Study On The Method Of Grain Detection In Rice Panicle Based On X - Ray Digital Imaging

Rice is one of the main food in China, with the increase of population, the demand of rice increased gradually. It is the central problem of rice breeding to cultivate fine varieties and increase rice yield. Ear of rice is the ultimate expression of rice yield, and the relationship between its characters and yield has always been the focus of breeding. Rice panicle traits have been an important reference information when evaluating the rice quality. The characters of ear of rice include seed setting rate, grain weight, and grain length etc. Seed setting rate affected by empty-grain and grain weight affected by grain length are important factor effecting the yield.With the improvement of fabrication process of X-ray digital imaging equipment.After analyzing existing methods of yield of rice, the paper presents a method of detecting rice grain parameter based on the X-ray digital imaging technology, the method can achieve the rice panicle total grain number, filled grain number, empty unfilled grain number and grain length by detecting in body.Image processing technologies are used to analysis the X-ray image of rice in this paper, and image denoising, segmentation and morphological processing are carried on the images. The paper adopts threshold segmentation algorithm based on gray level histogram which can improve the accuracy of parameter extraction of rice grain. At the same time, on the basis of analysis of radial distortion in X-ray image, algorithm for real time measurement of rice grain length is improved. In order to validate stability and accuracy of algorithm, a real-time detection platform for rice spike detection is built. The main contents are as follows:Firstly, introduces the research background and the research status, analyzes the existing problems and defects of rice grain parameters analysis method. Secondly, elaborates the principle of X imaging, the structure of imaging system, and the time sequence of mechanical motion of real-time detection platform is introduced emphatically. Thirdly the image processing algorithms are introduced, and the improved algorithm based on Gauss smoothing, multi threshold segmentation and morphological processing is proposed by analyzing different image processing algorithms. Finally, lossless acquisition of grain parameters was realized on the analysis of image data achieving through the actual running detection system.