种蛋外形的计算机视觉识别研究

提出一种利用计算机视觉对种蛋进行筛选的方法。蛋形指数有两个指标：质量与形状，将质量的测量转为对尺寸的测量，采用第一个神经网络检测种蛋的主要特征点，并根据视觉测量原理得到其尺寸大小。利用带小波变换对原始的边缘形状参数进行特征提取，第二个神经网络以此为输入来识别种蛋外形的规则性。试验表明：该方法与人工筛选相比较，其一致性可达到93%。

Shape Identification of Fertile Eggs by Using Computer Vision

An approach of grading fertile eggs based on computer vision was presented. The weight and contour regularity were used as identification parameters of the shape of fertile eggs. According to computer vision measurement principle, the first artificial neural network(ANN) was used to detect dominant points of fertile eggs, such as size of fertile eggs obtained, then m-bands wavelet transform was used for features extraction of origin edge data. The second ANN identif regularity of shape. Results from experiments confirmed that the consistency identified by computer vision grading can reach over 93% in comparison with manual grading.