Due to the limitation of the short baseline configuration in Mobile mapping systems ofier rapid acquisition of geothe VrsATTM mobile mapping system (Tao, 1999), the accuracy referenced image sequences. With the vast amounts of image of 3~ object coordinates degrades fast along the depth (vehicle sequences collected by these systems, efficient and accumte moving) direction. Figure 1 dmws the current imaging geomeinformation extraction from image sequences becomes a try of the V & U ' ~ system. The baseline of the across-track stereo critical issue. It affects the operational cost of the application image pair is about 2 m* while the along-track stereopsis has a of the mobile mapping technology, The goal of this research 6 to 10-m baseline. In the manual object-measurement scheme, is to develop an efficient and robust approach for accumte in order to attain an increased 3D coordinate accuracy, a human object measurement from mobile mapping image sequences Operator needs to n~easure the multiple corresponding points of road corridors. The paper describes a semi-automated object visible in the consecutive image pairs and then include them measurement approach based on the use of a multiple-image into a photogrammetric intersection. This ~rocedure is timematching stmtegy. Once an object point in one image is cons~ming, and, therefore, hardly acceptable for practical measured manually, the corresponding points in consecutive image pairs can be determined automatically, and then, the Automatic image measurement of objects from imagery 3D coordinates of the object point can be calculated by phorequires matching of corresponding features. Extensive studtogrammetric intersection using correspon~ing ies have been conducted in this area. The interested reader may points. Based on the test results, the peqormance of this refer these review Papers (Lemmens, 1988; Dhond and approach was operationally satisfactory in terms of reliability, Aggarwal, 1989; Baltsavias, 1991; Jones, 1997). It has been recaccuracy, and efficiency. The effective use of stereoscopic and ognized that the application of fully automatic object measuresequential image infomation, known image geo-referncing ment is still heavily restricted and is only possible when parameters, and multiple-baseline geometry is a key to the tailored to specific objects. It is understood that the cooperadevelopment of this semi-automated object measurement tive work between the two agents-human operator and cornapproach. puter-will play a key role in intelligent information processing (Tao and Lin, 1994). A successful semi-automated Introduction approach will be beneficial not only because of its reliability ~h~ development of mobile mapping systems has evolved from but also because of its efficiency. Therefore, a semi-automated the research stage into the operational stage. The V I S A T ~ landwas researched and based mobile mapping system, originally developed by the The goal of this research was to develop an efficient and university of~algary in conjunction with ~ ~ ~ f i ~ I ~ c . , robust approach to accurate object measurement from mobile has been in commercial operations since 1996. The system image sequences corridors. There are two key integrating sensors, the Global Positioning System techniques involved in this approach: (1) multiple-image (GPS), inertial navigation systems (INS), and digital imaging matching, i.e., automatic determination of multiple correscameras, is capable of collecting geo-referenced imagery along ond ding points in image pairs; and (2) multipleroad corridors at a vehicle speed of 50 to 60 km per hour baseline-based photogrammetric intersection, i.e., optimized (schwarz et al., 1993; ~ l ~ h ~ i ~ ~ , 1996; ~ i , 1997; T ~ ~ , 1998). calculation of 3D object coordinates using multiple-baseline With the vast amounts of image sequences collected, rapid and geometry. This Paper only describesthe multi~le-image matchaccurate object measurement becomes a critical issue during ing and its test post-mission processing. It affects the operational cost of the application of mobile mapping technology. Muhiplelmage Matching Methodology By means of direct georeferencing (Schwarz and ElReview of Related Work Sheimy, 1996)~ mobile mapping image sequence scan be €Pore1, the recent past, multiple-image matching has been of conferenced to a global coordinate system. Usually, an object in a siderable interest in sn object reconstruction. ~ h ~ ~ d and road scene can appear in a number of stereo image pairs. For Aggarwal (1989) performed a cost-benefit analysis of the use example, in the VISAP system, normally three to four image 1 pairs covering the same object along the road corridor are available. With this imaging geometry, multiple-view imagery can be utilized to achieve more accurate object measurement. Photogrammetric Engineering & Remote Sensing Vol. 66, No. 12, December 2000, pp. 1477-1485. Department of Geomatics Engineering, University of Calgary, 0099-1112/00/6612-1477$3.00/
[1]
Emmanuel P. Baltsavias,et al.
Multiphoto geometrically constrained matching
,
1991
.
[2]
Nicholas Ayache,et al.
Trinocular Stereo Vision for Robotics
,
1991,
IEEE Trans. Pattern Anal. Mach. Intell..
[3]
Masatsugu Kidode,et al.
An iterative prediction and correction method for automatic stereocomparison
,
1973,
Comput. Graph. Image Process..
[4]
Rongxing Li,et al.
Mobile Mapping: An Emerging Technology for Spatial Data Acquisition
,
1997
.
[5]
T. Schenk,et al.
Reconstructing small surface patches from multiple images
,
1992
.
[6]
Hans P. Moravec.
Visual Mapping by a Robot Rover
,
1979,
IJCAI.