基于K-means聚类的木材运输行为的可视化监管

Visualized Management of Wood Transportation Behaviors Based on K-means Clustering Method

视频监控是有效的木材运输行为监管方法,但是随着摄像头数量的增长,人工监视实时视频的方法出现了较多的监管事故.为了提高监管效率,对比了多种图像分割技术,对K-means聚类分析算法实现了改良.通过将帧图像的色彩空间从RGB变换为YCbCr后,忽略Y(亮度因子),仅在Cb-Cr平面上进行K-means聚类分析,兼容了木材货物在复杂光照条件下形成的明暗斑纹,获得了较好的聚类效果.系统投入应用后,对空车及其它货物车辆自动忽略,仅在出现木材运输车辆时提示值守人员,并提供精确的车、货分割图像,实现了可视化管理,减少人工判识的错误.随着系统的运行,将累积车、货图像大数据,可进一步通过神经网络技术实现车型和木材运输量的识别.

Video supervision is an effective method to monitor wood transportation behaviors,but with the increase of the number of cameras,the manual inspection of real-time videos has encountered increasingly more accidents.In order to improve the efficiency of supervision,this study compared a variety of image segmentation techniques to optimize the K-means clustering analysis algorithm.After converting the color space of frame images from RGB to YCbCr and neglecting Y（luminance factor）,the K-means clustering analysis can be performed only on the Cb-Cr plane.This method successfully neglects the light and dark stripes of wood goods formed by complex light conditions,and the clustering results are satisfactory.The new system can automatically ignore empty vehicles and vehicles carrying other goods,and only remind the operators on duty when wood transportation vehicles appear.It is also possible to provide accurate pictures that separate goods from vehicles,which can realize visualized management and reduce human errors.With the operation of this system,the big data containing vehicles and goods will be accumulated gradually to further achieve identification of vehicle types and wood transportation volume based on neural network.