Dynamic programming approach to visual place recognition in changing environments

Recognising places under extreme perceptual changes is a challenging problem. A new dynamic programming method to align sequences of image features extracted from a deep convolutional auto-encoder to efficiently solve this problem is proposed. As this method considers not only environmental variations, but also the motion constraint of the mobile robot, places from changing environment can be successfully recognised by finding the most likely path sequence. Experimental results show improved precision–recall performance compared with other algorithms.