基于SFFS多特征选择的航拍图像目标检测  

Image Target Detection Based on SFFS Multi-feature Selection

作　　者：陈苏婷[1,2] 牛宇宁 张闯[1,2] 张艳艳[1,2] CHEN Suting;NIU Yuning;ZHANG Chuang;ZHANG Yanyan(Jiangsu Key Laboratory of Meteorological Observation and Information Processing,Nanjing University of Information Science & Technology, Nanjing 210044, China;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology (CICAEET),Nanjing University of Information Science & Technology, Nanjing 210044, China)

机构地区：[1]南京信息工程大学江苏省气象探测与信息处理重点实验室,南京210044 [2]南京信息工程大学江苏省大气环境与装备技术协同创新中心,南京210044

出　　处：《实验室研究与探索》2019年第3期5-10,28,共7页Research and Exploration In Laboratory

基　　金：国家自然科学基金(61705019);江苏省高校重大自然科学基金(12KJA510001)

摘　　要：针对航拍图像目标多变、目标角度变化丰富以及背景复杂等特点,提出基于SFFS(Sequential Floating Forward Selection)多特征选择的航拍图像目标检测方法。首先,引入BING(Binarized Normed Gradients)算法进行目标粗定位,克服航拍图像数据量大、检测效率低等问题。其次,提取航拍图像的RI-LBP (Rotation Invariant LBP)特征、FD(Fourier Descriptor)形状特征和RIFF(Rotation Invariant Fast Feature)特征对航拍图像进行描述,避免单一特征描述能力不足带来的误检测率高的问题。最后,基于SFFS贝叶斯分类算法对RI-LBP特征、FD特征和RIFF特征集合进行最优子集选择,以进一步提高特征的泛化性与鲁棒性。实验结果表明,该多特征航拍图像目标检测算法不仅对航拍角度变化具有更强的鲁棒性,且在减少特征维数的同时,提高了目标检测的平均准确率,满足实时性的需求。In view of aerial image has the features of variable objects, rich target angle and complicated background, a method of aerial image object detection based on SFFS (Sequential floating forward selection) selection of multi features is proposed. Firstly, in order to overcome the aerial image processing with large amount of data and the low efficiency of detection, the BING (Binarized Normed Gradients) algorithm is introduced for preliminary location of target detection. Secondly, the RI-LBP (rotation invariant local binary pattern), FD(Fourier descriptor) and RIFF (rotation invariant fast feature) of aerial images are extracted to describe it comprehensively, avoiding the error of the single feature inadequate to describe. Lastly, with the purpose of improving generalization and robustness of features, Sequential floating forward selection Bayesian classification method is proposed to select above three features sets for a better set. The experiment result indicates that this multi features target detection method not only is more robust to the changes of aerial angles, but also improves the average accuracy of target detection at the same time of reducing the feature dimension, meeting the needs of real-time performance.

关 键 词：航拍图像 目标检测 BING算法粗定位 多特征提取 SFFS特征 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]