This work introduces a framework for contactless fingerprint recognition, addressing fundamental challenges in biometric authentication through the novel integration of feature processing methodologies. The proposed architecture seamlessly combines a dual-path autoencoder with a graph convolutional network, leveraging the inherent complementarity between texture patterns and minutiae features. Departing from traditional single-feature approaches, the framework implements sophisticated parallel processing streams: the dual-path encoder utilizes multiple kernel sizes for comprehensive multiscale texture analysis, while the graph convolutional network meticulously processes the geometric relationships between minutiae points. These distinct processing streams converge into a unified feature representation, capturing the multifaceted characteristics of fingerprints across different scales. Extensive experimental validation demonstrates that this integrated approach significantly outperforms conventional single-feature methodologies, marking a substantial advancement in contactless fingerprint recognition systems.