CAJ | 학술논문

针对物体轮廓分类准确性和实时性的要求，提出了一种新的基于傅立叶描述子主要系数的分类树构造方法。首先构建梯度递增决策树( GBDT)模型对轮廓傅立叶描述子进行特征选择，得到其主要系数；然后利用主要系数构建分类和回归树( CART)模型对轮廓分类。实验表明，在保证较高分类准确率的情况下，此分类方法平均耗时仅为0．03 s。
침대물체륜곽분류준학성화실시성적요구，제출료일충신적기우부립협묘술자주요계수적분류수구조방법。수선구건제도체증결책수( GBDT)모형대륜곽부립협묘술자진행특정선택，득도기주요계수；연후이용주요계수구건분류화회귀수( CART)모형대륜곽분류。실험표명，재보증교고분류준학솔적정황하，차분류방법평균모시부위0．03 s。
Aiming at the precision and real-time of classification methods of object contours, a novel classification tree of object contours was presented based on Fourier descriptor's principal coefficients. First a Gradient Boost Decision Tree ( GBDT) model was constructed to select the principal coefficients of Fourier descriptor; then the principle coefficients were used to creat Classification And Regression Tree ( CART) model and classify the contours. Experimental results verify the effectiveness of this method which classification's mean time is only 0. 03 seconds with high accuracy .