| 28 | 0 | 143 |
| 下载次数 | 被引频次 | 阅读次数 |
目的:针对图像匹配技术在实际应用中面临的多姿态、多目标场景适配、匹配速度快等需求，提出了一种基于零均值归一化互相关(zero-mean normalized cross-correlation, ZNCC)的灰度图像匹配算法，以提升复杂场景下的匹配性能。方法:首先依据预设角度范围与金字塔层数，在金字塔顶层构建初始模板库；采用改进积分图与掩模图融合的方法，实现非矩形感兴趣区域(region of interest, ROI)的快速匹配；通过非极大值抑制技术消除匹配框紧邻引发的误匹配问题；以顶层模板库的初步匹配结果为基础，将其映射至金字塔下一层级，动态生成缩小后的搜索空间与对应模板，经逐层迭代直至金字塔底层，完成精确匹配。结果:在单目标匹配任务中(待匹配图像尺寸为1 920×1 278),当匹配交并比大于95%时，算法平均耗时仅7.52 ms;在包含多目标及旋转变换的复杂场景中(待匹配图像分辨率为646×482),满足相同匹配交并比要求的前提下，算法平均耗时52.91 ms。结论:所提算法能够在多姿态、多目标及旋转变换等复杂场景下，实现快速且精准的灰度图像匹配，具备良好的实际应用价值。
Abstract：Aims: For the requirements of in image matching applications, such as varying poses, multiple targets, and high-speed, a grayscale image matching method based on ZNCC was designed. Methods: According to the preset angle range and the number of pyramid layers, an initial template library was constructed at the top layer of the pyramid. A method combining improved integral images and mask images was adopted to realize fast matching of non-rectangular regions of interest(ROI). The non-maximum suppression method was used to eliminate false matches caused by proximal detection regions. The top-layer template library was used for preliminary matching; and the preliminary matching results were mapped to the next level of the pyramid to dynamically generate a reduced search space and corresponding templates, which were iterated layer by layer to the bottom layer of the pyramid to achieve accurate matching. Results: In the single-target matching task(the size of the image to be matched is 1 920×1 278), under the condition that the matching intersection over union(IoU) was greater than 95%, the average time consumption of the algorithm was only 7.52 ms. In complex scenarios involving multiple targets, and rotation transformations(the resolution of the image to be matched is 646×482), the average time consumption was 52.91 ms while meeting the same requirement of matching IoU. Conclusions: The proposed algorithm is capable of achieving fast and accurate grayscale image matching in complex scenarios such as multi-pose, multi-target, and rotation transformations, demonstrating strong practical application value.
[1] KORMAN S,REICHMAN D,TSUR G,et al.Fast-match:Fast affine template matching[J].International Journal of Computer Vision,2017,121:111-125.
[2] 贾迪，杨宁华，孙劲光.像对匹配的模板选择与匹配[J].中国图象图形学报，2017,22(11):1512-1520.JIA D,YANG N H,SUN J G.Template selection and matching algorithm for image matching[J].Journal of Image and Graphics,2017,22(11):1512-1520.
[3] 李长有，付惠琪，陈国玺.改进模板匹配的快速识别算法[J].河南理工大学学报，2020,39(4):106-111.LI C Y,FU H Q,CHEN G X.A fast recognition algorithm for improved template matching[J].Journal of Henan Polytechnic University,2020,39(4):106-111.
[4] 王世勇，乾国康，李迪，等.面向边缘特征的实时模板匹配方法[J].华南理工大学学报，2023,51(9):1-10.WANG S Y,QIAN G K,LI D,et al.Real-time template matching method for edge features[J].Journal of South China University of Technology,2023,51(9):1-10.
[5] 刘常杰，吕倩云，范文龙.液化天然气自动卸料臂末端视觉定位方法研究[J].激光与光电子学进展，2025,62(20):100-108.LIU C J,LYU Q Y,FAN W L.Research on visual positioning method for the end of liquefied natural gas automatic unloading arm[J].Laser & Optoelectronics Progress,2025,62(20):100-108.
[6] 贾迪，朱宁丹，杨宁华，等.图像匹配方法研究综述[J].中国图象图形学报，2019,24(5):677-699.JIA D,ZHU N D,YANG N H,et al.Image matching methods[J].Journal of Image and Graphics,2019,24(5):677-699.
[7] 陈雨婷，谢胜龙，邹伟.基于标准距离端点数据重组的标定优化方法[J].中国计量大学学报，2023,34(3):357-363.CHEN Y T,XIE S L,ZOU W.Calibration optimization method based on the reorganization of standard distance endpoint data[J].Journal of China University of Metrology,2023,34(3):357-363.
[8] 彭向前，张贻鹏，胡小平，等.图像失真条件下的模板匹配算法研究[J].传感技术学报，2024,37(3):507-512.PENG X Q,ZHANG Y P,HU X P,et al.Research on characteristics of template matching algorithm under image distortion[J].Journal of Sensors and Actuators,2024,37(3):507-512.
[9] 褚冬亭，刘丙友，齐晶晶，等.基于改进边缘信息的ZNCC位移提取算法研究[J].电子测量技术，2024,47(8):120-125.CHU D T,LIU B Y,QI J J,et al.Research on ZNCC displacement extraction algorithm based on improved edge information[J].Electronic Measurement Technology,2024,47(8):120-125.
[10] 董晶.模板图像快速可靠匹配技术研究[D].长沙：国防科学技术大学，2015:43-56.DONG J.Study on Fast and Reliable Pattern Match[D].Changsha:National University of Defense Technology,2015:43-56.
[11] ZHANG Y F,ZHANG Z,PENG S H,et al.A rotation invariant template matching algorithm based on Sub-NCC[J].Mathematical Biosciences and Engineering,2022,19(9):9505-9519.
[12] 练文聪.面向超大图像的快速模板匹配算法及其应用研究[D].苏州：苏州大学，2023:9-17.LIAN W C.Research on Fast Template Matching Algorithm for Ultra Large Images and Its Application[D].Suzhou:Soochow University,2023:9-17.
[13] CHANG Q,ZHA A L,WANG W M,et al.Z2-ZNCC:Zigzag scanning based zero-means normalized cross correlation for fast and accurate stereo matching on embedded GPU[C]//2020 IEEE 38th International Conference on Computer Design (ICCD).Hartford,CT,USA:IEEE,2020:597-600.
[14] 霍冠群，陆金波，罗圣翔.基于CLAHE和改进ZNCC的图像拼接研究[J].激光与光电子学进展，2022,59(12):226-234.HUO G Q,LU J B,LUO S X.Image stitching based on CLAHE and improved ZNCC[J].Laser & Optoelectronics Progress,2022,59(12):226-234.
[15] 袁狄剑，许新科，刘通，等.基于自构卷积核改进模板匹配的红外十字靶标中心定位算法[J].红外与毫米波学报，2025,44(3):469-476.YUAN D J,XU X K,LIU T,et al.An improved template matching algorithm for infrared cross-target center positioning based on self-constructed convolution kernels[J].Journal of Infrared and Millimeter Waves,2025,44(3):469-476.
基本信息:
中图分类号:TP391.41
引用信息:
[1]池成焕,陈锡爱.基于改进ZNCC的灰度快速图像匹配算法[J].中国计量大学学报,2025,36(04):631-641.
2025-12-15
2025-12-15