基于Sauvola与Otsu算法的秸秆覆盖率图像检测方法

Straw Coverage Detection Method Based on Sauvola and Otsu Segmentation Algorithm

针对自然光照下粗秸秆中空、细碎秸秆细节丢失导致秸秆覆盖率图像检测法精度低的问题,提出了一种基于Sauvola与Otsu算法相结合的秸秆覆盖率检测方法。首先对彩色分量空间距离灰度化后的图像采用Sauvola阈值分割来提取秸秆区域细节图像,然后对色差法灰度化后的图像采用Otsu阈值分割来解决秸秆区域中空问题,最后对不同阈值分割图像相加,计算其平均值,从而得到秸秆覆盖率的大小。田间试验结果表明,采用此方法对不同情况秸秆覆盖与不同地区的秸秆覆盖均具有较好的识别效果,不同情况秸秆覆盖率的最大平均误差约为1.9%,不同地区秸秆覆盖率的最大平均误差约为2.5%,有效地提高了秸秆辨识的精确度。

The low accuracy image detection in natural light, the hollow of coarse straw, and the loss of finely straw details, all these problems may cause low accuracy straw coverage detection. A straw coverage detection method, which based on Sauvola and Otsu segmentation algorithm, was proposed. Firstly, Sauvola image threshold segmentation was used to extract detail area of grayscale image which based on color component space distance, then Otsu threshold segmentation was used to solve problem of straw hollow area which based on color difference. Finally, different threshold segmentation image was added, and straw cover- agewas calculated. Field experiments results showed that better segmentation results and more accurate results of straw coverage had been achieved by proposed method. The maximum average error of different conditions of straw coverage was about 1.9%. The maximum average error of straw coverage in different regions was about 2. 5%. Proposed method could effectively improve precision of straw identification.