采用物理可靠性门控的隧道偏振法线增强方法OA

To address the issues of intensity-imaging degradation and polarization normal noise amplification caused by alternating glare and low-illumination conditions in highway tunnels,a polarization normal enhance-ment method for tunnel traffic scenes is proposed.Based on the Fresnel reflection relationship,the relationship between the degree of linear polarization(DoLP)and the reliability of normal estimation is analyzed.A com-posite reliability model integrating DoLP and intensity saturation is then constructed and used as a soft-gating weight to adaptively fuse raw normals with spatially smoothed priors.The results show that on the POLAR-GLV tunnel polarization dataset,the angular standard deviation of normals in the road-surface region has de-creased from 2.03°to 1.47°,representing a 27.6%reduction compared with the raw shape from polarization(SfP)and outperforming traditional methods such as Gaussian filtering and bilateral filtering.The proposed method suppresses noise in low-reliability regions while preserving geometric details in high-reliability regions,providing an effective polarization normal preprocessing method for 3D perception in tunnel traffic environ-ments.