Dynamical analysis, FPGA implementation and optical remote sensing image sensitive targets encryption of locally active Memristive HNN

Satellite remote sensing images acquired from space are widely used in national economic development and military. Remote sensing images related to marine scenarios often contain sensitive military information. However, ensuring secure transmission in the field of remote sensing images remains a crucial challenge. In this paper, a novel multistable locally-active memristor is proposed, exhibiting rich hysteresis loops and both locally active and passive behaviors under different parameters. Based on this model, a three-neuron memristive Hopfield neural network (MHNN) is constructed by incorporating three bio-inspired neural mechanisms: electromagnetic radiation effects, self-regulating synapses, and memristive inter-neuronal synaptic coupling. The effects of memristor parameters and coupling weights on the neural network dynamics are investigated. In addition, the digital circuit implementation of the memristor-based HNN is realized and functionally verified using a field-programmable gate array (FPGA). Finally, an optical remote sensing image encryption algorithm is proposed to enhance the security of sensitive targets in marine-related scenarios. The proposed method consists of two stages: object detection and image encryption. This method selectively encrypts the color images by combining multi-channel permutation and bidirectional diffusion. Simulation results demonstrate the effectiveness of the proposed scheme and its resistance to chosen-plaintext attacks.