基于归一化二维相关技术的图像不变目标检测
计算机视觉与模式识别
2023-12-08 v2 图像与视频处理
摘要
归一化二维相关技术是一种在图像中检测目标的鲁棒方法,因其能够在旋转、平移和缩放下保持不变性。本文考察了平移和缩放对图像中目标识别的影响。结果表明,即使在目标位置和大小发生变化时,检测仍具有较高准确率。结果表明,随着缩放比增大,图像与所用两个目标之间的相似度提高。所有统计估计量均显示出原始目标与提取目标之间的强相似性。所有场景的耗时分别在 (44.75-44.85) 和 (37.48-37.73) 秒范围内(bird 和 children 目标 respectively),相关系数呈现稳定关系,其取值分别在 (0.90-0.98) 和 (0.87-0.93) 范围内(bird 和 children 目标 respectively)。
引用
@article{arxiv.2302.11196,
title = {Invariant Target Detection in Images through the Normalized 2-D Correlation Technique},
author = {Fatin E. M. Al-Obaidi and Anwar H. Al-Saleh and Shaymaa H. Kafi and Ali J. Karam and Ali A. D. Al-Zuky},
journal= {arXiv preprint arXiv:2302.11196},
year = {2023}
}
备注
The paper in its uploaded format contained errors in analysis as in Abstract-4th and 5th line. In section 4 ; results and discussion and in conclusion section. So, in order not to adopt the research as a source by other researchers I wish to withdraw it completely