Image Retrieval Method Based on Visual Map Pre-Sampling Construction in Indoor Positioning

In visual indoor positioning systems, the method of constructing a map by point-by-point sampling is widely used due to its characteristics clear static images and simple coordinate calculation. However, too small interval will cause image redundancy, while large lead absence any scene images, which result in worse efficiency inferior accuracy. As result, this paper proposed construction based on pre-sampled features matching, according epipolar geometry adjacent position determine optimal spacing within constraints effectively control database size ensuring integrity information. addition, order realize rapid retrieval reduce error caused time overhead, an deep hashing was also designed paper. This convolutional neural network extract construct semantic similarity structure guide generation hash code. Based log-cosh function, loss function whose curve smooth not affected outliers, then integrated it into optimize parameters, for fast accurate retrieval. Experiments FLICKR25K dataset proved that could achieve sub-second with guaranteed accuracy, thereby demonstrating promising performance.