. -■ • ■ 
. 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. PartBl. Beijing 2008 
387 
conjugate points in overlapping strips will show systematic 
discrepancies. The following conclusion could be drawn from 
the simulation experiments: 
(1) The discrepancies caused by the bore-sighting offset and 
angular biases can be modelled by shifts and a rotation 
across the flight direction. Therefore, a six-parameter 
rigid-body transformation (three shifts and three rotations) 
is sufficient for modelling the introduced discrepancies and 
for aligning overlapping strips. 
(2) A rigid-body transformation, on the other hand, cannot be 
used to align the strips relative to the ground coordinate 
system. 
(3) In the presence of systematic errors in the bore-sighting 
parameters, averaging the spatial coordinates in 
overlapping strips will lead to a surface which is closer to 
the ground truth (the effects of the systematic errors are 
cancelled out or minimised). 
3. STRIP ADJUSTMENT 
The main goal of strip adjustment is to minimize the impact of 
systematic errors in the LiDAR system parameters by 
improving the compatibility among neighbouring strips. In 
addition, the estimated transformation parameters relating 
overlapping strips can be used to verify the quality of the 
system calibration. In the absence of biases in the system 
parameters, overlapping strips should coincide with each other 
without the need for any shifts or rotations. In other words, 
significant deviations from zero shifts and rotations can be used 
as an indication of the presence of systematic errors in the data 
acquisition system. Improving the compatibility between 
neighbouring strips can be viewed as the co-alignment of the 
different strips to a common reference frame. Therefore, the 
strip adjustment can be thought of as a registration procedure. 
An effective registration process should deal with four main 
issues: the registration primitives, establishing the 
correspondence between conjugate primitives, the 
transformation function relating the reference frames of the 
involved datasets, and the similarity measure which utilizes 
conjugate primitives for the estimation of the involved 
parameters in the transformation function. As it has been 
mentioned in the previous section, a six-parameter rigid-body 
transformation can be used as the transformation function 
relating overlapping strips in the presence of biases in the bore 
sighting parameters. Traditional registration procedures (e.g., 
photogrammetric Block Adjustment of Independent Models - 
BAIM) are usually based on point primitives. These primitives, 
however, are not suitable when dealing with LiDAR data since 
it is quite difficult to establish the correspondence between 
distinct points in the irregularly-distributed footprints. 
Therefore, the use of linear features is proposed in this work. In 
the following sub-sections, the extraction and matching of 
primitives will be described. Also, the similarity measure, 
which incorporates the extracted primitives for the estimation of 
the parameters of the transformation function, will be presented. 
3.1 Primitives Extraction and Matching 
Since the LiDAR footprints are irregularly distributed, no point- 
to-point correspondence can be assumed between overlapping 
strips. In this regard, other primitives must be investigated. In 
this work, the use of linear features derived from the 
intersection of neighbouring planar patches is proposed. LiDAR 
provides high redundancy in planar surfaces. Therefore, the 
plane parameters can be derived with high accuracy using an 
adjustment procedure (e.g. plane fitting). The larger the planar 
surface, the greater will be the point cloud noise reduction. 
Therefore, high accuracy linear features can be extracted by 
intersecting neighbouring planes. To do so, an environment for 
the extraction and matching of linear features in overlapping 
strips was developed. The process starts by displaying the 
LiDAR intensity images for overlapping strips where the 
operator selects an area where linear features might exist (e.g. 
roof ridge line). The user clicks on the centre of the area after 
defining the radius of a circle, within which the original LiDAR 
footprints will be extracted. It should be noted that the LiDAR 
intensity images are only used for visualization purposes. The 
user needs to establish the area of interest in one of the strips 
and the corresponding areas in the other strips are automatically 
defined. Figure 2a shows the specified area in one of the strips 
as well as the original LiDAR footprints in that area. Then a 
segmentation technique (Kim et al., 2007) is used to identify 
planar patches in the point cloud within the selected area. This 
segmentation procedure is independently run on the point cloud 
for all the overlapping strips. The outcome from such 
segmentation is aggregated sets of points representing planar 
patches in the selected area (bottom right portion in Figure 2b). 
For the linear features extraction, neighbouring planar patches 
are identified and the plane parameters determined. Then the 
neighbouring planes are intersected to produce an infinite 
straight-line. Then, using the segmented patches, the infinite 
line and a given buffer, the end points for the intersected line 
can be defined (top left portion in Figure 2b). This procedure is 
repeated for several areas within the overlap portion in the 
involved strips. 
•* 3t> *q 
* * ftfe 4 
■* \ * 
(a) 
* l 
* <? ’ 
Figure 2. Area of interest selection and LiDAR point cloud 
extraction (a), and extracted linear features by intersection of 
segmented planar patches in the area of interest (b) 
The outcome of the extraction procedure is a set of linear 
features in overlapping strips. Due to the nature of the LiDAR 
data acquisition (e.g., scan angle, surface normal, surface 
reflectivity, occlusions), there is no guarantee that there is one- 
to-one correspondence between the extracted primitives from 
overlapping strips. To solve the correspondence problem, one 
has to utilize the attributes of the extracted lines. Conjugate 
lines can be automatically matched using the normal distance, 
parallelism, and the percentage of overlap between candidate 
lines in overlapping strips (Figure 3). A graphic visualization of 
matched linear features is presented to the user for final 
confirmation of the validity of the matched primitives.