低空视频影像自适应关键帧提取与快速拼接

Adaptive Key Frames Extraction and Fast Image Mosaic of Low Altitude Aerial Video

摘要: 无人飞行器超低空飞行，获取图像幅宽小，数量多，重复率高，为获取全测区的图像需要图像拼接。目前航空影像拼接广泛使用基于特征匹配的图像配准方法。特征匹配的计算量大，错误率高，影响了图像拼接的速度。针对上述问题提出了一种自适应关键帧的图像拼接方法。该方法根据基准影像和给定的重叠度自适应提取关键帧影像，使用正解和反解相结合的方法完成对关键帧影像的正射校正。通过坐标解算得到每个关键帧影像中需要参与拼接的部分，然后使用横向流型拼接方法将每个关键帧中参与拼接的部分拼接形成测区全图。实验结果表明本文方法在保证较高的拼接精度下可以实现航拍影像的快速实时拼接，很大程度上提高了图像拼接的效率。本文方法可以实现航拍影像快速拼接，满足现场数据检验以及其它快速应急响应的需求，在灾害应急保障与救援中有重要应用意义。

Abstract: Images captured by unmanned aerial vehicle have small image width, large quantity, and high overlap rate. In order to get the image of whole measurement area, image mosaicking is necessary. Feature matching method is widely used in the field of aerial image mosaicing. Feature matching has many disadvantages such as heavy burden of calculation, high error rate, so affected the speed of image stitching greatly. To solve those problems, an adaptive key frame was proposed. Extracting method extracts key frame images according to overlap rate of images and gets the orthographical correction of key frame images by using the combination of direct method and inverse solution method. The part involved in image mosaicing of every key frame was obtained by using its coordinates. Then we get the mosaiced image of the whole area by using flow pattern stitching method. Experimental results show that the method proposed in this paper can realize real-time aerial image stitching with high precision and can improve the efficiency of image mosacing sig-nificantly. The method proposed in this paper can realize real-time aerial image stitching and meet requirements of rapid response, especially in disaster emergency response and rescue.