| 摘要: |
| 针对复杂自然背景下的多目标检测，提出了结合颜色和分形特征的多目标检测算法.将RGB颜色空间转换到Lab颜色空间，采用改进K-means聚类算法，去除大片背景区域，计算区域分形维数和分形拟合误差.两种分形特征相结合能够准确排除小面积背景奇异区域的干扰，检测出待测图像中的多个目标.仿真结果表明：该算法能够正确检测出复杂自然背景下的多个目标，对彩色图像分割后的保留区域求分形特征，避免了搜索目标带来的计算量.相比于对全图提取分形特征的方法，本算法在时间上缩短约80%. |
| 关键词: 复杂自然背景 多目标检测 颜色特征 分形特征 改进K-means聚类算法 |
| DOI：10.3969/J.ISSN.1000-5137.2020.01.011 |
| 分类号:TP391 |
| 基金项目:上海市自然科学基金（16ZR1424500）；上海市地方院校能力建设（19070502900） |
|
| Multi-target detection by using combined color and fractal features in complex background |
|
ZHENG Linping1, WANG Bin1, LIU Huawei2, ZHOU Xiaoping1, HUANG Jifeng1
|
|
1.College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 201418, China;2.Key Laboratory of Microsystem Technology, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China
|
| Abstract: |
| A multi-target detection algorithm combining color with fractal features was proposed for multi-objective detection in complex natural background.The RGB color space was converted to the Lab color space and the improved K-means clustering algorithm was used to remove the large background area.The regional fractal dimension and the fractal fitting error were calculated,which could accurately exclude the small area background singular region interference and detect multiple targets in the tested image.The simulation results showed that the algorithm could correctly detect multiple targets in complex natural backgrounds.Besides,the fractal features for the regions retained by color image segmentation was captured,avoiding the computation of searching targets.Compared with the algorithm of extracting fractal features on the whole graph,time was shortened by about 80% in the proposed algorithm. |
| Key words: complex natural background multi-target detection color feature fractal feature improved K-means clustering algorithm |
