CAJ | 학술논문

为了从 LiDAR 点云数据中准确可靠地提取汽车目标，本文基于 LiDAR 高程数据和强度信息，提出一种基于像素和面向对象分类的汽车提取方法。该方法在形态学开重建滤波的基础上，首先根据无植被覆盖的地面点强度低的特点，利用 Otsu 方法对城区 LiDAR 强度数据进行分类，结合开重建的滤波结果完成基于像素的初步分类；然后根据汽车的尺寸特征和其行驶或停靠在无植被覆盖的地面上的特点，利用形态学开重建获取符合汽车部分特征的区域对象，最后根据汽车对象的形状特征及其高程属性，采用面向对象的分类方法提取汽车目标。实验结果证明了本文方法的可行性和有效性。
위료종 LiDAR 점운수거중준학가고지제취기차목표，본문기우 LiDAR 고정수거화강도신식，제출일충기우상소화면향대상분류적기차제취방법。해방법재형태학개중건려파적기출상，수선근거무식피복개적지면점강도저적특점，이용 Otsu 방법대성구 LiDAR 강도수거진행분류，결합개중건적려파결과완성기우상소적초보분류；연후근거기차적척촌특정화기행사혹정고재무식피복개적지면상적특점，이용형태학개중건획취부합기차부분특정적구역대상，최후근거기차대상적형상특정급기고정속성，채용면향대상적분류방법제취기차목표。실험결과증명료본문방법적가행성화유효성。
To extract car targets from LiDAR data exactly and reliably,a pixel and object oriented method of car extraction is proposed on the basis of LiDAR elevation and intensity data after filtering.Firstly,according to the low intensity of non-vegetation ground,this method classifies pixel point by Otsu based on intensity.Combined with the result of morphological opening by reconstruction filtering,it achieves the pixel classification.Secondly,according to the characteristics of car driven and stopped only on non-vegetation ground,it gets the satisfied object region by morphological opening by reconstruction for the size and area feature of car.Finally,it uses object-oriented method to extract car for its shape feature and height property.It can be seen that this method is feasible and effective from experimental results.