工业热力管道广泛应用于社会生产中，由于其特殊的高温工作环境，易引发爆炸、泄漏等突发事件，影响社会安全和经济效益。传统的视频检测技术不能直观、精确地显示出真实的管道内壁信息，需要对视频连续帧截取的管道探伤图像进行展开、拼接等处理。并且对于环境恶劣的热工管道而言，视频图像相关传统算法存在适应性不强、实时性较差等问题，因此研究此类算法、精准地实时监测管道内壁的探伤情况具有重要意义。

       针对采集图像的复杂背景，对噪声来源进行分析；利用图像去噪、增强等方法改善管道探伤图像质量。由于工业内窥镜采集的管道图像存在扭曲变形，且不适合人眼观察，因此提出一种图像展开方法。具体步骤为：首先采用自适应阈值分割和霍夫圆检测确定图像中心点并提取圆环区域；然后利用Bresenham直线扫描对圆环图像进行切向展开；最后结合双线性插值算法完成图像径向拉伸，提高展开图像的信息丰富度和清晰度。

       分析传统SURF匹配算法存在计算量大、配准精度低等问题，结合热工管实际情况，为充分提取管道内壁的探伤特征，提出一种改进SURF-RANSAC特征匹配算法。在特征点邻域内采用圆形模板提取32维描述符，实现特征描述符的降维并减少数据复杂度。为了减少误匹配对数以及人为设定阈值对匹配结果的影响，利用自适应阈值方法完成特征点粗匹配；采取特征向量构建余弦约束优化RANSAC算法实现匹配点对的提纯。为解决图像融合过程中出现“错位”和“重影”现象，采用改进的最佳缝合线算法完成探伤图像拼接实验。通过仿真结果分析，相比于传统算法，改进的匹配算法不仅保证了运算速度，且精度达到95%以上；本文的融合算法使得拼接图像过渡平滑、自然，对管道探伤图像具有良好的鲁棒性和适用性。

     Industrial thermal pipelines are widely used in social production. Due to their special high-temperature working environment, they are prone to explosions, leaks and other emergencies, which affect social safety and economic benefits. Traditional video detection technology cannot display the real pipeline inner wall information intuitively and accurately. It needs to expand and splice the pipeline flaw detection images of continuous frames of video. And for industrial pipelines with harsh environments, related processing algorithms have problems such as poor adaptability and poor real-time performance. Therefore, it is of great significance to study the image algorithm of the inner wall of the pipeline to accurately monitor the flaw detection of the inner wall of the pipeline in real time.

    In view of the complex background of the collected images, analyze the source of noise, use methods such as image denoising and enhancement to improve the quality of pipeline flaw detection images. Because the flaw detection images collected by industrial endoscopes are distorted and unsuitable for human observation, the Bresenham linear scanning method combined with the bilinear interpolation method is used to expand the collected flaw detection images. The specific methods of image expansion are as follows: First, use adaptive threshold segmentation algorithm and Hough circle detection to determine the center point of the image and extract the ring area. Then, use the Bresenham line scanning algorithm to expand the ring image tangentially. Finally, combine the bilinear interpolation algorithm, the image is stretched radially to improve the information richness and clarity of the expanded image.

    Analyzing the traditional SURF matching algorithm, which has the problems of large calculation amount and low registration accuracy. Combined with the actual situation of thermal pipes, an improved SURF-RANSAC algorithm is proposed to extract the flaw detection characteristics of the inner wall of the pipe. In the feature point neighborhood, a circular neighborhood is used instead of a rectangular neighborhood to extract 32-dimensional descriptors to achieve dimensionality reduction of feature descriptors and reduce data complexity. In order to reduce the impact of mismatch logarithm and artificial thresholds on the matching results. The adaptive threshold method is used to achieve the simple match. The feature vector is used to construct the cosine constraint optimization RANSAC algorithm to achieve the purification of matching point pairs. In order to solve the phenomenon of "dislocation" and "ghosting" in the fusion process, an improved best seam-line algorithm was used to complete the splicing experiment of flaw detection images. The simulation results show that compared with the traditional SURF algorithm, the improved matching algorithm not only guarantees the operation speed, but also has an accuracy of more than 95%. The fusion algorithm in this paper makes the stitched image transition smooth and natural, and has good robustness and applicability to the pipeline flaw detection image. 

[1]卢泓方, 吴晓南, 彭善碧. 国外天然气管道检测技术现状及启示[J]. 天然气工业, 2018, 38(02): 103-111.

[2]胡铁华, 郭静波. 油气管道内检测新技术与装备的开发及应用[J]. 天然气工业, 2019, 39(01): 118-124.

[3]Wen Hao Png, Horng Sheng Lin, Chang Hong Pua, Faidz Abd Rahman. Pipeline monitoring and leak detection using Loop integrated Mach Zehnder Interferometer optical fiber sensor [J]. Optical Fiber Technology, 2018: 221-225.

[4]I. A. Daniyan, O. A. Dahunsi, O. B. Oguntuase, O. L. Daniyan, K. Mpofu. Development of a Prototype Test Rig for Leak Detection in Pipelines[J]. Procedia CIRP, 2019: 524-529.

[5]张顺, 龚怡宏, 王进军. 深度卷积神经网络的发展及其在计算机视觉领域的应用[J].计算机学报, 2019, 42(03): 453-482.

[6]李华, 刘阳, 孟祥鹏, 张文学, 金俞鑫. 基于机器视觉的管道通径内检测技术[J]. 油气储运, 2019, 38(10): 1165-1169.

[7]Keiji Shibata et al. High-Quality Panoramic Image Generation Using Multiple PAL Images[J]. Electronics and Communications in Japan, 2014, 97(6) : 58-66.

[8]Shibata K, Araki S, Maeda K, et al. High-Quality Panoramic Image Generation Using Multiple PAL Images[J]. Electronics and Communications in Japan, 2014, 97(6): 58-66.

[9]Valérian Némesin, Stéphane Derrode. Quality-driven and real-time iris recognition from close-up eye videos[J]. Signal, Image and Video Processing, 2016, 10(1): 153-160.

[10]Heinz Hofbauer, Fernando Alonso-Fernandez, Josef Bigun, Andreas Uhl. Experimental analysis regarding the influence of iris segmentation on the recognition rate[J]. IET Biometrics, 2016, 5(3): 200-211.

[11]Reinoso O, Paya L, Gil A. A State-of-the-Art Review on Mapping and Localization of Mobile Robots Using Omnidirectional Vision Sensors[J]. Journal of Sensors, 2017:1-20.

[12]Paulson B, Lee S H, Kim Y, et al. Miniaturized omnidirectional flexible side-view endoscope for rapid monitoring of thin tubular biostructures[J]. Biomedical optics express, 2019, 10(5): 2264-2274.

[13]何四华, 邵晓方, 杨绍清, 吴春光. 双曲面折反射全景成像柱面展开图像质量分析[J].激光与红外, 2012, 42(02): 187-191.

[14]唐爱平. 管道内壁全景图像自适应展开算法研究[J]. 图学学报, 2015, 36(6): 973-977.

[15]吕德锐, 车德福, 贾庆仁, 李佳徐, 修春华. 钻孔图像的煤矿地质信息获取方法[J]. 测绘科学, 2016, 41(12): 288-301.

[16]曹茂永, 高炜, 邓兆鹏, 赵猛. 基于前视钻孔摄像提取全景图像的方法[J]. 郑州大学学报(理学版), 2017, 49(02): 79-83.

[17]彭斌, 祝志恒, 阳军生, 傅金阳, 何洪波. 基于全景展开图像的隧道衬砌渗漏水数字化识别方法研究[J]. 现代隧道技术, 2019, 56(03): 31-37+44.

[18]孙卫红, 廖艺真, 梁曼, 邵铁锋, 毕海忠. 蚕茧表面图像等效阶梯柱面展开[J]. 中国图象图形学报, 2020, 25(03): 498-506.

[19]夏丹, 周睿. 视差图像配准技术研究综述[J]. 计算机工程与应用, 2021, 57(02): 18-27.

[20]陈靓婷. 基于特征点的图像拼接算法的研究[D]. 湖南师范大学, 2017. 

[21]刘杰, 游品鸿, 占建斌, 刘金凤. 改进SIFT快速图像拼接和重影优化[J]. 光学精密工程, 2020, 28(09): 2076-2084.

[22]岳广, 孙文邦, 张星铭, 李铜哨. 航空面阵图像拼接的累积误差消除方法[J/OL]. 红外与激光工程: 2021, (22)03: 1-9.

[23]Seung Hyeon Cheon et al. An enhanced SURF algorithm based on new interest point detection procedure and fast computation technique[J]. Journal of Real-Time Image Processing, 2019, 16(4) : 1177-1187.

[24]Lowe D G. Distinctive  Image  Features  from  Scale-Invariant  Keypoints[J].  International Journal of Computer Vision, 2004, 60(2): 91-110. 

[25]Bay H, Tuytelaars T, Gool L V. SURF: Speeded Up Robust Features[J]. Computer Vision & Image Understanding, 2006, 110(3). 404-417.

[26]Zaragoza J, Chin T, Tran Q, et al. As-Projective-As-Possible Image Stitching with Moving DLT[J]. IEEE Trans Pattern Anal Mach Intell, 2014, 36(7):1285-1298.

[27]C.C. Lin, S.U. Pankanti, R.K. Natesan, et al.Adaptive as-natural-as-possible image stitching[C].In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015,1155-1163.

[28]Y. Ye, J. Shan, L. Bruzzone, et al. Robust Registration of Multimodal Remote Sensing Images Based on Structural Similarity[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017,55(5): 2941-2958.

[29]B. D. deVos, F. F. Berendsen, M.A.Viergever, etal. Adeep learning framework for unsupervised affine and deformable image registration[J]. Medical image analysis, 2019: 128-143.

[30]Sharma S K, Jain K. Image Stitching using AKAZE Features[J]. Journal of the Indian Society of Remote Sensing, 2020, 48(10): 1-13.

[31]董强, 刘晶红, 周前飞. 用于遥感图像拼接的改进SURF算法[J]. 吉林大学学报(工学版), 2017, 47(05): 1644-1652.

[32]沈跃, 朱嘉慧, 刘慧, 崔业民, 张炳南. 基于彩色和深度信息结合K-means聚类算法快速拼接植株图像[J]. 农业工程学报, 2018, 34(23): 134-141.

[33]徐弘祯, 李世超, 季宇寒, 曹如月, 张漫, 李寒. 基于特征点匹配的全景相机图像拼接方法研究[J]. 农业机械学报, 2019, 50(S1): 150-158.

[34]毛峡, 李亚豪. 光伏阵列故障检测中的无人机红外图像拼接[J]. 太阳能学报, 2020, 41(03): 262-269.

[35]杨赞伟, 郑亮亮, 曲宏松, 吴勇. 联合均值滤波与泊松核双边滤波降噪算法研究[J].计算机仿真, 2020, 37(09): 460-463+468.

[36]Justin Varghese, Saudia Subhash,Kamalraj Subramaniam, Kuttaiyur Palaniswamy Sridhar. Adaptive Gaussian notch filter for removing periodic noise from digital images[J]. IET Image Processing, 2020, 14(8): 1529-1538.

[37]SheXiang Jiang, Ri-Gui Zhou, WenWen Hu, YaoChong Li. Improved Quantum Image Median Filtering in the Spatial Domain[J]. International Journal of Theoretical Physics, 2019, 58(7):2115-2133.

[38]强彦, 张晓慧. 车载手势识别中基于小波变换和双边滤波的图像去噪方法[J]. 北京理工大学学报, 2017, 37(04):376-380.

[39]曹海杰, 刘宁, 许吉, 彭杰, 刘宇昕. 红外图像自适应逆直方图增强技术[J]. 红外与激光工程, 2020, 49(04): 256-262.

[40]张立亚, 郝博南, 孟庆勇, 温良, 吴文臻. 基于HSV空间改进融合Retinex算法的井下图像增强方法[J]. 煤炭学报, 2020, 45(S1): 532-540.

[41]Jamshid Soltani-Nabipour, Abdollah Khorshidi, Behrooz Noorian. Lung tumor segmentation using improved region growing algorithm[J]. Nuclear Engineering and Technology, 2020, 52(10): 356-369.

[42]段锁林, 殷聪聪, 李大伟. 改进的自适应Canny边缘检测算法[J]. 计算机工程与设计, 2018, 39(06): 1645-1652.

[43]Bo Chen et al. Detect concrete cracks based on OTSU algorithm with differential image[J]. The Journal of Engineering, 2019, 2019(23): 9088-9091.

[44]许益成, 谭文安, 陈丽婷. 基于Hough变换检测前大灯的夜间车辆检测[J]. 控制工程, 2018, 25(08): 1360-1363.

[45]陶宏江, 金龙旭. 基于各向异性插值的全景图像径向展开算法[J]. 液晶与显示, 2014, 29(03): 435-440.

[46]王宝仁, 李金松, 吕国娜, 董洁. 基于Bresenham算法的快速直线脉冲增量插补算法[J]. 机床与液压, 2018, 46(20): 155-158.

[47]Yang Jungang et al. High-precision refocusing method with one interpolation for camera array images[J]. IET Image Processing, 2021, 14(15): 3899-3908.

[48]范清兰, 张云峰, 包芳勋, 沈晓红, 姚勋祥.基于NSCT的区域自适应图像插值算法[J].计算机研究与发展, 2018, 55(03): 629-642.

[49]郝勇, 耿佩, 温钦华, 吴文辉. 滚动轴承保持架缺陷的图像处理及模式识别方法研究[J]. 仪器仪表学报, 2019, 40(09): 162-169.

[50]Seung Hyeon Cheon et al. An enhanced SURF algorithm based on new interest point detection procedure and fast computation technique[J]. Journal of Real-Time Image Processing, 2019, 16(4) : 1177-1187.

[51]唐颖复, 王忠静, 张子雄. 基于改进SIFT和SURF算法的沙丘图像配准[J]. 清华大学学报(自然科学版), 2021, 61(02): 161-169.

[52]柳长安, 艾壮, 赵丽娟. 基于网格运动统计的自适应图像特征匹配算法[J]. 华中科技大学学报(自然科学版), 2020, 48(01): 37-40+54.

[53]丁小奇, 李健, 胡雅婷等. 基于改进SURF算法无人机影像特征匹配的研究[J]. 中国农机化学报, 2020, 41(02): 147-154.

[54]Tao Song et al. Research on image feature matching algorithm based on feature optical flow and corner feature[J]. The Journal of Engineering, 2020(13) : 529-534.

[55]何玉青, 王雪, 王思远, 刘明奇, 诸加丹, 金伟其. Modified SIFT descriptor and key-point matching for fast and robust image mosaic[J]. Journal of Beijing Institute of Technology, 2016, 25(04): 562-570.

[56]Xiaobin, Zhongwen, Fei, et al. Improved calibration method based on the RANSAC approach and an improved gray centroid method for a laser-line-based structured light system.[J]. Applied optics, 2019, 58(35): 9603-9613.

[57]产叶, 林胡新平. 基于FAST和SURF的特征点快速匹配算法[J]. 计算机工程与设计, 2019, 40(12): 3500-3504.

[58]郭方锐. 基于管道内部探伤图像拼接融合问题研究[D]. 西安科技大学, 2020.

[59]Zhiqin Zhu et al. A novel multi-modality image fusion method based on image decomposition and sparse representation[J]. Information Sciences, 2018(432) : 516-529.

[60]黄樟钦, 母召, 岑陈, 高寒. 一种多路视频的360度实时拼接技术[J]. 西安电子科技大学学报, 2020, 47(02): 1-8+31.