In this paper, we propose a location-based large-scale landmark image recognition scheme for mobile devices such as smart phones. To achieve this goal, we collected landmark images all around the world, which were available on the web. For each landmark, we detected interest points and constructed their feature descriptors using SURF. Next, we performed a statistical analysis on the local features to select representative points among them. Intuitively, the representative points of an object are the interest points that best characterize the object. Similar representative points are merged for filtering and fast matching purposes. These points are indexed using an R-tree based on GPS information. Our scheme is based on client-server architecture. When the user takes a picture of a landmark using a mobile device, the client module on the mobile device extracts the local features from the image and sends them to the server, along with location and other sensor data. For the query, the server searches its index using the location data first to find nearby landmarks and then compares their local features. Matched landmark images are sent back to the client. We implemented a prototype system and performed various experiments. Through experiments, we showed that our scheme achieves reasonable performance.