Normalization of LIDAR intensity

There are a lot of research today about laser intensity, because we find a huge potential at Lidar intensity datasets. These information could help by different processes, like segmentation, classification or visualization. All of lidar applicatnions based on x,y,z parameters, and I think, in the future, we will use this extra information (i=intensity), which could mean 4th dimension of lidar.

Normalization of intensity is a very useful direction, here is a new paper:

NORMALIZATION OF LIDAR INTENSITY DATA BASED ON
RANGE AND SURFACE INCIDENCE ANGLE
B. Jutzi, H. Gross

" ABSTRACT:
The analysis of airborne laser scanner data to extract surface features is of great interest in photogrammetric research. Especially for applications based on airborne measurements, where the intensity is crucial (e.g. for segmentation, classification or visualization purposes), a normalization considering the beam divergence, the incidence angle and the atmospheric attenuation is required. Our investigations show that the same material of a surface (e.g. gabled roof) yields to different measured values for the intensity. These values are strongly correlated to the incidence angle of the laser beam on the surface. Therefore the intensity value is improved with the incidence angle derived by the sensor and object position as well as its surface orientation. The surface orientation is estimated by the eigenvectors of the covariance matrix including all object points inside a close environment. Further the atmospheric attenuation is estimated. The adaptation of vegetation areas is disregarded in this study. After these improvements the intensity does no longer depend on the incidence angle but may be influenced by the material of the object surface only. For surface modelling the Phong model is introduced, considering diffuse and specular backscattering characteristics of the surface. A measurement campaign was carried out to investigate the influences of the incidence angle on the measured intensity. By considering the incidence angle and the distance between sensor and object the laser data captured from different flight paths (data stripes) can be successfully fused. In our experiments it could be shown that the radiometric normalization of the intensity for the investigated areas are improved.


DISCUSSION AND CONCLUSION
For assessing the normalized intensity values nearly
homogenous regions have been selected interactively. The
variation parameter is selected as measure for the comparison of
the values before and after normalization. Mean and standard
deviation of this measure over all regions decreases by the
normalization, especially if all flights are included. For pulsed
laser systems a strong intensity variation could be observed. The
intensity inside a region shows a high variance even for a
constant incidence angle. This may caused by material features
or local surface effects. For nearly all regions the results for the intensity have been improved, even with region disturbances on the roofs like chimneys. The Lambertian model fits the investigated surfaces well. For specular reflectance based on the Phong model no significant improvements could be derived. This might depend on the diffuse backscattering characteristic of the material. Further investigations for this study were not possible because only one data set with surfaces of a single material with different orientations was available. For terrestrial laser data enhanced results can be expected, with a lower variance of the intensity, due to a better signal-to-noise ratio for the measured data. This paper proposes a general approach for intensity normalization considering diffuse and specular scattering characteristics of the surface. This assumption should be proved in future by investigating reference targets where the
backscattering characteristic is known or could be measured by reference measurements."