Full-waveform airborne laser scanning has shown increasing utility for earth feature extraction through enhanced physical object recognition. Such data provides users with additional physical observables of the earth's surface. This information has the potential to be exploited alongside geometric information to overcome signal inconsistencies between overlapping flightlines and to improve existing segmentation methodologies. However, because the laser signal is influenced by many variables during travel between the sensor and the target, direct use of this information is not recommended without performing echo amplitude normalization as a function of the incidence angle effect. While existing normalization approaches have proven to be valid over planar features, they tend to perform poorly over nonplanar surfaces. This is primarily due to the lack of robust local surface normal estimation. Realizing these shortcomings, this paper proposes a new echo amplitude normalization approach, where each echo's incidence angle is estimated based on illumination direction and local surface orientation. The local surface orientation estimation method computes the normal to an individual point using the minimum number of points. 3-D moment invariants are used to deliver the normal vector using a weighting function. Thereafter, a vector dot product in 3-D space is adopted to check planarity, ensuring robustness. This method is shown to overcome the weaknesses of existing approaches, performing strongly in challenging areas of rough natural terrain, as well as for planar features. Consequently, this method could be adopted in order to compensate incidence angle effects in any laser scanning physical signals for a range of downstream radiometric calibration and point cloud segmentation applications.
展开▼
