研究掺杂牛奶杂质的准确检测问题.牛奶中混入颜色激起相似的异性纤维,如化纤、毛发等杂质,由于杂质柔软、细小,杂质颜色像素与牛奶接近.传统的基于像素差异的杂质检测方法无法准确区分两者之间的细微差异,导致根据像素差异的阈值很难准确确定,造成牛奶杂质检测漏检率过高的问题.为了解决上述问题,提出一种基于多层细微像素模型的牛奶杂质检测算法.通过提取图像中的关键细节特点,运用边沿多层细化理论,对异样像素多层迭代验证,确定牛奶中杂质的边缘像素空间位置,从而实现牛奶杂质的检测,提高了检测的准确率.实验证明,上述方式能够准确检测出牛奶中的细微杂质,提高了检测的准确率,取得了令人满意的效果.%This is a study on accurate detection of fine impurities mingled in milk. The colour mingled in milk can bring out similar aniso-tropic fibres such as chemical fibre, hair and other impurities, which are soft and tiny with similar colour pixels to the milk. Traditional impurity detection methods based on pixels difference can not distinguish the gradation between the impurities and the milk, therefore the threshold based on pixels difference is vary hard to be precisely defined, which causes the problem of an exorbitant high level of omission ratio. In order to solve the above mentioned problems, this thesis puts forward a detection algorithm for fine impurities in milk based on multilayer subtle pixel model. Through collecting the key details from the images, and using edge multilayer refinement theory to validate the different pixels by multilayer iteration, the new algorithm identifies the spatial location of the edge pixels of impurities in milk, thereby realises the detection of the impurities in milk and improves the detection accuracy. According to the experiments, with the abovementioned method, we can detect the fine impurities in milk accurately. The improved accuracy procures satisfactory effect.
展开▼
