LASSIE (Land mark Acquisition SubSystem Improving Exploration) is a program which recognizes a car from several points of view. Trapezoids are formed from groups of lines extracted from a color image; instances of the car model are found by searching for sets of trapezoids which fit the constraints between parts of the car such as the windows, roof, and trunk. No information is used about the dimensions of the various parts, setting LASSIE apart from many other programs which use features extracted from images to recognize objects. 1 The remainder of the introduction sketches out the varied functions of object recognition in the navigation system of an autonomous vehicle and discusses the choice of domain for this work. The next section of the report discusses related work, followed by a section giving a detailed description of the LASSIE program. Results are shown for a number of images. The report closes with a discussion of directions for further work on LASSIE.
[1]
D. W. Thompson,et al.
Three-dimensional model matching from an unconstrained viewpoint
,
1987,
Proceedings. 1987 IEEE International Conference on Robotics and Automation.
[2]
Rodney A. Brooks,et al.
Symbolic Reasoning Among 3-D Models and 2-D Images
,
1981,
Artif. Intell..
[3]
Y. Goto,et al.
CMU Sidewalk Navigation System: A Blackboard-Based Outdoor Navigation System Using Sensor Fusion with Colored-Range Images
,
1986,
FJCC.
[4]
Yuichi Ohta.
A Region-Oriented Image-Analysis System by Computer
,
1980
.
[5]
Arnie Kwong,et al.
Fall Joint Computer Conference
,
1987,
Data Base.
[6]
Ramesh C. Jain,et al.
Three-dimensional object recognition
,
1985,
CSUR.