基于Hough变换的成熟草莓识别技术

Ripened Strawberry Recognition Based on Hough Transform

基于图像区域特征来识别被遮掩、重叠或紧靠的草莓果实非常困难,提出一种基于Hough变换的成熟草莓识别方法。先对Lab彩色模型下a通道图像进行分割,利用提取的草莓轮廓信息,根据草莓轮廓的数学模型进行Hough变换,实现成熟草莓的识别。为减少运算量,在Hough变换之前,先进行区域标记,获取有效图像信息区域。草莓轮廓信息提取和Hough变换在各个有效区域中进行,由于参数空间大大压缩,运算量也得到减少。试验表明:当成熟草莓轮廓信息丢失小于1/2时,无论单个分离的成熟草莓,还是被遮掩、重叠或紧靠的成熟草莓,皆有很好的识别效果,识别平均相对偏差为4.8%,能满足草莓采摘机器人对目标识别精度的要求。

Strawberry recognition is a very important part in strawberry harvesting robot. The method being used is based on the features （such as area, compact, etc. ） of the image area. It is effective for single strawberry recognition, but unsatisfactory for complex situation, in which the strawberry may be hidden by other objects or covered each other. In this paper, we put forward a new method for strawberry recognition based on Hough transform. It labels the region of the image after segmented, gets the minimum enclosing rectangle （MER） and recognizes it based on Hough transform using the contour of strawberry in the MER. The experimental results show that the strawberry recognition method based on Hough transform is effective both for single strawberry and complex situation when the strawberry contour loss is less than 1/2. The recognition warp on average is 4.8% and can meet the requirement of strawberry harvesting robot.