为解决复杂场景目标识别中伪目标的干扰问题,采用基于AdaBoost分类的方法分析疑似目标的三维轨迹,结合真实目标共有的特征信息,进一步分类真实目标与伪目标.首先,根据深度相机获取的深度图像提取疑似目标的人头区域,利用Kalman滤波跟踪得到二维轨迹;其次,通过摄像机标定将目标的二维轨迹转换为空间中的三维轨迹;最后,利用AdaBoost训练正负样本得到强分类器,进一步分类真实目标与伪目标.实验结果表明,该方法能够有效地提高目标识别的精度,对复杂场景下的目标识别具有良好的适应性.%In order to solve the interference of pseudo target for the object recognition in complex scenes,this paper used the AdaBoost classifier to analyze the 3D trajectory of suspected object.It combined the feature information of the objects to classify the real objects from the false objects.Firstly,it used a depth camera to obtain the depth image and extracted the head region of suspected objects,and it adopted a Kalman filter to track the 2D trajectory on the image.Secondly,the method converted the 2D trajectory to 3D trajectory in space using the camera calibration.Finally,it applied the AdaBoost classifier to train the positive and negative samples,and it got a strong classifier.It used the strong classifier to further distinguish real targets from pseudo target.Experimental results show that this method can improve the accuracy of object identification effectively.It has a good adaptability for object recognition in complex scenes.
展开▼
