Research And Application Of Assistant Decision Key Technology For Plant Breeding

For large-scale breeding, facing the current challenge of developing technologies of obtaining high-throughput trait data, efficiently designing and planning of experiments, labor intensively rating breeding data, under the goal of increasing plant breeding efficiency, this paper researches to quickly obtain crop breeding information, assistant filtering analysis and whole process management skills. As the convenient, highly efficient and full-scale informational and technological service skills, the information management and decision system of plant breeding is also developed. Focus on the work in the following four areas:(1) Phenotype feature selection for plant breeding evaluation based on ranking relevanceTraditional breeding evaluation methods focus on information of crop traits, while ignoring the previous evaluation results. In order to enhance the efficiency of material evaluation under the condition of large-scale breeding, the comprehensive evaluation of crop traits is integrated into the breeding evaluation, and a kind of method of phenotype feature selection for crop breeding evaluation based on ranking relevance is proposed. Firstly, the training sample set and the candidate feature set were selected from breeding data, and calculated the correlation between the phenotype feature and the results of evaluation and the similarity of the agronomic traits. Then, considering the characteristics of the maximum correlation coefficient and the minimum similarity, a model of phenotype feature selection for breeding evaluation based on ranking relevance was constructed. This model can be used for different breeding objectives focused on collection of characters and the validity of model was verified by using the three kinds of soybean identification trial. This model also can be used as the pre process of breeding evaluation method to determine the weights of the traits accurately.(2) Plant breeding evaluation with rank entropy-based decision treeWith the rapid development of automatic machine and software, plant breeding has entered a stage of large-scale, in which much work needs to be done by software automatically. According to the characteristics, plant breeding evaluation can be considered as an ordinal classification task. Therefore, this paper proposes a rank entropy-based decision tree algorithm to do this classification. The algorithm uses historical trait performance and evaluation data to construct decision trees based on rank entropy. The constructed decision trees are then used to generate evaluations for future cultivars with their trait performances. To demonstrate the effectiveness of proposed algorithm, experiments are carried out on three breeding datasets of soybean with early-maturity, medium-maturity, and green soybean as breeding object, respectively. This work can assist breeders to complete a large number of basic work, and enhance the efficiency of plant breeding.(3) Assisted identification methods of disease and insect pest in plant resistance breedingIn the process of resistance identification of breeding materials, the workload of data acquisition is large, and the error rate of artificial factors is higher. In this paper, a novel method to segment the whitefly in the field environment by the Discrete Cosine Transformation (DCT) and region growing methods is proposed. The experiments were conducted on whitefly images by comparison with the methods based on thresholding and Gaussian Mixture Model (GMM). Experimental results show that our proposed method can effectively separate pests apart from normal part of leaves and background. Our method provides higher precision as well as the accurate and closed boundaries, which is beneficial in the processing of whitefly images.To improve the generalization and ccuracy of pests counting algorithm, a novel counting algorithm for whiteflies based on k-means clustering and ellipse fitting method was proposed. It combined k-means clustering algorithm with ellipse fitting and automatically learned the features of whiteflies and background to segment and count whitefly images accurately. The proposed method is implemented on mobile smart devices and tested with field experiments. The experimental results show that the proposed method has good recognition performance with high efficiency.(4) Information management and assistant decision system for plant breedingConsidering the characteristics of large number of breeding materials and wide experimental bases, and intergrating the methods of breeding evaluate proposed in this paper, software is developed for information management and assistant decision of plant breeding. The software plays a big role in germplasm management, parental combination, breeding, trial design, data analysis and materials evaluation, and has been applied in breeding enterprises, breeding scientific research institute, testing consortium, experiment station, seed extension station, et al. The methods and software are practical and valueable as were shown in the application cases.