基于改进Hough变换的农田作物行快速检测算法

Rapid Crop-row Detection Based on Improved Hough Transformation

选取苗期农田作为研究对象,采集了包含行栽作物和土壤背景的农田图像,针对现有作物行定位方法易受外界干扰和处理速度较慢的不足,提出将投影法和直接Hough变换法相结合检测作物行的算法。采用2G-R-B法和OTSU法将图像二值化,通过快速中值滤波算法去除噪声,再利用垂直直方图投影将图像进行水平条划分获取作物垄平均定位点,最后通过Hough变换检测垄定位点,得到作物行中心线。试验结果表明:基于垂直直方图投影的Hough变换检测作物行中心线的算法在保证高定位精度的同时,算法处理速度比直接Hough变换检测法提高了3倍,得到的定位基准线能代表作物行走向。

A rapid detection method of crop-row based on improved Hough transformation was suggested to detect the position of crop-rows after analyzing the characteristics of field crop planting. The true color images containing crop-row and soil background were captured as the research objects. Since the histogram projection is easily disturbed by external factors and the Hough transformation has a low speed, a combination of these two algorithms was developed. First, a true color image of field crops was processed with 2G -R -B gray and OTSU to separate plants from soil. Second, the noise was eliminated by fast median filtering algorithm, and then the image was divided into ten horizontal strips to get the anchor points according to the geometric distribution of the crop-rows. Finally, the centerline of crop rows was obtained by Hough transformation using those points. Test results show that the processing speed of the improved Hough transformation method is four times faster than that of original Hough transformation with the similar positioning accuracy, and the centerline detected can highly match the crop-row.