Object–attribute–relation model-based semantic coding for image transmission

The development of digital, networked and intelligent communication technology has led to significant growth in data and services. Especially in the communication scenario of monitoring data, a variety of downstream tasks bring significant technical challenges for high-performance data transmission, information mining, and data distribution. The problem to be solved is the coding and low bit-rate transmission of large amounts of surveillance data. Concepts and methods that utilize semantic communication are needed. This paper proposes a communication architecture based on the object–attribute–relation (OAR) semantic knowledge base. The sender uses the model parameters and prior knowledge in the knowledge base to drive the OAR extraction model to encode the image. After semantic coding and channel coding, the signal is transmitted to the receiver through the physical channel. At the receiver, the signal is decoded into a semantic vector. The OAR global vector is generated by structured processing, which is used for downstream tasks such as image retrieval or image reconstruction. Experimental comparison and analysis illustrate the effectiveness and superiority of each part of the communication framework. This approach provides a new solution for improving data utilization and processing efficiency.