In this paper we propose an innovative method for the automatic detection and tracking of road traffic signs using an onboard
stereo camera. It involves a combination of monocular and stereo analysis strategies to increase the reliability of
the detections such that it can boost the performance of any traffic sign recognition scheme. Firstly, an adaptive color
and appearance based detection is applied at single camera level to generate a set of traffic sign hypotheses. In turn,
stereo information allows for sparse 3D reconstruction of potential traffic signs through a SURF-based matching
strategy. Namely, the plane that best fits the cloud of 3D points traced back from feature matches is estimated using a
RANSAC based approach to improve robustness to outliers. Temporal consistency of the 3D information is ensured
through a Kalman-based tracking stage. This also allows for the generation of a predicted 3D traffic sign model, which is
in turn used to enhance the previously mentioned color-based detector through a feedback loop, thus improving detection
accuracy. The proposed solution has been tested with real sequences under several illumination conditions and in both
urban areas and highways, achieving very high detection rates in challenging environments, including rapid motion and
significant perspective distortion.
In recent years advanced driver assistance systems (ADAS) have received increasing interest to confront car accidents.
In particular, video processing based vehicle detection methods are emerging as an efficient way to address accident
prevention. Many video-based approaches are proposed in the literature for vehicle detection, involving sophisticated
and costly computer vision techniques. Most of these methods require ad hoc hardware implementations to attain
real-time operation. Alternatively, other approaches perform a domain change --via transforms like FFT, inverse
perspective mapping (IPM) or Hough transform-- that simplifies otherwise complex feature detection. In this work, a
cooperative strategy between two domains, the original perspective space and the transformed non-perspective space
computed trough IPM, is proposed in order to alleviate the processing load in each domain by maximizing the
information exchange between the two domains. A system is designed upon this framework that computes the location
and dimension of the vehicles in a video sequence. Additionally, the system is made scalable to the complexity imposed
by the scenario. As a result, real-time vehicle detection and tracking is accomplished in a general purpose platform. The
system has been tested for sequences comprising a wide variety of scenarios, showing robust and accurate performance.