algorithm - 山脉识别算法

Question

如果我有一张显示山脉的照片，是否有一种算法或方法可以尝试搜索并找到该山脉？例如，假设我有一张这样的旧照片(片段):

所以在这里我们可以在背景中看到 3 个不同的山脉，我们可以用手勾勒出它们在天空或背后山脉中的轮廓。

将这些轮廓线作为输入，是否有一种算法可以将其与 DEM 相匹配？总体目标是找出照片的拍摄地点。

Answer 1

这样的算法确实存在，至少对于受限区域而言。例如，参见论文:

使用数字高程模型对未标记沙漠图像进行用户驱动的地理定位，Tzeng, E. 等人，计算机视觉和模式识别研讨会 (CVPRW)，2013 年 6 月 23 日至 28 日，俄勒冈州波特兰。

(Abstract): We propose a system for user-aided visuallocalization of desert imagery without the use of any metadata such asGPS readings, camera focal length, or field-of-view. The system makesuse only of publicly available digital elevation models (DEMs) torapidly and accurately locate photographs in non-urban environmentssuch as deserts. Our system generates synthetic skyline views from aDEM and extracts stable concavity-based features from these skylinesto form a database. To localize queries, a user manually traces theskyline on an input photograph. The skyline is automatically refinedbased on this estimate, and the same concavity-based features areextracted. We then apply a variety of geometrically constrainedmatching techniques to efficiently and accurately match the queryskyline to a database skyline, thereby localizing the query image. Weevaluate our system using a test set of 44 ground-truthed images overa 10, 000 km2 region of interest in a desert and show that in manycases, queries can be localized with precision as fine as 100 m2.

全文也是available .

当然，这种技术的规模(例如全局范围)是另一回事......

另一个relevant paper是:

山地地形图像的大规模视觉地理定位，Georges Baatz 等人，Proc。 2012 年欧洲计算机视觉 session

Abstract. Given a picture taken somewhere in the world, automaticgeo-localization of that image is a task that would be extremelyuseful e.g. for historical and forensic sciences, documentation purposes, organization of the world’s photo material and also intelligenceapplications. While tremendous progress has been made over the lastyears in visual location recognition within a single city,localization in natural environments is much more difficult, sincevegetation, illumination, seasonal changes make appearance-onlyapproaches impractical. In this work, we target mountainous terrainand use digital elevation models to extract representations for fastvisual database lookup. We propose an automated approach for verylarge scale visual localization that can efficiently exploit visualinformation (contours) and geometric constraints (consistentorientation) at the same time. We validate the system on the scale ofa whole country (Switzerland, 40000km 2 ) using a new dataset of morethan 200 landscape query pictures with ground truth.