摘要
研究了彩色图像的分割算法,提出了运用彩色图像的颜色特征相似性进行图像的分割;然后将分割的目标进行检测,运用无监督网络,即竞争网络学习训练聚类,将图像中不同的目标提取出来。在识别方面,研究了一种运用空间距离变换方法来识别的算法,计算出区域半径测试样本点在区域内还是区域外,以达到识别的目的。实验结果表明,本文提出的分割算法和识别算法可以很好地运用到实际中,识别率可达90%以上。
The target recognition method is studied. In the proposed framework, the color characteristic comparability of the color image is used for image segmentation firstly, and a non-supervisory network, called the competition network based clustering, is then employed for the objection recognition. Finally, a novel classification algorithm is used for recognition. The algorithm uses space transformation distance with implicit nonlinear projection to calculate regional radius, and then tests whether samples are or not in the region. Experimental results show that the proposed method can be well applied in practice and the recognition rate is more than 90%.
出处
《南京航空航天大学学报》
EI
CAS
CSCD
北大核心
2007年第5期601-606,共6页
Journal of Nanjing University of Aeronautics & Astronautics
关键词
分割
检测
识别
聚类
segmentation
detection
recognition
clustering
