| GISdevelopment.net ---> Technology ---> Synthetic Aperture Radar (SAR) |
Principle of High-resolution Airborne SAR Aerial triangulation supported by sparse GCPs
Lei Pang1,3, Jixian Zhang1, Mingbo Zhang3 1
Chinese Academy of Surveying and Mapping
Tel: 86 10 88229545, Fax: 0086 10 68211420
Email: Panglei.mail@163.com, stecsm@public.bta.net.cn
Address: Beitaiping road 16, Haidian district, Beijing, P. R. China 100039
2Shandong University of Science and Technology
Phone: +86 538 6226813(0), Fax :( 0538) 6226163
Address: Institute of Geosciences, Shandong University of Science and Technology, Taian, Shandong, P. R. China, 271019
3Institute of Geographical Sciences and Natural Resources Research, CAS
Phone: +86 10 64861632-184, Fax :0086 10 64851844
Email: zhangmb@lreis.ac.cn
Address: Datun Road 3, Chaoyang district, Beijing, P. R. China 100101
Lei Pang1,3, Jixian Zhang1, Mingbo Zhang3 1
Chinese Academy of Surveying and Mapping
Tel: 86 10 88229545, Fax: 0086 10 68211420
Email: Panglei.mail@163.com, stecsm@public.bta.net.cn
Address: Beitaiping road 16, Haidian district, Beijing, P. R. China 100039
2Shandong University of Science and Technology
Phone: +86 538 6226813(0), Fax :( 0538) 6226163
Address: Institute of Geosciences, Shandong University of Science and Technology, Taian, Shandong, P. R. China, 271019
3Institute of Geographical Sciences and Natural Resources Research, CAS
Phone: +86 10 64861632-184, Fax :0086 10 64851844
Email: zhangmb@lreis.ac.cn
Address: Datun Road 3, Chaoyang district, Beijing, P. R. China 100101
1 Introduction
Presently, the Western Areas Mapping Mission (WAMM) is executing to map the relief in cloudy or soupy mountainous western area in China, where the high-resolution airborne SAR imagery has often been chosen as the irreplaceable remote sensing data resource. At the same time, the mission that how to acquire enough accuracy and quantity of ground control points (GCPs), which are difficult to be measured practically, has become an urgent demand for mapping tasks.
Although, the study to decrease the demand of GCPs in mosaicing RADARSAT or ERS SAR images or densing GCPs network has been done by Thierry Toutin(2003), Léonard Denise (1998) and Youssef Belgued (1998), there has no corresponding study on the method or models of high-resolution airborne SAR images triangulation.
Comparing with satellite-borne SAR images triangulation, the essence of high-resolution airborne SAR triangulation is still the 3D space positioning from SAR images (Chang Benyi, 2002) supported by only sparse GCPs or without GCPs. But, the method and mathematic models must be based on the characteristics of airborne SAR imaging conditions and flight tracks. So, we analyses the airborne SAR triangulation conditions, put forward the corresponding mathematic models and introduce the high-resolution airborne SAR triangulation test in China in the following parts.
2 Airborne SAR Triangulation Conditions
We all know that InSAR technique can extract the 3D terrain information, but, without GCPs, the absolute space positions of the ground objects are still unknown to us. So, different from InSAR technique, we studied the high-resolution airborne SAR images triangulation conditions, which can be summarized as the follows:
3.1 Independent model unit configuration
Fig.1 The independent model unit for airborne SAR imagery triangulation
To construct the independent model unit, we must pre-prepare the airborne SAR imagery. Firstly, we divide the imagery paths into many segments with some length, for example, the length of one synthetic aperture, and 20%~30% overlapping with adjacent segments in the same path. Then, take the flight direction as the X-axis, and range direction as Y. And, according to our flight course design, there is 60 percent overlap area between the two segments of independent model unit from adjacent imagery paths, so, the segments of the same overlap area in adjacent paths will construct the independent model unit for airborne SAR imagery triangulation.
3.2 Airborne SAR triangulation model based on independent models
Under the condition of only sparse GCPs available around the mapping area, the vector model of independent models based airborne SAR triangulation see as Fig.2. To make the processing simple, we adopt the stereo-pairs from the adjacent flight paths to buildup the independent model unit. And, each object P in the independent model unit will have the unique relative model coordinates, which can be transform into the absolute ground coordinates (X', Y', Z') through sequent space transform matrix. A more detailed explanation is as follows.
Fig.2 Independent models based airborne SAR triangulation vector model
The structure of an independent model unit is based on the two adjacent flight paths. The designed flight course plan satisfies the demand of 60% overlap degree, and the overlap parts between two imaging paths must be divided into several segments, two segments of the same area comprising a independent model. The model coordinates of ground point P based on the three axes: X- azimuth direction, Y-range direction, Z-sky direction, with point O as the origin, and can be computed according for the formula (1) as follows:
Where, (X, Y, Z) are model coordinates of P; Si are space position of radar station on i flight trajectory, corresponding to the zero Doppler frequency and the position of ground object point P; Vi are corresponding flight velocity.
Thus, all of the independent models connect into a block network by tie points between adjacent models, once we have computed out the model coordinates of ground point P, for example, in Model i, we can obtain the absolute ground coordinates through block adjustment. The transform mathematic relations see as formula (2):
Where,?i is the ground coordinate scale of Model i; Ri is the transform matrix from Model i to the ground coordinates; O'is the origin of ground coordinates.
What should be pointed out is that, independent models based SAR triangulation demands only sparse GCPs around the mapping area, but the using of a lot of tie points maybe bring computation burden.
3.3 GPS/INS supported airborne SAR triangulation model without GCPs
An adaptive method of airborne SAR triangulation is supported by GPS and INS inertial navigation data. GPS data provided with the space position coordinates of the radar station Si on flight trajectory, and INS data present the corresponding instant velocity Vi. So that, airborne SAR triangulation become simple and direct when provided with the flight trajectory control vector parameters. The vector model for it sees as Fig. 3.
Fig.3 GPS/INS supported airborne SAR triangulation vector model
The 3D ground coordinates can be computed out through formula (1) without GCPs supported by flight control parameters, and what is important is the preprocessing of GPS/INS data. For acquiring accurate ground objects 3D coordinates, the spatial and temporal calibration of GPS and INS data is essential. Then, we can obtain the instant values through interpolator computation, and absolute space coordinates from least square adjustment.
4 Test Site and Flight Design
We chose a test site on the southwest of Zhengzhou, Henan Province, in China. The total area is about 20km×80km,and in the western area, almost 35 percent of total area with height from 500m to 1400m, are covered with poplar and pinaster. In other plain area, cultivated land is the most land use. Our airborne SAR remote sensing data set are acquired from the SAR sensor system carried on YUN-12 type airplane (see as Fig.4), with single-antenna, 1m-resolution and same-side stereo imaging mode provided with GPS and inertial navigation system equipments synchronously. We designed the fight courses with 60% overlap between adjacent paths see as Fig.5, At the same time, we distributed and mapped 12 GCPs around the area (see as Fig.6) utilizing homemade aluminous radar corner-reflector. and acquired the test data source of one path just as showed in Fig.7.
Fig.4 YUN-12 airplane
Fig.5 the flight course
Fig.6 12 GCPs distribution
Fig.7 the test data source of one path
Though, it has not the final accuracy analysis of the test by now, but the first results has verified the mathematical models and presented the potential of mapping application.
6 Conclusions
In this paper, the application perspective of high-resolution airborne SAR triangulation technique and its corresponding mathematical vector models has been described in detail. Though, supported by sparse GCPs distributed around the test area, the first results are not so satisfying, we still found that independent models based and GPS/INS supported airborne SAR triangulation technique can reduce the demand of GCPs, and realize dense GCP network of some area difficult to survey, or SAR imagery mosaicing of large area.
References
Presently, the Western Areas Mapping Mission (WAMM) is executing to map the relief in cloudy or soupy mountainous western area in China, where the high-resolution airborne SAR imagery has often been chosen as the irreplaceable remote sensing data resource. At the same time, the mission that how to acquire enough accuracy and quantity of ground control points (GCPs), which are difficult to be measured practically, has become an urgent demand for mapping tasks.
Although, the study to decrease the demand of GCPs in mosaicing RADARSAT or ERS SAR images or densing GCPs network has been done by Thierry Toutin(2003), Léonard Denise (1998) and Youssef Belgued (1998), there has no corresponding study on the method or models of high-resolution airborne SAR images triangulation.
Comparing with satellite-borne SAR images triangulation, the essence of high-resolution airborne SAR triangulation is still the 3D space positioning from SAR images (Chang Benyi, 2002) supported by only sparse GCPs or without GCPs. But, the method and mathematic models must be based on the characteristics of airborne SAR imaging conditions and flight tracks. So, we analyses the airborne SAR triangulation conditions, put forward the corresponding mathematic models and introduce the high-resolution airborne SAR triangulation test in China in the following parts.
2 Airborne SAR Triangulation Conditions
We all know that InSAR technique can extract the 3D terrain information, but, without GCPs, the absolute space positions of the ground objects are still unknown to us. So, different from InSAR technique, we studied the high-resolution airborne SAR images triangulation conditions, which can be summarized as the follows:
- Airborne SAR images stereo-pairs based triangulation models;
- 60% overlap degree between the adjacent flight paths;
- Independent model block adjustment to compute target 3D coordinates only supported by sparse GCPs available;
- GPS/INS data supported airborne SAR triangulation without GCPs;
3.1 Independent model unit configuration
Fig.1 The independent model unit for airborne SAR imagery triangulation
To construct the independent model unit, we must pre-prepare the airborne SAR imagery. Firstly, we divide the imagery paths into many segments with some length, for example, the length of one synthetic aperture, and 20%~30% overlapping with adjacent segments in the same path. Then, take the flight direction as the X-axis, and range direction as Y. And, according to our flight course design, there is 60 percent overlap area between the two segments of independent model unit from adjacent imagery paths, so, the segments of the same overlap area in adjacent paths will construct the independent model unit for airborne SAR imagery triangulation.
3.2 Airborne SAR triangulation model based on independent models
Under the condition of only sparse GCPs available around the mapping area, the vector model of independent models based airborne SAR triangulation see as Fig.2. To make the processing simple, we adopt the stereo-pairs from the adjacent flight paths to buildup the independent model unit. And, each object P in the independent model unit will have the unique relative model coordinates, which can be transform into the absolute ground coordinates (X', Y', Z') through sequent space transform matrix. A more detailed explanation is as follows.
Fig.2 Independent models based airborne SAR triangulation vector model
The structure of an independent model unit is based on the two adjacent flight paths. The designed flight course plan satisfies the demand of 60% overlap degree, and the overlap parts between two imaging paths must be divided into several segments, two segments of the same area comprising a independent model. The model coordinates of ground point P based on the three axes: X- azimuth direction, Y-range direction, Z-sky direction, with point O as the origin, and can be computed according for the formula (1) as follows:
Where, (X, Y, Z) are model coordinates of P; Si are space position of radar station on i flight trajectory, corresponding to the zero Doppler frequency and the position of ground object point P; Vi are corresponding flight velocity.
Thus, all of the independent models connect into a block network by tie points between adjacent models, once we have computed out the model coordinates of ground point P, for example, in Model i, we can obtain the absolute ground coordinates through block adjustment. The transform mathematic relations see as formula (2):
Where,?i is the ground coordinate scale of Model i; Ri is the transform matrix from Model i to the ground coordinates; O'is the origin of ground coordinates.
What should be pointed out is that, independent models based SAR triangulation demands only sparse GCPs around the mapping area, but the using of a lot of tie points maybe bring computation burden.
3.3 GPS/INS supported airborne SAR triangulation model without GCPs
An adaptive method of airborne SAR triangulation is supported by GPS and INS inertial navigation data. GPS data provided with the space position coordinates of the radar station Si on flight trajectory, and INS data present the corresponding instant velocity Vi. So that, airborne SAR triangulation become simple and direct when provided with the flight trajectory control vector parameters. The vector model for it sees as Fig. 3.
Fig.3 GPS/INS supported airborne SAR triangulation vector model
The 3D ground coordinates can be computed out through formula (1) without GCPs supported by flight control parameters, and what is important is the preprocessing of GPS/INS data. For acquiring accurate ground objects 3D coordinates, the spatial and temporal calibration of GPS and INS data is essential. Then, we can obtain the instant values through interpolator computation, and absolute space coordinates from least square adjustment.
4 Test Site and Flight Design
We chose a test site on the southwest of Zhengzhou, Henan Province, in China. The total area is about 20km×80km,and in the western area, almost 35 percent of total area with height from 500m to 1400m, are covered with poplar and pinaster. In other plain area, cultivated land is the most land use. Our airborne SAR remote sensing data set are acquired from the SAR sensor system carried on YUN-12 type airplane (see as Fig.4), with single-antenna, 1m-resolution and same-side stereo imaging mode provided with GPS and inertial navigation system equipments synchronously. We designed the fight courses with 60% overlap between adjacent paths see as Fig.5, At the same time, we distributed and mapped 12 GCPs around the area (see as Fig.6) utilizing homemade aluminous radar corner-reflector. and acquired the test data source of one path just as showed in Fig.7.
Fig.4 YUN-12 airplane
Fig.5 the flight course
Fig.6 12 GCPs distribution
Fig.7 the test data source of one path
Though, it has not the final accuracy analysis of the test by now, but the first results has verified the mathematical models and presented the potential of mapping application.
6 Conclusions
In this paper, the application perspective of high-resolution airborne SAR triangulation technique and its corresponding mathematical vector models has been described in detail. Though, supported by sparse GCPs distributed around the test area, the first results are not so satisfying, we still found that independent models based and GPS/INS supported airborne SAR triangulation technique can reduce the demand of GCPs, and realize dense GCP network of some area difficult to survey, or SAR imagery mosaicing of large area.
References
- Léonard Denise et al. Application of radar space triangulation for 3D localization, European conference on synthetic aperture radar, 25-27 May 1998, Friedrichshafen, Germany. pp.351-354.
- Youssef Belgued et al. Application of radar triangulation to the calibration of interferometric DEM, Geoscience and remote sensing symposium proceedings, 1998. IGARSS’98.1998 IEEE international published :Vol.5, pp.2665-2667.
- Toutin Th. Path processing and block adjustment with RADARSAT-1 SAR images, IEEE Transactions on Geoscience & Remote Sensing, 2003, 41(10), pp. 2320-2328.
- Toutin Th. Error tracking of Radargrammetric DEM from RADARSAT Images, IEEE Transactions on Geoscience & Remote Sensing , 1999, 37(5), pp. 2227-2238.
- Thierry Toutin. Block bundle adjustment of IKONOS in-track images, International Journal of Remote Sensing, Feb. 2003, 24(4): 851-857.
- Chang Benyi; Fang Yong. The principles of positioning with space-borne SAR images, Geoscience and Remote Sensing Symposium, 2002. IGARSS '02. 2002 IEEE International Pages: 1950 - 1952 vol.4 , 24-28 June 2002.
- De Grandi, G.F.; Mayaux, P.; Simard, M.; Saatchi, S. Fusion of the L-band GRFM and C-band CAMP wide area Central Africa radar mosaics: a new data set with unprecedented potential for regional scale vegetation mapping, Geoscience and Remote Sensing Symposium, 2000. Proceedings. IGARSS 2000. IEEE 2000 International Pages: 4 - 6, vol.1, 24-28 July 2000.