果型综合评价系统中的退火演化和神经网络融合方法研究

Study on Annealing Evolutionary Algorithms and Neural Network Fusion Method Used in the Synthetic Evaluation of Fruit Shapes

摘要: 应用计算机视觉技术获取了国光、富士和黄元帅三个不同品种苹果的形状特征参数，采用退火演化算法及神经网络“融合”方法建立了分级合作的层次处理自动评价系统，实现了对果型的快速、准确综合判别。实验表明，该方法准确率达到９５％，并具有较好的鲁棒性和灵活性，能够满足多品种、大批量实时分级的需要。

Abstract: The shape feature parameters of three different breeds of apples were extracted with the computer vision technology. Using the Simulated Evolutionary Algorithm and the Neural Network Fusion Methods, a hierarchical automatic evaluation system was established, which can achieve highspeed and accurate synthetic judgement of the apple shapes . This method possesses a fairly good robustness and flexibility, and can satisfy the need of multi breed realtime fruit grading in big scale.