College of Mechanical Engineering, Xi'an University of Science and Technology
Shaanxi Key Laboratory of Mine Electromechanical Equipment Intelligent Detection and Control, Xi'an University of Science and Technology

煤矿井下掘进工作面高粉尘、低照度的恶劣环境导致图像信噪比较低，且有效特征点数量严重减少，处理后的图像存在较大色差和噪声，在使用最佳缝合线算法进行图像拼接时出现细节错位、缝合线处过渡不自然或拼接痕迹明显的现象。针对上述问题，提出了一种基于改进最佳缝合线的矿井图像拼接方法。首先，对原始图像进行HSV空间变换，采用改进的Retinex算法对亮度分量进行增强，利用双边滤波函数代替中心环绕函数，以解决亮度差异大处产生的光晕问题，通过增强算法有效提高特征点提取数量。然后，采用SIFT算法提取特征点，并以余弦距离作为匹配度指标；引入像素余弦相似度作为约束项，并采用形态学操作对颜色差异强度进行改进，利用动态规划法对最佳缝合线进行搜索，以避免图像拼接处的错位现象。最后，结合渐入渐出融合算法，使图像过渡平滑，实现煤矿井下掘进工作面的图像融合。模拟井下实际工况环境进行实验验证，结果表明：基于改进最佳缝合线的矿井图像拼接方法与传统最佳缝合线算法相比，避免了颜色差异和噪声引起的错位拼接现象，拼接缝处的图像过渡更加自然，避免了“鬼影”和明显拼接缝的产生，且图像平均梯度提高2.38%，拼接时间提高32.5%，使得融合区域更加平滑自然，提高了拼接质量。

The harsh environment of high dust and low lighting in the coal mine underground excavation working face results in low signal-to-noise ratio of the image, and a serious reduction in the number of effective feature points. The processed image has significant color difference and noise. When using the best seam-line algorithm for image stitching, there are problems such as fine section misalignment, unnatural transitions at the seam line, or obvious stitching traces. In order to solve the above problems, a mine image stitching method based on improved best seam-line is proposed. Firstly, the original image is subjected to HSV spatial transformation, and an improved Retinex algorithm is used for enhancement on the luminance component. Bilateral filtering is used instead of the center surround function to solve the halo problem caused by large brightness differences. The number of feature points extracted is effectively increased through the enhancement algorithm. Secondly, the SIFT algorithm is used to extract feature points, and cosine distance is used as the matching degree indicator. The method introduces pixel cosine similarity as a constraint, and uses morphological operations to improve color difference intensity, uses dynamic programming to search for the best seam-line to avoid misalignment at image stitching. Finally, combined with the gradual in and out algorithm, the image transition is smooth to achieve image fusion of the underground excavation working face. Experimental verification is conducted by simulating the actual working environment underground. The results show that the mine image stitching method based on the improved best seam-line avoids the phenomenon of misalignment stitching caused by color differences and noise compared to the traditional best seam-line algorithm. The image transition at the stitching seam is more natural, avoiding the generation of 'ghosts' and obvious stitching seams. The average gradient of the image is increased by about 2.38%, and the stitching time is increased by about 32.5%, making the fusion area smoother and more natural, improving the stitching quality.

国家自然科学基金青年项目（52104166）；陕煤联合基金项目（2021JLM-03）；中国博士后科学基金面上项目（2022MD723826）；陕西省重点研发计划项目（2023-YBGY-063）。

张旭辉，王悦，杨文娟，等. 基于改进最佳缝合线的矿井图像拼接方法[J]. 工矿自动化，2024，50（4）：9-17.

ZHANG Xuhui, WANG Yue, YANG Wenjuan, et al. A mine image stitching method based on improved best seam-line[J]. Journal of Mine Automation，2024，50（4）：9-17.