基于Riverbed的机载电子板卡红外图像分割  

The Segmentation of Infrared Image of Airborne Electronic Board Based on Riverbed

作　　者：王坤[1] 陆文斌[1] 王力[1] 张红颖[1] 

机构地区：[1]中国民航大学航空自动化学院,天津300300

出　　处：《控制工程》2014年第5期726-730,共5页Control Engineering of China

基　　金：国家自然科学基金(U1333111);中国民航大学校级项目基金(08QD17X;2010kyE07);中央高校基本科研项目支持(ZXH2009C006;ZXB2011A003);天津市自然基金(12JCQNJC00600);中国民航机务维修科研基地资助

摘　　要：在机载电子板卡故障诊断过程中,需要对故障芯片进行准确的定位。针对机载电子板卡芯片布局紧密、工作芯片的热辐射范围广的问题,提出一种新型的人机交互式红外图像分割方法。首先,利用基于图像森林变换的Riverbed算法对故障目标图像进行边界追踪,并将Live Wire的特性与之相结合,优化了对边界模糊的目标的轮廓的提取,显著减少了所需的定位点数;然后,鉴于芯片的外形特征,采用基于梯度幅度直方图和类内方差最小化的自适应方法找到芯片的边界,用得到的芯片边界对目标轮廓进行约束,提高目标区域定位的精度。实验结果表明:该方法有效提高了故障边界提取的速度,与其他经典算法相比较有效提高了芯片的核心发热区域提取的精度,避免了多分割的现象。The accurate positioning of faulty chip is needed in the process of fault diagnosis of airborne electronic boards. For the densely layout and the wide scope of the thermal radiation of chips in the airborne electronic board, a new method of user-steered infrared image segmentation is proposed. The image is segmented by Riverbed with Image Foresting Transform （ IFF）, and combining the character of Live Wire to Riverbed, the contour extraction of the object with undefined boundary is optimized, moreover, it can significantly reduce the number of anchor points as compared to Riverbed for objects; In view of the shape characteristic of the chip, a histogram of gradient and adaptive minimum within-class variance-based method is used to obtain the boundary of the chip, which is taken to restrain the counter of the object to enhance the accuracy of positioning the target. Experiment results show that the method effectively improve the speed and accuracy of the extraction of the core of the heat generating region of the chip compared with other classic algorithms, avoid the phenomenon of multi-segmentation.

关 键 词：交互式 分割 RIVERBED LIVE WIRE 多分割 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]