Joint scalable quantum convolutional neural network and reverse fidelity training for high-accurate recognition in unmanned aerial vehicle surveillance

Emily Jimin Roh, Joongheon Kim, Soyi Jung, Soohyun Park

Abstract: Unmanned aerial vehicles (UAVs) have demonstrated considerable potential in surveillance tasks, primarily owing to their high mobility and adaptability. However, these vehicles are often constrained by the extensive number of parameters required in conventional neural networks, which poses significant challenges under wireless communication conditions. To address these limitations, quantum neural networks (QNNs) have emerged as a promising solution, demonstrating capabilities that classical neural networks are unable to fully realize. Specifically, quantum convolutional neural networks (QCNNs) have garnered significant attention due to their capacity to process high-dimensional vector inputs. However, the capabilities of QCNNs in feature extraction are currently constrained by the limitations of quantum computing. In response to these challenges, a scalable quantum convolutional neural network (sQCNN) is proposed, accompanied by a reverse fidelity-Train (RF-Train). This algorithm utilizes quantum fidelity to enhance the performance of the sQCNN. In the context of UAV surveillance systems, the sQCNN and RF-train framework offer a lightweight, high-performance approach that is specifically optimized for UAV applications, ensuring efficient operation under demanding conditions.

External IDs:dblp:journals/tjs/RohKJP25