CAJ | 학술논문

确定描述机场跑道属性的特征是机场检测的关键问题。针对此，研究了机场跑道区域的特征，利用统计矩、Hu不变矩、灰度共生矩阵、Zernike矩、傅里叶频谱、Gabor频谱、HSV颜色空间构建123维特征向量描述跑道。采用CART-Adaboost算法检测机场跑道区域，在检测结果的二值图上进行长直线提取，实验结果表明机场跑道检测算法在复杂背景下具有较高的检测率。利用Adaboost确定跑道的最佳描述特征，实现特征降维，实验结果表明使用选择出的14维特征在检测性能相近时，能大幅提高特征提取和分类器训练的效率。
학정묘술궤장포도속성적특정시궤장검측적관건문제。침대차，연구료궤장포도구역적특정，이용통계구、Hu불변구、회도공생구진、Zernike구、부리협빈보、Gabor빈보、HSV안색공간구건123유특정향량묘술포도。채용CART-Adaboost산법검측궤장포도구역，재검측결과적이치도상진행장직선제취，실험결과표명궤장포도검측산법재복잡배경하구유교고적검측솔。이용Adaboost학정포도적최가묘술특정，실현특정강유，실험결과표명사용선택출적14유특정재검측성능상근시，능대폭제고특정제취화분류기훈련적효솔。
Selecting descriptive features of runway properties is the key to airport detection.Aiming at this issue,we studied on the air-port runway features and constructed a 123 dimensional feature vector,including statistical moment,Hu invariant moment,Gray-level co-occurrence matrix,Zernike moment,Fourier spectrum,Gabor spectrum and color information in HSV space,to describe the runway. The CART-Adaboost algorithm is employed to classify the runway region,and the airport is detected by performing line extraction on the classification result binary image.The experiment results show that the airport runway detection algorithm has high detection rate in the image with complex background.The Adaboost algorithm is employed to select the most descriptive features of the runway and the select-ed subset is also used to detect airport;the experiments show that using the selected 14 dimensional features can greatly improve the effi-ciency of the feature extraction and classifier training on the premise of similar detection performance.