基于计算机视觉的弹着点坐标远程测量系统

系统采用双目立体视觉原理，通过捕捉、处理弹落击起的烟尘图像精确地获取弹着点信息。给出了一种适于野外环境构建的视觉测量模型，并提出了一种仅对焦距及光轴水平偏角进行现场标定的相机参数快速获取方法。采用序列图像实时帧间差分阈值法作为主要的图像处理手段，通过构造高斯背景统计模型获得精确的分割阈值。为排除野外环境的各种干扰，提出了依据时空相关性检测的帧间多级差分方法，为解决烟尘快速飘移对后续弹点探测的影响，给出了一种伪目标边缘跟踪算法。远程视觉传感器距靶心1.2 km对?准400 m靶区实测时，图像处理时间小于8 ms，系统响应时间小于75 ms，报靶精度优于2 m，完全满足实时准确报靶的要求。

Remote measuring system of bomb-fall based on computer vision

Based on the binocular stereo vision theory,the smoke images were caught and processed to get the bomb position information.A practical vision measuring model was given which model adapts to field environment.And a high speed camera calibration method was proposed which just calibrate two key camera parameters: the focal length and the horizontal angle of the optical axis.A real-time image processing method based on frame difference was used to process the images in system.A Gaussian background model based on background subtracting method was presented to get the segmentation threshold.To eliminate interference brought by field environment,a multilevel difference method based on detection of space-time relativity was proposed.To avoid the effect caused by the quickly moving dust,a false object boundary tracking method was presented.The model is accurate to 2 meter when scoring within a 200 meter radius of the impact zone.