Comprehensive Quality Evaluation of Airborne LiDAR Data Jianwei Wu*a, Wei Yaob, Wei Chia ,Xiangdong Zhaoc a School of remote sensing and information engineering Wuhan University, Wuhan, China; b Photogrammetry and Remote Sensing, Technische Universitaet Muenchen, Munich, Germany; c Military Surveying and Mapping Team, Dalian 116000, China ABSTRACT As a new data source of remote sensing, airborne LiDAR data quality evaluation is of great importance. This paper focused on the comprehensive quality evaluation of airborne LiDAR data in the following aspects: data integrity、data accuracy 、data density、interpretation ability of intensity and the quality of spatial data products from LiDAR data, which aimed to provide a complete reference to airborne LiDAR data quality for practical applications. For data integrity, a data void extraction method based on region growing was proposed to check the completeness of spatial distribution. Data accuracy is a key quality index for LiDAR data quality. Height offset statistics among overlapping strips were calculated to evaluate the relative accuracy. Data density is an important factor influencing the products quality from LiDAR data. The average point spacing can only demonstrate the whole density of the data, whereas the local density is much more important for specific applications. In this paper, the density distribution map is used to reflect the density variations for the whole data. Besides, interpretation ability of intensity is also used to evaluate the quality of the airborne LiDAR data. Quality of DTM was used to evaluate the LiDAR data at last. Keywords: Airborne LiDAR, data integrity, data accuracy, quality evaluation
1. INTRODUCTION Airborne LiDAR (Light Detection and Ranging) has been emerging as an advanced and fast technique for earth observation. Nowadays, airborne LiDAR has been used in many different areas such as digital city, forest management, city planning, and transportation planning, gas lines planning, hydraulic applications and so on. For the efficient use and applications of airborne LiDAR, good data quality is the most important basis. However, as a new remote sensing data source, the data quality evaluation is quite different from traditional remote sensing images and the corresponding methods are still under research. Compared to traditional remote sensing imaging sensors, airborne LiDAR can receive more than one pulse return and the returns are translated into 3D object space coordinates which were called as point cloud. Besides the echo information and 3D coordinates, the LiDAR raw data include intensity information which is helpful for the interpretation of the LiDAR data. Another different aspect between the traditional remote sensing images and the LiDAR data is that the images cover the earth continuously while the LiDAR data cover the surface in a active semi-random way, which can be called as different sampling method. The semi-random sampling method caused some difficulties for airborne LiDAR data quality evaluation:(1) semi-random sampling method causes the feature points may not be observed, which makes it difficult to evaluate the external accuracy (or absolute accuracy);(2)the active semirandom sampling method may cause some data voids and different data density due to object absorption and water mirror reflections. Since airborne LiDAR data is much different from traditional remote sensing images and used in many different areas, the corresponding efficient data quality evaluation method in a practical use aspect is of great importance to the development of airborne LiDAR data applications. This paper aimed to provide a complete reference to airborne LiDAR data quality for practical applications by evaluating the airborne LiDAR data in the following aspects: data integrity、 data accuracy 、data density、interpretation ability of LiDAR intensity and the quality of spatial data products from LiDAR data. This paper was organized as following: in the introduction part, the related work about airborne LiDAR data evaluation was discussed and the data evaluation method used in this paper was simply introduced; in the methodology part, the airborne LiDAR data evaluation method was depicted in detail with some practical examples; the third part included some discussions and future work about the airborne LiDAR data evaluation. *
[email protected]
International Symposium on Lidar and Radar Mapping 2011: Technologies and Applications, edited by Xiufeng He, Jia Xu, Vagner G. Ferreira, Proc. of SPIE Vol. 8286, 828604 © 2011 SPIE · CCC code: 0277-786X/11/$18 · doi: 10.1117/12.912588 Proc. of SPIE Vol. 8286 828604-1 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
1.1 Related work about airborne LiDAR data quality evaluation The geometric data accuracy assessment is one of the most important aspects for airborne LiDAR data quality evaluation and it was almost equal to quality evaluation in the early time of LiDAR development. With the development of LiDAR technique such as data processing、hardware、applications and so on, .more about LiDAR data quality was researched such as the intensity calibration 、the repeatability of laser scanning 、the matching of the corresponding features from adjacent strips 、the quality of DTM from LiDAR 、the quality of building model from LiDAR and so on. In all, the related work about airborne LiDAR data quality evaluation can be categorized into two types: one was only concerned with the geometric accuracy including relative accuracy and absolute accuracy ;the other was concerned with both the geometric accuracy and other quality factors such as data density 、data coverage 、data void and so on. For the geometric accuracy evaluation, the methods can be categorized into LSM (least squares matching)-based approaches and non-LSM based approaches. For the LSM based approaches, they can be classified in the way the laser points are treated in the matching procedure: a) LSM is applied on the original irregular ground points by utilizing a TIN structure; e.g. [Maas 2002], [Kalian et al. 1996]. b) before applying LSM, a regular raster is interpolated from the irregular points; e.g. [Burman 2000], [Behan 2000]. According to [Maas 2002], the TIN structure avoiding of the two interpolated error is superior to the interpolated raster structure and is utilized more in practice. For the non-LSM based approaches, some used the images or linear features to evaluate the plan metric accuracy of LiDAR; some used the designed objects or ICP (iterative closest point) matching methods to evaluate the 3D accuracy. Besides, the height offset statistics were also used to check the relative accuracy of the airborne LiDAR data. Other quality factors also attracted the researchers’ and the data producers’ attention. Some used the data density map to evaluate the quality of LiDAR data and the derived DTM accuracy was assessed based on the data density [E.Ahokas, 2005, J.Luethy. 2004] and topographic features such as slope. 1.2 Outline of the method The data quality evaluation should be specific application oriented because of the different data quality requirement from the practical projects. Therefore, this paper firstly proposed an application requirement oriented airborne LiDAR data quality evaluation framework. Then, the data quality in different aspects such as data integrity, data density, data accuracy was discussed in detail. The corresponding quality indexes were derived in both local and total sides. For example, the data density was divided into local density and the whole density; the local density was calculated by the adjacent points of the evaluated point, which was in the form of density distribution map which was in different color for different local density. The whole density was calculated by the whole data point number and the corresponding coverage area, which was called mean density too. For the data accuracy, the local accuracy and mean accuracy were both used too; the relative accuracy and absolute accuracy were both evaluated. For relative accuracy, the height offset statistics were used and calculated by comparing the height differences of the areas with smooth surfaces. For absolute accuracy, a specific- target in concentric form with different height was designed and a simulation experiment was conducted to demonstrate its efficiency for absolute accuracy evaluation. Finally, the product quality derived from LiDAR data was discussed.
2. METHODOLOGIES FOR LIDAR DATA QULATIY EVALUATION 2.1 Framework of airborne LiDAR data quality evaluation Since the requirement of airborne LiDAR data quality was different for different practical projects, the framework of airborne LiDAR data quality evaluation should be extensible and application oriented. It consisted of the following parts: the basic tools for airborne LiDAR data quality evaluation, the application oriented data quality evaluation request, the main data quality evaluation procedure and quality evaluation report:
Proc. of SPIE Vol. 8286 828604-2 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
The basic tools for airborne LiDAR data quality evaluation included some key processing such as point cloud data matching, data outline extraction, the matching of LiDAR and image features, LiDAR data segmentation and so on. The tools were the basis of LiDAR data quality methods and they could be enriched according to the requirement of data quality evaluation. The application oriented data quality evaluation request means a flexible evaluation mode. The users could propose the suitable quality evaluation task request according to their project requirement to data quality based on this part. The main data quality evaluation procedure included the following steps: data integrity checking, data density checking, raw LiDAR data accuracy checking and the product quality checking. If the data integrity especially the spatial distribution can’t meet the project requirement, the following quality checking wasn’t needed. After the data integrity checking, the data density checking was conducted for different applications and projects had different requirements to LiDAR data density. If data density of the whole area or most part of the area is enough, data accuracy should be checked. To drive products with consistent accuracy, the adjacent strips must match well, which means a high relative accuracy. Besides relative accuracy, the absolute accuracy should be checked too by using absolute control information. Finally, the derived product quality should be checked to determine whether the collected airborne LiDAR data quality is good enough for the specific application to some extend. The output of final quality evaluation report. This part combined the above evaluation, derived the final comprehensive evaluation of the collected airborne LiDAR data and concluded whether it was suitable for the specific application. According to above mentioned, the specific application oriented airborne LiDAR data quality evaluation is depicted as figure 1.
C}JC](J1J
Lcbot
B'qi'i
)LJWKHIUDPHZRUNRIDLUERUQH/L'$5GDWDTXDOLW\HYDOXDWLRQ
Proc. of SPIE Vol. 8286 828604-3 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
2.2 Airborne LiDAR data integrity checking Airborne LiDAR data contain much information such as 3D coordinates, intensity, echo information, flight line information and so on. For one collected airborne LiDAR data, it should be first to check whether such information is valid and correct. Then, the data’s spatial distribution should be checked whether it covered the whole project area. Therefore, the outline of the data should be extracted by the flight line edge information or by constructing the convex hull of the data. Because of the LiDAR signal absorption by some objects, some area may have not return point in the project area, causing data voids which were harmful for specific project application. Data void detection is very important for the data integrity checking. Based on the idea of image region growing, a data void detection was proposed as following: a)
Dividing the whole data area into grids with size of average point distance;
b)
Making statistics of the grids with no LiDAR point in it;
c)
Choosing one of the grids with no LiDAR point in as the seed, then region growing was applied;
d)
For the region generated in c), if its size was larger than a threshold, then the region was taken as data void;
e)
Repeating c) and d) until all grids processed;
2.3 Airborne LiDAR data density checking Data density is a key quality index of airborne LiDAR data. There are two types of data density: local density and mean density. In usual applications, the mean density is used and it reflects the whole density of the project area. The mean data density is calculated by all the point number dividing the whole project area. However, the mean data density can’t reflect the local density information which determines the quality of processed products directly. Therefore the local data density is needed. The local data density is calculated by point number dividing area in a small area. All the local density values in an area form a density distribution map. For density in different range, different color is applied in the density distribution map. An example of density distribution map is as figure2. For density below one value, a specific color can be assigned. In figure2: density below one point per square meter is assigned blue color; therefore, for applications needing data density larger than one point per square meter, the blue areas in the density distribution map can’t meet the density requirement.
Proc. of SPIE Vol. 8286 828604-4 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
Fig. 2.Data density distribution map of airborne LiDAR data 2.4 Airborne LiDAR data accuracy checking There are two accuracy checking methods for LiDAR data: absolute accuracy checking and relative accuracy checking. For the absolute height accuracy checking, plane areas such as plane bare earth, road, highway, the roof of the plane house, basketball square and so on. Comparing the control points in the plane areas with the corresponding LiDAR points, then the absolute height accuracy can be derived .For the relative height accuracy checking, a statistics method[Latypov D,2002, Camillo Ressl 2008] for surfaces is used. First, some statistics are calculated to check whether the surface is suitable for accuracy checking based on its characteristics. Second, the mean height of the selected surfaces in overlapping areas are calculated. At last, the mean value of the corresponding surfaces is derived and taken as the reference value; therefore the height offset is derived by the difference between the reference value and the point surface value. Using the height offset and difference, a height offset map(fig 3) can be derived for the users and customers.
Proc. of SPIE Vol. 8286 828604-5 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
Height offset /m
Fig. 3.Height offset distribution map of airborne LiDAR data Compared to the height accuracy checking, the planimetric accuracy checking is much more difficult. It still lacks of efficient method for the planimetric accuracy checking of the airborne LiDAR data. Some specific designed targets are often used. The targets are always circle and con-centre shaped [Csanyi N,2007]. In practical projects, some T shape targets are also used for the absolute accuracy checking. Besides, the ortho-image is often used as reference for the absolute planimetric accuracy checking for there are many salient features in the images. 2.5 Airborne LiDAR data product quality evaluation The quality of the products derived from airborne LiDAR data is also an important index for LiDAR data quality evaluation [E.Ahokas, 2005]. Here discuss some aspects of them:1) the accuracy of DEM reflects the quality of LiDAR data directly especially in the areas with various height; 2) the completeness and accuracy of the building model from the LiDAR data which reflects the data density and accuracy directly;3) the interpretation ability of the LiDAR intensity and its usefulness for LiDAR data classification, which reflects the quality of the intensity directly. In the following fig4 ,Fig 4-b is much clear than Fig 4-a which means that the interpretation ability of Fig 4-b is better. With the deeper and deeper applications of the LiDAR data in different areas, more products such as pipe model, highway model and so on will be derived from the LiDAR data and they also can be used for the comprehensive quality evaluation of the airborne LiDAR data.
Proc. of SPIE Vol. 8286 828604-6 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
Fig 4 --a
Fig 4--b Fig 4 intensity of the LiDAR data
3. CONCLUTION AND FUTURE WORKS This paper discussed the comprehensive quality evaluation of the airborne LiDAR data in following aspects:1)data integrity;2) data density;3)data accuracy and the product quality of the LiDAR. A complete LiDAR data quality evaluation framework was proposed and a complete evaluation report should be made and given to the users and customers, which is helpful to instruct the users how to use the LiDAR data. For LiDAR data management, the quality
Proc. of SPIE Vol. 8286 828604-7 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
evaluation report should also be available for different customers. Although the LiDAR has been used in many different areas, the quality evaluation has not gained enough attention and more effort should be taken. Besides, there still exist some technology problems in LiDAR data quality evaluation. First, the absolute accuracy checking especially the planimetric accuracy still needs further research in specific target design, the matching between the image and LiDAR features, and the matching between the LiDAR points. Second, the efficient data organization methods for large LiDAR data still need deeper study. At last, the calibration of the intensity should be paid more attention and the intensity evaluation should be evaluated more combined with the geometric features. ACKNOWLEDGMENT The work was supported by Project 41001257 supported by NSFC and “the Fundamental Research Funds for the Central Universities”.
REFERENCES [1]
[2]
[3] [4]
[5]
[6]
[7]
[8]
[9]
[10]
Behan, A., 2000. On The Matching Accuracy of Rasterised Scanning Laser Altimeter Data. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Science, Amsterdam, The Netherlands, Vol. XXXIII, Part 3A, pp. 548 - 555. Burman, H., 2000. Adjustment of Laser Scanner Data for Correction of Orientation Errors. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Science, Amsterdam, The Netherlands, Vol. XXXIII, Part 3A, pp. 548 – 555. Camillo Ressl. Quality Checking Of ALS Project Using Statics of Strip Differences[C]. ISPRS 2008, Beijing.(2008) Csanyi N,C.Toth. Improvement Of LIDAR Data Accuracy Using LIDAR-Specific Ground Targets[J]. Photogrammetric Engineering & Remote Sensing, 2007，73 (4):385-396. E.Ahokas, X.Yu, H.Kaartinen. Quality of laser scanning. http://www.enge.ucl.be./EARSEL/workshops/3DRS/Paper/Ahokas.pdf .availble 2011-05-30 J.Luethy. How To Evaluate The Quality Airborne Laser Scanning Data [C]. International Archives of the Photogrammetry,Remote Sensing and Spatial information science:XXXVI-8/W2.(2004) Kilian, J., Haala, N., Eglich, M., 1996. Capture and Evaluation of Airborne Laser Scanner Data. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Science, Vienna, Austria, Vol. XXXI, Part B3. Latypov D. Estimating Relative Lidar Accuracy Information from Overlapping FlightLines [J].ISPRS Journal of Photogrammetry & Remote Sensing,56: 236—245. (2002) Maas, H.-G., 2000. Least-squares matching with airborne laser scanning data in a TIN structure. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Science, Amsterdam, The Netherlands, Vol. XXXIII, Part 3A, pp. 548 - 555. Maas, H.-G., 2002. Methods for Measuring Height and Planimetry Discrepancies in Airborne Laserscanner Data. Photogrammetric Engineering and Remote Sensing, Vol. 68, No. 9, pp. 933-940
Proc. of SPIE Vol. 8286 828604-8 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 09/17/2013 Terms of Use: http://spiedl.org/terms
