MonoPCFlow: Enabling Efficient Scene Flow Estimation From Monocular View
Chichao Cheng, Guangming Wang, Yin-Dong Zheng, Lu Liu, Hesheng Wang · IEEE Transactions on Instrumentation and Measurement · 2025
Scene flow captures the dynamic changes of points in a 3D scene, essential for understanding motion in physical environments. LiDAR-based scene flow estimation methods face challenges related to resolution, refresh rate, and cost. In contrast, monocular image-based methods estimate optical flow and depth separately at different stages. This fragmented approach inevitably compromises spatial-temporal consistency and introduces error accumulation. We proposeMonocular Point Cloud FlowNet(MonoPCFlow), a novel framework for scene flow estimation directly from a pair of consecutive monocular images. We integrate pseudo-LiDAR representations with dense 3D scene flow estimation, effectively bridging the 2D-to-3D domain gap for monocular motion analysis. We develop a depth-enhanced refinement module that mitigates information loss in pseudo-LiDAR generation, preserving critical geometric and appearance features to improve scene flow accuracy. Experimental validation demonstrates MonoPCFlow’s superior performance, achieving 37.0% (FlyingThings3D) and 39.7% (KITTI) relative reductions in End-Point-Error compared to contemporary benchmarks.