基于图像识别技术对玉米种子品种识别探究

Research on Recognition of Corn Seed Varieties Based on Image Recognition Technology

随着计算机技术的不断发展,图像采集和图像处理技术的应用越来越广,利用图像识别技术来识别农作物种子已经成为了可能。本研究基于图像识别技术,选取‘郑单958’、‘农大108’、‘鲁单981’、‘京科25’、‘京丰8号’共5个品种的玉米种子为研究对象,通过对玉米种子图像的采集、灰度化、中值滤波、阈值分割和形态学处理,提取了玉米种子的轮廓点数、面积、周长、圆形度、长短轴、直径、紧凑度和偏心率8个特征参数,利用人工神经网络方法对5个品质共250粒玉米种子的品种进行了识别测试。结果表明,‘郑单958’、‘农大108’、‘鲁单981’、‘京科25’、‘京丰8号’共5个品种玉米种子的识别正确率分别为92%、90%、92%、94%、94%,具有较高的识别正确率,为推动自动化检测玉米品种的手段提供了研究基础,具有重要的应用价值。

With the continuous development of computer technology, the application of image acquisition and image processing technology has become more and more widespread. It has become possible to use image recognition technology to identify crop seeds. Based on image recognition technology, this paper selected ’Zhengdan958’,’Nongda108’, ’Ludan981’, ’Jingke25’, and ’Jingfeng8’ as research objects. Through the collection, grayscale, median filtering, threshold segmentation and morphological processing of corn seed images, eight characteristic parameters,such as contour points, area, circumference, roundness, long and short axis, diameter, compactness and eccentricity,were extracted from corn seeds, and 250 corn seed varieties with five qualities were identified by artificial neural network method. The results showed that ’Zhengdan958’, ’Nongda108’, ’Ludan981’, ’Jingke25’ and ’Jingfeng8’ had high recognition accuracy of 92%, 90%, 92%, 94% and 94%, respectively. It provides the research foundation for promoting the means of automatic detection of maize varieties and has important application value.