensive,and their efficiency and accuracy were low.Aiming at these problems,therefore,it is of great value in engineering application to develop the research of X-ray computed tomography in the characterization of rice panicle characters.First,the paper analyzed the characteristics of each part of the rice panicle in the CT image.Then,in view of the problem that the connection between the shoots and stems of the rice panicle and their overlap makes it impossible to accurately count,the corresponding characterization method of the panicle character was studied.The distance transformation watershed algorithm was used to segment the rice to remove the overlap between the rice and the stems of the spikes.In view of the diversity of rice panicle structure,which cannot satisfy all the rice panicle segmentation and extraction at the same time,the feature extraction and recognition method of rice panicle CT image based on Mask R-CNN(Mask Region-based Convolutional Neural Network)algorithm is studied;According to the rice mask features in the segmentation results,an algorithm for extracting rice in three-dimensional space is proposed to achieve characterization.Based on the principle of classification in the deep learning algorithm,the method of recognizing the blighted grain studied.For the eight different sets of rice panicles,the above two methods were used to characterize the traits.The experimental results prove that the characterization method of X-ray CT can use nondestructive detection method to identify and count rice,and the function of rice grain recognition can realize the characterization of rice seed setting rate.The Mask R-CNN algorithm used at the same time is versatile and can automatically realize the characterization and recognition of different rice panicles.