机器视觉在稻米粒型检测中的应用

Detection of Rice Shape Based on Machine Vision

针对目前在稻米粒型检测中依靠人工费时、费力、精度难于控制等问题,利用机器视觉技术建立了稻米粒型实时检测系统。提出了以改进的最大类间方差法来自动确定图像分割阈值,采用开运算去除图像中的噪声,使用最小外接矩形方法计算稻米粒型。与直尺法、微粒子计法比较,该系统具有精度高、鲁棒性好、处理快速的特点,能满足稻米粒型检测的实际需求。

Artificial detection of rice kernels takes a lot of time and energy and its accuracy is hard to be controlled. To seek a convenient method for detection of rice shape, a real-time vision-based detection approach was developed, which used an improved Ostu algorithm to segmentate the image, eliminated the noise of image by open operation, and calculated rice shape characteristic values by the minimum enclosing rectangle method. Comparing with the methods by ruler and microparticle, the new detection method was of higher accuracy, stronger robustness and better efficiency and could be applied to practical measurement.