Research And Design Of Crop Germplasm Resources Quality Comprehensive Evaluation System

The crop germplasm resources are the material bases of the plant breeding and the biological technology. They are strategic resources of living and developing for human beings, and the guarantees for the national food security and the agricultural sustainable development. At present, people are to pay more attention to the collection, arrangement, evaluation and utilization of crop gennplasm resourcesDomestic and foreign researchers have already completed many germplasm resources information platform. But the resources data in the majority systems and database are primitive. Majority of the crop gennplasm resources data can’t be used for the shortage of comprehensive evaluation. Therefore researches and discussions of comprehensive evaluation methods and techniques are more and more important for mining some association rules between the performance character and gene character, providing the accurate breeding decision-making.This research improved on the Project "Crop Special Genetic Resources standardization finishing, integration and sharing platform" subsidized by Chinese Ministry of Education. Neural network has the capacity to learn, self-organize and self-adapt. In this paper a comprehensive evaluation model based on BP neural network which is built up for rice gennplasm resources data. Aimed at to enhance the effect, Liebenberg-Marquardt algorithm is adopted to optimize the BPNN. Based on the evaluation model, this paper proposes the WebGIS-based rice germplasm resources evaluation system. Experimental results and resources distribution maps prove to the LM-BP evaluation model is validity. The main works are as follows:Firstly, this paper analysis the comprehensive evaluation methods based on the basic theory of neural network technology. According to the rules of registration for crop variety, a comprehensive evaluation system is built up for rice gennplasm resources.Secondly, research and discuss the structure, activation function and superior generating ability of BPNN through consulting correlation references and massive experiments. Considering influence of initial values, poor stability, slow convergent rate and the local minimum problem, the BPNN based on Levenberg-Marquardt algorithm is combined with the evaluation model. Case study shows that the proposed model is superior to traditional one.At last, a WebGIS-based rice germplasm resources comprehensive evaluation system was set up based on LM-BP neural network. It contains basic statistical analysis, comprehensive evaluation, attribute queries and maps demonstration. The validity and security of evaluating results on practical sampling data are analyzed and proven by the resources distribution maps.