| 引用本文: | 王静静,秦世引.互补增强式空间运动目标高精度检测与分割[J].哈尔滨工业大学学报,2016,48(3):26.DOI:10.11918/j.issn.0367-6234.2016.03.005 |
| WANG Jingjing,QIN Shiyin.High accuracy detection and segmentation of space moving target by complementary enhancement[J].Journal of Harbin Institute of Technology,2016,48(3):26.DOI:10.11918/j.issn.0367-6234.2016.03.005 |
|
| 摘要: |
| 针对空间监视系统中的运动目标自主高精度的检测问题，提出一种显著性计算与光流检测互补增强式算法.以mean shift初检结果为导引，通过显著性区域检测与光流检测的互补增强而实现空间运动目标的高精度检测.首先通过梯度信息和频率调谐滤波的互补方式分别计算整幅图像和mean shift分割后各个分块区域的显著性，再以整幅图像显著性均值为参考，确定合理的阈值以检测候选目标；与此相并行，也采用光流计算及相应的阈值检测方法获取候选目标.进而通过对显著性计算与光流检测分别得到的两个不同候选目标分布图的合取运算进行目标确认，最后再辅之以形态学滤波使确认目标得以增强，从而实现空间运动目标的高精度检测与分割.研究结果表明，在无需知道任何场景和目标先验信息，也无需人工干预的条件下，所提算法能够有效实现空间运动目标的精确检测和分割，对光照变化和噪声干扰也有很强的适应能力.
|
| 关键词: 显著性计算 光流检测 空间运动目标 均值平移 |
| DOI：10.11918/j.issn.0367-6234.2016.03.005 |
| 分类号:V391 |
| 文献标识码:A |
| 基金项目:国家自然科学基金(0,2). |
|
| High accuracy detection and segmentation of space moving target by complementary enhancement |
|
WANG Jingjing, QIN Shiyin
|
|
（ School of Automation Science and Electrical Engineering, Beihang University, 100191 Beijing, China）
|
| Abstract: |
| In view of the problem of automatic and high accuracy detection of space moving targets for space surveillance system, an algorithm of complementary enhancement by salience computation with optical flow detection is presented, in which the elementary detecting result by mean shift is taken as a guidance to realize high accuracy detection and segmentation through the complementary enhancement of salience computation and optical flow detection. Firstly, the salience computation by complementary enhancement of frequency tuned filter with gradient information is carried out for the whole frame of video image and various regions produced by mean shift segmentation separately so as to obtain their respective salient regions, then a proper threshold value is determined based on the mean salience of whole frame image to detect candidate targets. Meanwhile, in a concurrent way, another set of candidate targets are acquired by threshold based optical flow detection. Afterwards, a conjunction operation is used to obtain an intersection of two different distributed maps of candidate targets as confirmatory detection result, then the morphological filter is employed to enhance the confirmatory targets to achieve a high accuracy detection and segmentation of space moving targets. A series of experimental results with video images demonstrate that the proposed algorithm can accurately detect and segment space moving targets with neither priori knowledge of scene and targets nor manual intervention, which qualifies well performance of adaptability to illumination changes and noise disturbance.
|
| Key words: salience computation optical flow detection space moving target mean shift |
