Efficient Detection of Non-Lambertian Reflectance Behavior and Lambertian Surface Characterization Using Minimal Angular Sampling in Light Fields

This paper presents a method for classifying Lambertian, Non-Lambertian, and Partial Lambertian surfaces in light field data by extracting angular variance features from multiple views of 3D scenes. The approach uses K-means clustering to group scenes based on feature similarity, followed by training a Random Forest classifier to predict surface categories. The classifier is evaluated on unseen test scenes, and its performance is analyzed using evaluation metrics such as accuracy, precision, and recall. The results are visualized through PCA plots and radar charts, which provide a clear understanding of how scenes cluster according to their reflective behavior. Variance maps and feature importance plots offer further insights into the key characteristics that differentiate Lambertian and Non-Lambertian surfaces. This methodology demonstrates strong performance in distinguishing between uniform and variable reflectance behaviors and is applicable to 3D scene classification, with potential applications in augmented reality, robotics, and environmental modeling.