论文部分内容阅读
为了防止井下机车撞人事故发生,提出了一种基于图像处理的井下机车行人检测技术。为矿用摄像机选择窄带滤光片,通过Hough变换并提出了极角极径约束法提升轨道检测的有效性,并将获取的轨道参数信息与实际的轨道间距相结合动态的标定矩形感兴趣区,引入卷积神经网络并与SVM和Adaboost进行对比,在训练样本的选择中,为了降低错检率,提出了负样本筛选法,最后为了消除轨道外围行人的干扰设计黑色带,结果表明可以很好地识别机车正前方行人并作出预警判断。
In order to prevent underground locomotive from colliding with an accident, a locomotive pedestrian detection technology based on image processing is proposed. The selection of narrow-band filters for mining cameras, the Hough transform and the polar-radius constraint method are proposed to enhance the effectiveness of the orbit detection. The obtained orbital parameters are combined with the actual orbital spacing to dynamically calibrate the rectangular region of interest , Introduced the convolutional neural network and compared with SVM and Adaboost. In the selection of training samples, in order to reduce the mis-detection rate, a negative sample screening method was proposed. Finally, in order to eliminate the interference of the pedestrian around the track, the black band was designed. Good recognition of locomotive in front of pedestrians and make early warning judgments.