Abstract: Accurate retrieval of power grid data information plays a very important role in ensuring the normal operation of the power grid system. It can effectively improve the work efficiency of power grid staff to quickly and accurately find images with high similarity to the target image from the power grid image database. Aiming at the low retrieval accuracy of traditional retrieval methods, an end-to-end hash coding method based on time-domain and frequency-domain is proposed. The experimental results on two datasets show the effectiveness of the proposed method. The model combines the frequency domain information innovatively to improve the prediction accuracy, and adds multi-task learning and distance circle loss to constrain the training process of hash coding task more clearly, which makes the image retrieval results more accurate.

Key words: power grid image data, deep learning, image retrieval, multi-task learning, hash coding ,