| “This study addresses the challenges inherent in traditional Autism Spectrum Disorder (ASD) diagnostics by proposing an innovative approach leveraging multi depth cameras and advanced 3D point cloud analysis. Focusing on comprehensive head measurements crucial for ASD diagnosis, the research introduces a surround setup with multiple Kinect Azure depth cameras to ensure accurate and detailed data acquisition. The calibration process aligns and synchronizes point clouds seamlessly, forming the basis for high-fidelity head meshes. To enhance diagnostic precision, the study employs 2D landmark recognition techniques, projecting facial features onto 3D models, and integrates machine learning with PointNet++ to extract nuanced features. Notably, the research addresses data scarcity by incorporating synthetic head models generated through “”MakeHuman.”” This novel approach not only overcomes limitations in real head data but also augments the adaptability and robustness of the feature extraction model. The potential impact of this methodology extends beyond diagnostics, offering a non-invasive and efficient means of identifying ASD patients. By combining advanced sensor technology, meticulous calibration, and innovative synthetic data integration, this research aims to revolutionize ASD diagnostics, providing healthcare professionals with a valuable tool for early and accurate detection, ultimately contributing to improved outcomes for individuals with ASD.” |
