Hierarchical Content Based Key Retrieval

The loss of keys poses a huge security risk to an organization especially if the keys are tagged. This paper proposes a new approach for key management that allows for the identification of keys based on their photographs such that dependency on tags is reduced. If all keys in an institution are visually scanned and their details captured in a database, then key identification can theoretically be achieved through content based image retrieval. A key retrieval system based on archived boundary and region based features from processed key images, is developed and evaluated to demonstrate the viability of the idea. A cascaded approach towards key search achieves a mean average precision of 90% on a dataset of 50 key images.