Research on Rice Ear Segmentation Technology in Datian Community Based on Deep Learning and Superpixel
Rice ear segmentation under different growth stages from top camera angle Recently, Huazhong Agricultural University, Huazhong University of Science and Crop phenotype joint research team entitled on "Plant Methods" magazine: Panicle-SEG: A robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimization article, For the first time, this paper combines deep learning technology and superpixel clustering algorithm to achieve high robustness of rice paddy segmentation in Datian Community. The experimental results show that the algorithm can be applied to different light environments, different rice varieties, different The rice growth period and the ability to process rice images of different imaging scenes and angles. Rice is the main food crop of most of the world's population. Rice ear is an important agronomic organ, which is closely related to rice yield, disease detection, and growth period judgment. The Daejeon environment is the real growing environment for rice growth, so how to solve the problem of how to solve the problem is how to solve the problem. Usually the environment in the field is very complicated. The changing illumination, the disturbance of the wind field, the influence of the weather, the reflection of water and the phenomenon of overlapping and blocking of stems and leaves all bring difficulties to the study of rice ear. At the same time, due to the differences in rice varieties, the rice ears have great differences in color, texture, shape, size and posture. Different photographing angles will make the rice ear images very different. These problems are all given to rice paddy fields. The segmentation of the ear brings no small challenge. Therefore, the general segmentation algorithm is no longer suitable, and a robust array of rice paddy segmentation algorithms must be designed. In recent years, deep learning methods have been widely used in many fields. Due to its powerful automatic feature extraction, complex model construction and image processing capabilities, it is very suitable for new problems in biological image processing. In this study, rice in the field was planted in a community manner, and 20 rice varieties of the same variety were planted in each plot (90 × 90 cm 2 ), and the cells were separated from the plot by a protective line. Two photos are taken for each cell, that is, the rice plots are imaged from top view and top view, respectively. The acquired top view image and top view image are divided into a training set, a verification set, and a test set. For the training samples (including the training set and the verification set), the simple trained iterative clustering (SLIC) method is used to extract the final trained image blocks, and the convolutional neural network is used to train the classification model. The test samples are then classified and tested using the trained CNN model. The segmentation image of the final rice ear is obtained after optimization of the rough segmentation result. Based on the deep learning method, this study only uses the ordinary monocular RGB camera to obtain the rice image of Daejeon, and proposes a new idea to convert the rice ear division task into a classification task. The segmentation of the ear also ensures the integrity of the segmentation edge. At the same time, for complex and varied scenes and environments (such as different illumination, etc.), the algorithm has good segmentation robustness. For images with different imaging angles, The algorithm can also be applied, and the average segmentation accuracy of the test samples is above 80%. Panicle-SEG algorithm flow chart The work was supported by the National Program on High Technology Development (2013AA102403), the National Key Research and Development Program (2016YFD0100101-18), and the National Natural Science Foundation. Of China (31701317, 31770397)), Hubei Provincial Natural Science Foundation of China (2017 CFB208), Scientific Conditions and Resources Research Program of Hubei Province of China (2015BCE044) And funding from the Fundamental Research Funds for the Central Universities (2662015QC006, 2662017PY058). Source: Xiong Xiong, Lingfeng Duan, Lingbo Liu, Wanneng Yang and Qian Li (2017), Panicle-SEG: A robust image segmentation method for rice panicles in the field based on deep learning and superpixel optimization, Plant Methods, 13:104. We are a professional Chinese manufacturer of Plants Peptides; we supply various products of Plants Peptides, and can providing product images and basic parameters with each Oligopeptides Powder and Polypeptide Powder. We have the perfect after-sales service and technical support. Please do not hesitate to contact us, Look forward to your cooperation! Oligopeptides Powder,Polypeptide Powder,Hydrolyzed Vegetable Protein,Soybean Peptide,Sunflower Plate Peptide Xi'an Quanao Biotech Co., Ltd. , https://www.quanaobiotech.com