面向野外火灾监测的图像拼接技术研究

Research on Image Stitching Technology for Field Fire Monitoring

传统的火灾检测系统主要依靠传感器实现,由于设备长时间接触灰尘等其他因素,有很大的可能发生失灵或故障。基于上述的缺陷,将图像拼接系统应用在野外防火检测系统上将会产生更好的效果。通过摄像头采集的图像信息往往具有一定的视差,拼接后得到的全景图像易产生重影等现象。在进行多幅图像拼接时,拼接后的全景图像不能呈现出规则的矩形形状。针对上述问题,采用一种图像变形的方法来实现精确对齐,并利用Seam Carving算法将不规则的全景图像矩形化输出为规则的全景图像,最后将规则全景图像与原场景图像进行背景对比,从而达到火灾检测的目的。

The traditional fire detection system relies mainly on sensors. Due to other factors such as dust exposure to equipment for a long time,there is a great possibility of failure or malfunction.Based on the above defects,the application of the image splicing system to the field fire detection system may produce good results. The image information collected by the camera often has a certain visual error and the panoramic image obtained after the stitching is prone to ghosting. When stitching a plurality of images,the spliced panoramic image is not a regular rectangular shape. In response to the above problem,an image distortion method is used to achieve precise alignment and the irregular panoramic image is rectangularized and output as a regular panoramic image by using the Seam Carving algorithm. Finally,the regular panoramic image is compared with the original scene image to achieve the purpose of fire detection.