Real Time Object Detection For Smart Vehicles
REAL-TIME OBJECT DETECTION FOR "SMART" VEHICLES
D.M. Gavrila
V. Philomin
Image Understanding Systems
DaimlerChrysler Research
Ulm 89081, Germany dariu.gavrila@DaimlerChrysler.com Computer Vision Laboratory
University of Maryland
College Park, MD 20742, U.S.A. vasi@cs.umd.edu ABSTRACT
This paper presents an e cient shape-based object detection method based on Distance Transforms and describes its use for real-time vision on-board vehicles.
The method uses a template hierarchy to capture the variety of object shapes; e cient hierarchies can be generated o ine for given shape distributions using stochastic optimization techniques i.e. simulated annealing . Online, matching involves a simultaneous coarse-to- ne approach over the shape hierarchy and over the transformation parameters. Very large speedup factors are typically obtained when comparing this approach with the equivalent brute-force formulation; we have measured gains of several orders of magnitudes.
We present experimental results on the real-time detection of tra c signs and pedestrians from a moving vehicle. Because of the highly time sensitive nature of these vision tasks, we also discuss some hardwarespeci c implementations of the proposed method as far as SIMD parallelism is concerned.
if he makes a wrong turn in a one-way street or if he is speeding. Alternatively, a system to detect pedestrians might reduce the accident rate by taking either passive or active measures to deal with upcoming collisions.
In this paper, we present a shape-based method which can be used for that purpose; it is general enough to detect objects of arbitrary shapes. Models or other parametrizations need not to be established explicitly, which is an advantage when dealing with non-rigid objects such as pedestrians. Instead, the method is able to generate an e cient representation from example shapes o -line; matching proceeds on-line using a novel variant of Distance Transform DT - based matching.
The outline of the
D.M. Gavrila
V. Philomin
Image Understanding Systems
DaimlerChrysler Research
Ulm 89081, Germany dariu.gavrila@DaimlerChrysler.com Computer Vision Laboratory
University of Maryland
College Park, MD 20742, U.S.A. vasi@cs.umd.edu ABSTRACT
This paper presents an e cient shape-based object detection method based on Distance Transforms and describes its use for real-time vision on-board vehicles.
The method uses a template hierarchy to capture the variety of object shapes; e cient hierarchies can be generated o ine for given shape distributions using stochastic optimization techniques i.e. simulated annealing . Online, matching involves a simultaneous coarse-to- ne approach over the shape hierarchy and over the transformation parameters. Very large speedup factors are typically obtained when comparing this approach with the equivalent brute-force formulation; we have measured gains of several orders of magnitudes.
We present experimental results on the real-time detection of tra c signs and pedestrians from a moving vehicle. Because of the highly time sensitive nature of these vision tasks, we also discuss some hardwarespeci c implementations of the proposed method as far as SIMD parallelism is concerned.
if he makes a wrong turn in a one-way street or if he is speeding. Alternatively, a system to detect pedestrians might reduce the accident rate by taking either passive or active measures to deal with upcoming collisions.
In this paper, we present a shape-based method which can be used for that purpose; it is general enough to detect objects of arbitrary shapes. Models or other parametrizations need not to be established explicitly, which is an advantage when dealing with non-rigid objects such as pedestrians. Instead, the method is able to generate an e cient representation from example shapes o -line; matching proceeds on-line using a novel variant of Distance Transform DT - based matching.
The outline of the
References: In International Joint Conference on Arti cial Intelligence, pages 659 663, 1977. 3 G. Piccioli et. al. Robust method for road sign detection and recognition. Image and Vision Computing, 14:209 223, 1998. :40 48, 1998. 6 D. Gavrila. Multi-feature template matching using distance transforms. In International Conference on Pattern Recognition, pages 439 444, Brisbane, 1998. :82 98, 1999. 9 :850 863, 1993. 9 S. Kirkpatrick, Jr. C.D. Gelatt, and M.P. Vecchi. Optimization by simulated annealing. Science, 220:671 680, 1993. and Pattern Recognition, Santa Barbara, CA, 1998. Pattern Recognition, pages 193 199, San Juan, 1997. 747, 1994. Vision, pages 457 464, 1995.