基于近红外图像技术的水果轻微损伤检测

Detection slight bruised fruit base on near infrared imaging

水果轻微损伤检测对提高水果档次、延长货架期有着重要的经济价值,如果不把轻微损伤水果检测出来加以剔除,随着贮藏时间的延长就很容易霉烂,并影响其他正常水果。提出了在900—1100nm处利用近红外图像处理技术对水果轻微损伤进行检测的方法。采用四周扫描法去除背景,针对无明显双峰图像分割的难点开发出了一种新的分割算法对轻微损伤进行分割,最后对轻微损伤的个数和面积进行了识别计算并对面积的畸变校正进行了探讨。实验结果表明,该方法对有轻微损伤的图像识别正确率达91.3%,为水果的后续在线检测奠定了基础。

The detection of slight bruised fruit was important economic worthiness to enhance grade of fruit and prolong shelf life. With the time going they were very easy rot and contaminated other good fruit, if the slight bruised fruit were not eliminated. In this paper using the method of near infrared （NIR） imaging processing to detect slight bruised fruit were presented between 700 and 1100nm. The background was removed from the imaging by around-scanning method. And adopt a new method to segment the slight bruised imaging which was not evidence double apex. At last the number and area of slight bruised was computed in every picture and revising aberrant area of them was discussed. For the image with slight bruised the result of experiment indicate that identify ratio of this method is 91.3%, which is the foundation to subsequent on-line detection of fruit.