Research And Application Of Key Technologies For High-Throughput Crop Phenotype Detection

With the advent of technology, the development of information science and life sciences have made remarkable achievements. The intersection of information science and life science has become a new type of science field with rapid development. One of the key problems in this field is the mutual relationship between phenotypic information and genetic information in crop breeding. This problem is an urgent problem that 21st century breeders need to solve. Effective solutions to this problem can help breeding scientists predict crop phenotypes based on the genetic information of breeding materials. However, this is a very difficult task. Especially, thousands of complex genes and ever-changing plant growth environments control the phenotypic information of breeding materials.High-throughput information acquisition and large-scale information processing are important ways to solve this problem. The correlation mining relying on large-scale genotype data and phenotypic data (the observed phenotype of the plant) will eventually be able to predict the variety of performance. At present, the progress of the new generation of sequencing technology can provide a wealth of genomic information, but the scientific and efficient ability to measure crop phenotype develops slowly. This situation directly leads to data imbalance between genotype and phenotypic information.The lack of high-throughput phenotypic information data creates a bottleneck in hereditary association analysis and genome selection.The aspects of basic theory and application on large-scale data processing and information acquisition on high-throughput phenotypic are researched in this paper.Finally, the solution of high-throughput bio-breeding network platform is given.1?Anew rough fuzzy set model is formed based on inclusion degree of fuzzy set theory applied to the rough fuzzy set. Batch dynamically incremental clustering algorithm is proposed based on new rough fuzzy set model. In this algorithm, new incremental data is added to the initial database in form of cluster rather than one by one, so this algorithm greatly improves the efficiency of building large data warehouse and achieves the purpose of large-scale data efficient processing.2?Aiming at the problem of quality classification of germplasm resources in the process of germplasm database construction, k-means clustering algorithm based on stacked sparse autoencoder is proposed to cluster the data, and the clustering results are compared to known quality resources which have had category labeling, so as to achieve the purpose of classification of breeding data quality. Compared to the traditional K-means clustering algorithm, the stacked sparse autoencoder k-means clustering algorithm is used to extract the key data feature, and the dimension of sample is reduced gradually. The mixed feature data is constructed and used as the initial center of the k-means clustering algorithm. The sensitivity of the initial center in the k-means clustering algorithm is overcome. The experimental results show that the accuracy of the clustering algorithm is obviously improved.3?Based on the high-throughput crop phenotypic information acquisition task,phenotypic image data is obtained by the fixed-grid image acquisition matrix device equipment and the unmanned aerial vehicle (UAV) are equipped with visible, multi-spectral and thermal imaging sensors, then breeding area orthophoto images are obtained by using image correction and orthophoto generating technology. The method of calculating the leaf area index (LAI) is obtained innovatively by using the machine vision method of crop canopy green coverage extraction and the measured height of the crop based on trigonometric function. High-throughput, low-cost and efficient crop phenotype information is obtained, including plant height, emergence rate, lodging,growth potential, etc.4?The basic construction content and the current progress of the high-throughput bio-breeding network platform from the perspective of platform construction are studied. The four major platforms which are the basic management system of breeding information, the research and development of breeding agricultural sensors, crop field phenotypic visual measurement device development and breeding of phenotypic rapid detection of mobile robot are illustrated. The main functions and technical parameters of high - throughput breeding information management system and breeding phenotype for rapid detection of mobile robots are highlighted.