基于改进Canny算子的煤矿井下物体图像边缘检测方法研究

针对井下低照度环境下采煤机截割部边缘检测任务中存在的边缘缺失、细节模糊等问题，提出一种基于分数阶微分的边缘检测Lif算法。首先采用更大的检测模板尺寸，根据Grünwald-Letnikov分数阶定义构造最初的分数阶掩膜算子；然后根据Pascal三角形理论确定掩膜算子上各位置的权重系数，并将掩膜算子扩展到4个不同方向；最后将得到的掩膜算子与图像进行卷积，利用图像的局部特征信息对每个方向的微分结果进行后处理。结果表明：(1) 在进行多个不同场景的井下低照度图像上的实验时，Lif算法可以更全面地获取图像中不同方向上的边缘信息，在处理低照度图像时具备更强的抗噪性能，并且提取的边缘线条比其余边缘检测算法更加清晰、完整，保留了更多的纹理细节信息。(2) 在客观指标评价的对比上，与基于分数阶灰色系统模型的边缘检测算法以及改进的分数阶Sobel边缘检测算法相比，Lif算法在Entropy指标上分别提高了43%、11%，AG指标上分别提高了23%、23%，SSIM指标上分别提高了152%、6%。表明Lif算法在进行采煤机截割部的边缘检测任务时更具优势，研究对井下设备工作运行时的安全性和可靠性提升具有重要意义。

Image edge detection method of underground objects based on improved Canny operator

In this paper, a fractional differentiation-based edge detection algorithm named Lif is proposed to address the edge detection problems of missing edge and fuzzy details in the cutting unit of shearers working in low-light underground environments. First, the initial fractional mask operator was built using a larger detection template according to the Grünwald-Letnikov definition of the fractional derivative. Then, the weight coefficients at various positions of the mask operator were determined according to the theory of Pascal’s triangle, and the mask operator was extended to four different directions. Finally, the mask operator was convolved with the image, and the local feature information of the image was used to process the differentiation results in all directions. The results show that (1) the Lif algorithm can obtain the edge information in different directions in the image more comprehensively when conducting experiments on low-light images in different scenarios, has stronger noise resistance when processing low-light images, and can retain more textural details; the edge lines extracted by this algorithm are clearer and more complete than those extracted by other edge detection algorithms. (2) Compared with the edge detection algorithm based on the fractional grey system model and the improved fractional Sobel edge detection algorithm, the Lif algorithm performs better than them by 43% and 11% in terms of Entropy, by 23% and 23% in terms of AG, and by 152% and 6% in terms of SSIM, indicating that the Lif algorithm has advantages when detecting the edge for cutting unit of shearers. This study is of great significance for improving the operational safety and reliability of underground equipment such as shearers.