For visual approaches to simultaneous localization and mapping, the use of visual keypoint features is a popular choice. Certain areas of the environment, however, may contain ambiguous architecture or areas of repetition, which may cause keypoints to be matched well at multiple locations; it would be useful to consider other sources of evidence in addition to keypoints to make decisions. We propose a system that samples from varying layers of a locations signature, from global or simple properties such as color and straight lines, to texture related characteristics in keypoints, to symbolic, human characteristics and semantic information such as text recognition and object detection and recognition. We propose how these visual feature "ridges" can be associated together and utilized to form our version of a visual "fingerprint" of a place, how these fingerprints can be compared for localization, and how they can be linked together to form a topological map of the environment. This new approach is called SPLINTR: Spatial Place Recognition in a Topologically Mapping Robot.