A Two-Stage Homography Matrix Prediction Approach for Trajectory Generation in Multi-Object Tracking on Sports Fields

IF 2 4区 计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE
Pan Zhang, Jiangtao Luo, Guoliang Xu, Xupeng Liang
{"title":"A Two-Stage Homography Matrix Prediction Approach for Trajectory Generation in Multi-Object Tracking on Sports Fields","authors":"Pan Zhang,&nbsp;Jiangtao Luo,&nbsp;Guoliang Xu,&nbsp;Xupeng Liang","doi":"10.1049/ipr2.70136","DOIUrl":null,"url":null,"abstract":"<p>Homography estimation is a fundamental topic in computer vision, especially in scenarios that require perspective changes for intelligent analysis of sports fields, where it plays a crucial role. Existing methods predict the homography matrix either indirectly by evaluating the 4-key-point coordinate deviation in paired images with the same visual content or directly by fine-tuning the 8 degrees of freedom numerical values that define the matrix. However, these approaches often fail to effectively incorporate coordinate positional information and overlook optimal application scenarios, leading to significant accuracy bottlenecks, particularly for paired images with differing visual content. To address these issues, we propose an approach that integrates both methods in a staged manner, leveraging their respective advantages. In the first stage, positional information is embedded to enhance convolutional computations, replacing serial concatenation in traditional feature fusion with parallel concatenation, while using 4-key-point coordinate deviation to predict the macroscopic homography matrix. In the second stage, positional information is further integrated into the input images to refine the direct 8 degrees of freedom numerical predictions, improving matrix fine-tuning accuracy. Comparative experiments with state-of-the-art methods demonstrate that our approach achieves superior performance, yielding a root mean square error as low as 1.25 and an average corner errror as low as 14.1 in homography transformation of competitive sports image pairs.</p>","PeriodicalId":56303,"journal":{"name":"IET Image Processing","volume":"19 1","pages":""},"PeriodicalIF":2.0000,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1049/ipr2.70136","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IET Image Processing","FirstCategoryId":"94","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1049/ipr2.70136","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0

Abstract

Homography estimation is a fundamental topic in computer vision, especially in scenarios that require perspective changes for intelligent analysis of sports fields, where it plays a crucial role. Existing methods predict the homography matrix either indirectly by evaluating the 4-key-point coordinate deviation in paired images with the same visual content or directly by fine-tuning the 8 degrees of freedom numerical values that define the matrix. However, these approaches often fail to effectively incorporate coordinate positional information and overlook optimal application scenarios, leading to significant accuracy bottlenecks, particularly for paired images with differing visual content. To address these issues, we propose an approach that integrates both methods in a staged manner, leveraging their respective advantages. In the first stage, positional information is embedded to enhance convolutional computations, replacing serial concatenation in traditional feature fusion with parallel concatenation, while using 4-key-point coordinate deviation to predict the macroscopic homography matrix. In the second stage, positional information is further integrated into the input images to refine the direct 8 degrees of freedom numerical predictions, improving matrix fine-tuning accuracy. Comparative experiments with state-of-the-art methods demonstrate that our approach achieves superior performance, yielding a root mean square error as low as 1.25 and an average corner errror as low as 14.1 in homography transformation of competitive sports image pairs.

Abstract Image

运动场上多目标跟踪轨迹生成的两阶段单应矩阵预测方法
单应性估计是计算机视觉中的一个基本问题,特别是在需要视角变化的场景中,对于体育领域的智能分析,它起着至关重要的作用。现有方法要么通过评估具有相同视觉内容的成对图像的4个关键点坐标偏差来间接预测单应性矩阵,要么通过微调定义矩阵的8个自由度数值来直接预测矩阵。然而,这些方法往往不能有效地结合坐标位置信息,并且忽略了最佳应用场景,导致显著的精度瓶颈,特别是对于具有不同视觉内容的成对图像。为了解决这些问题,我们提出了一种分阶段结合、发挥各自优势的方法。第一阶段,嵌入位置信息增强卷积计算能力,用并行拼接取代传统特征融合中的串行拼接,同时利用4点坐标偏差预测宏观单应性矩阵。第二阶段,将位置信息进一步整合到输入图像中,对直接的8自由度数值预测进行细化,提高矩阵微调精度。对比实验表明,本文方法在竞技体育图像对的单应性变换中取得了较好的效果,均方根误差低至1.25,平均角误差低至14.1。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
IET Image Processing
IET Image Processing 工程技术-工程:电子与电气
CiteScore
5.40
自引率
8.70%
发文量
282
审稿时长
6 months
期刊介绍: The IET Image Processing journal encompasses research areas related to the generation, processing and communication of visual information. The focus of the journal is the coverage of the latest research results in image and video processing, including image generation and display, enhancement and restoration, segmentation, colour and texture analysis, coding and communication, implementations and architectures as well as innovative applications. Principal topics include: Generation and Display - Imaging sensors and acquisition systems, illumination, sampling and scanning, quantization, colour reproduction, image rendering, display and printing systems, evaluation of image quality. Processing and Analysis - Image enhancement, restoration, segmentation, registration, multispectral, colour and texture processing, multiresolution processing and wavelets, morphological operations, stereoscopic and 3-D processing, motion detection and estimation, video and image sequence processing. Implementations and Architectures - Image and video processing hardware and software, design and construction, architectures and software, neural, adaptive, and fuzzy processing. Coding and Transmission - Image and video compression and coding, compression standards, noise modelling, visual information networks, streamed video. Retrieval and Multimedia - Storage of images and video, database design, image retrieval, video annotation and editing, mixed media incorporating visual information, multimedia systems and applications, image and video watermarking, steganography. Applications - Innovative application of image and video processing technologies to any field, including life sciences, earth sciences, astronomy, document processing and security. Current Special Issue Call for Papers: Evolutionary Computation for Image Processing - https://digital-library.theiet.org/files/IET_IPR_CFP_EC.pdf AI-Powered 3D Vision - https://digital-library.theiet.org/files/IET_IPR_CFP_AIPV.pdf Multidisciplinary advancement of Imaging Technologies: From Medical Diagnostics and Genomics to Cognitive Machine Vision, and Artificial Intelligence - https://digital-library.theiet.org/files/IET_IPR_CFP_IST.pdf Deep Learning for 3D Reconstruction - https://digital-library.theiet.org/files/IET_IPR_CFP_DLR.pdf
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信