天际线识别在森林烟火识别中的应用

Automatic detection of the skyline in forest fire watch

在森林烟火检测中,为减少烟火虚警率,需在监控图像中自动分割出森林区域。在森林区域分割过程中,首先将天空与森林区域分割开,以消除天空飞行物及云彩等的变化对烟火目标的干扰。针对这一问题,提出了一种新的基于 Mumford-Shah 泛函模型和曲线演化原理的天际线识别算法并将其应用到森林烟火识别系统中.这种算法不仅能有效地检测出具有较小梯度图像的天际线目标,而且还可直接给出天际线的曲线表示。实验数据显示算法具有良好的有效性和自适应性。

To reduce the rate of false alarm in fire watch,the forest area should be segmented.In this process, first the sky from the forest is separate in order to eliminate the effects of moving objects or the changing clouds in the sky.first,a new algorithm is presented based on both Munford-Shah model and the principle of curve evolution, and then the algorithm is applied to the system for fire watch.The algorithm not only can effectively detect the skyline with unobvious gradient image and also get directly the curve representation of the skyline at last. Experiments prove the efficiency and self-adaptation of our algorithm.