一种高斯混合模型的危化品堆垛目标提取方法

A method for picking up hazardous chemical stacking targets based on Gauss mixture model

危险化学品存储状态事关人民及城市的安全。目前危化品多以平仓堆垛的方式进行存储。为保证危险化学品的安全存储,需要对货物"五距"进行实时的安全监测。针对视频监测,提出了基于高斯混合模型的危化品堆垛目标提取方法。将双目摄像头所得图像首先采用高斯混合模型实现当前货物的检测,使用背景更新的方式实现运动货物监测,进一步采用相邻差分图像像素值关系来判断货物的进出状态,最终得到货物堆垛在图像中的准确位置,以便于实现后续边界提取以及五距检测。实验证明,与均值、中值背景估计结果相比,本算法所提取货物图像噪声减少,判断货物状态准确率高达85%。

The storage status of dangerous chemicals is related to the safety of people and cities. At present,hazardous chemicals are mostly stored in stacking way. In order to ensure the safe storage of dangerous chemicals, a real-time safety monitoring system for the "five-distance" of goods is needed to be developed.Aiming at video surveillance, a Gauss mixture model is proposed to extract hazardous chemicals stacking targets.The images obtained by the binocular camera is used to detect the current cargo by Gauss mixture model. The background update algorithm is used to monitor the contour of the goods. The status of goods is judged by adjacent difference image pixel value relation, and the position of the goods stacked in image is finally obtained,so as to facilitate subsequent boundary extraction and five distance detection. Experiments show that the algorithm is better than the mean and median background estimation results. The accuracy of judging the condition of goods is up to 85 %.