基于计算机视觉的库尔勒香梨脱萼果识别研究

摘要

[目的]针对库尔勒香梨的图像,对香梨脱萼果和宿萼果进行分类及识别研究.[方法]利用Matlab软件进行图像数据获取、灰度化、中值滤波,利用动态阀值分割二值化图像；为弥补二值化后图像中出现的孔洞,基于数学形态学算子填充孔洞.设计神经网络参数,然后利用试验样本对BP神经网络进行训练和测试.[结果]所提出的方法能够获得很高的正确识别率,能够有效地将脱萼果识别出来.[结论]用BP神经网络来分类识别香梨脱萼果可以达到较好的效果.%[Objective] Aiming to image of Korla fragrant pear,the identification research and classification of the calyx fruit and persistent calyx fruit were studied.[Method] Firstly,some pre-processing for fruit image such as data acquisition,the graying arithmetic,median filtering algorithm,were carried out by Matlab software,and binary image was cut up by dynamic threshold; In order to make up the broken edges and holes in the binary image,mathematical morphology operator was employed to fill holes.Design of network parameters,and then use the experimental samples to train and test the BP neural network.[Result] The proposed method obtained a high accurate identification rate and efficiently identify Calyx leaving from fruit.[Conclusion] Using BP neural network to identify and classify the Calyx leaving from fruit of Korla fragrant pear obtains good effects.