| 张东波,彭英辉,刘霞,宋迪,王耀南.基于线扫描分析的点状目标检测和分割[J].光电子激光,2014,(8):1586~1591 |
| 基于线扫描分析的点状目标检测和分割 |
| Spot target detection and segmentation with line scanning analysis |
| 投稿时间：2014-03-20 |
| DOI： |
| 中文关键词: 点目标检测 线扫描 局部二值分析 |
| 英文关键词:spot target detection line scanning local binary analysis |
| 基金项目:国家自然科学基金(60835004)、湖南省教育厅重点(14A137)和湖南省重点学科资助项目 (1．湘潭大学 信息工程学院,湖南 湘潭 411105； 2．湖南大学 电气与信息工程学院,湖南 长沙 410082) |
|
| 摘要点击次数: 1376 |
| 全文下载次数: 167 |
| 中文摘要: |
| 根据点状目标在图像空间呈现的局部峰值或极小 值区域一般具有全向对称和高斯分布特性,提出一种 新颖的旋转线扫描算子对点目标的截面进行特征分析,并结合局部峰值分析和梯度信息,实 现点目标的检 测与分割。实验结果表明,本文算法对点目标外观如大小、局部的对比度和灰度变化具 有良好的适应 性；在手机隔板砂粒检测中,算法检测准确率达到0.993；而在视网 膜眼底图像微脉瘤病变检测中,在可 临床接受的误判率(1.08FP/image)前提下,检测准确率为0.42。大量 实际图像测试表明,本文算法具有良好的检测性能和鲁棒性,并且节约资源和时间成本。 |
| 英文摘要: |
| The robust detection of spot object in image with complicated backgrou nd and interference is a challenging problem.The circular spots appear as round local peaks or valleys in image space,and they reflect the omnidirectional symmetry and Gaussian distribution characteristics.A ccording to these observations,a novel rotating line scanning operator is proposed to achieve prof ile feature analysis of the spot object.By combining local peak analysis and gradient information,the spot d et ection and segmentation can be achieved.Experimental tests show that the proposed algorithm is adaptive to spots′ appearance variation (size,local contrast and intensity).For mobile phone partition resid ual sand detection,the accuracy of the proposed method achieves 0.993,and for microaneurysms detection in retinal image,the accuracy is 0.42under the clinically acceptable false positive rate (1.08FP/im a ge).The method presents promising performance and excellent robustness in variou s real images. |
| 查看全文 下载PDF阅读器 |
| 关闭 |
|
|
|
|
|