CAJ | 학술논문

针对智能交通系统中车辆阴影严重影响运动目标检测的问题，在混合高斯前景检测的基础上，提出了一种基于极限学习机的阴影检测方法。该方法只需提取前景像素点处的梯度值及其邻域的方差值来作为特征，并通过极限学习机来对其进行分类，获得真实阴影区域，无需预先设置任何经验阈值。最后实验结果表明，与传统方法相比，该方法不仅实现简单，而且可以准确地检测出阴影内部区域和阴影边缘区域。
침대지능교통계통중차량음영엄중영향운동목표검측적문제，재혼합고사전경검측적기출상，제출료일충기우겁한학습궤적음영검측방법。해방법지수제취전경상소점처적제도치급기린역적방차치래작위특정，병통과겁한학습궤래대기진행분류，획득진실음영구역，무수예선설치임하경험역치。최후실험결과표명，여전통방법상비，해방법불부실현간단，이차가이준학지검측출음영내부구역화음영변연구역。
In the system of intelligent traffic, the shadow of moving cars cause terrible detection of targets, therefore, a kind of method about shadow detection based on mixed gauss model algorithm and extreme learning machine is proposed. This method only needs to extract the gradient and variance of neighbourhood about pixels, and classifies pixels using the extreme learning machine, without setting any empirical threshold. Finally, compared with the traditional method, the experimental results show that the method proposed is not only simple, but also can accurately detect the inside shadow area and the edge of shadow.