Nasser Kehtarnavaz

Visual control of an autonomous vehicle (BART)-the vehicle-following problem

The authors consider the problem of vehicle-following including automatic steering and speed control of an autonomous vehicle following the motion of a lead vehicle. A visual control system for vehicle-following is presented. The system consists of the following modules: image processing, recursive filtering, and a driving command generator. First, the range and heading signal of the lead vehicle are obtained by visually identifying a unique tracking feature on the lead vehicle. Based upon this information, appropriate steering wheel and speed commands for driving are generated, which are then downloaded and executed on a microprocessor controller. The visual control system was tested on BART (Binocular Autonomous Research Team), a testbed vehicle developed at Texas A&M University for autonomous mobility. Successful full-scale test runs have been accomplished for speeds up to 20 mi/h. >

Stop-sign recognition based on color/shape processing

This paper presents a robust vision-based stop-sign reconition technique based on sequential processing of color and shape. The primary red-green-blue color coordinate system is first transformed into the saturation-hue-brightness color coordinate system. This color coordinate system allows the red color area of a stop sign to be bounded under various brightness conditions caused by weather, sun angle, or shadows. A combination of a median filter, a morphological filter, Sobel edge operator, and Hough transform is then employed to obtain the boundary contour. It is demonstrated that the parameters of eight straight lines representing the octagonal sides are sufficient for this purpose. Experimental results indicate that stop signs are successfully distinguished from other traffic sighs and background clutter.

Traffic sign recognition in noisy outdoor scenes

This paper presents a noise-tolerant traffic sign recognition method by using both color and shape attributes. First a discriminant analysis is carried out to obtain the color coordinate system giving the best separation between traffic signs and other objects in the scene. Then a recognition algorithm is devised by cascading three modules: an ART2 neural network module to perform color segmentation, a log-polar-exponential grid and Fourier transformation module to extract invariant traffic sign signatures, a backpropagation neural network module to classify such signatures. The performance of this method is evaluated by examining the effect of various noise sources, which may occur in actual outdoor scenes, on the recognition rate. The results obtained indicate the noise-tolerance of the developed methodology.

A real-time histographic approach to road sign recognition

This paper presents the development and real-time implementation of an algorithm capable of recognizing stop, yield, and do-not-enter traffic warning signs. It consists of six modules: color segmentation, edge localization, RGB differencing, edge detection, histograph extraction, and classification. RGB transformed pixels are sparsely segmented and sequentially XOR-ed to localize edge areas. RGB differencing together with maxima edge detection is then deployed to locate edges in these areas. Recognition is achieved based on the angular histographic attribute extracted by a semi-rectangular histographic mask. All the modules are implemented on the TMS320C40 DSP processor allowing video data captured by a video camera to be processed in real-time. The devised real-time processing platform has led to an understanding of various environmental effects on video data.

An invariant traffic sign recognition system based on sequential color processing and geometrical transformation

One of the most noteworthy problems associated with conventional pattern recognition methods is that it is not easy to extract feature vectors from images which are not translation, rotation, and scale change invariant in outdoor noisy environments. This paper describes the development of an invariant traffic sign recognition system capable of tolerating the above variations. The signs are restricted to three types of warning signs and are all of red color. The developed method is insensitive to brightness changes as well as invariant to translation, rotation, scale change, and noise. The architecture of this system is based upon neural network supervised learning after geometrical transformations have been applied. The performance of this system is compared with other invariant approaches in terms of the percentage of correct decisions in outdoor noisy environments.<<ETX>>

Determining number of clusters and prototype locations via multi-scale clustering

Abstract In clustering algorithms, it is usually assumed that the number of clusters is known or given. In the absence of such a priori information, a procedure is needed to find an appropriate number of clusters. This paper presents a clustering algorithm that incorporates a mechanism for finding the appropriate number of clusters as well as the locations of cluster prototypes. This algorithm, called multi-scale clustering, is based on scale-space theory by considering that any prominent data structure ought to survive over many scales. The number of clusters as well as the locations of cluster prototypes are found in an objective manner by defining and using lifetime and drift speed clustering criteria. The outcome of this algorithm does not depend on the initial prototype locations that affect the outcome of many clustering algorithms. As an application of this algorithm, it is used to enhance the Hough transform technique.

A framework for estimation of motion parameters from range image

Abstract A framework for inferring motion parameters of a moving object from its range image frames is presented. This framework is based on a correspondence approach where correspondences of view-invariant patches on the object surface are sought. From each range image, a relational graph is set up such that a node represents a view-invariant patch and an edge the adjacency of two such patches. Then, the largest subgraph common to two relational graphs is detected. This subgraph represents the same region visible in the corresponding range images. Finally, the motion parameters are estimated by minimizing the sum of squares error between the centers of mass of the common nodes in the largest matched subgraph. Examples of real range images show the merit of this framework.

A string matching computer-assisted system for dolphin photoidentification.

AbstractThis paper presents a syntactic/semantic string representation scheme as well as a string matching method as part of a computer-assisted system to identify dolphins from photographs of their dorsal fins. A low-level string representation is constructed from the curvature function of a dolphins fin trailing edge, consisting of positive and negative curvature primitives. A high-level string representation is then built over the low-level string via merging appropriate groupings of primitives in order to have a less sensitive representation to curvature fluctuations or noise. A family of syntactic/semantic distance measures between two strings is introduced. A composite distance measure is then defined and used as a dissimilarity measure for database search, highlighting both the syntax (structure or sequence) and semantic (attribute or feature) differences. The syntax consists of an ordered sequence of significant protrusions and intrusions on the edge, while the semantics consist of seven attributes extracted from the edge and its curvature function. The matching results are reported for a database of 624 images corresponding to 164 individual dolphins. The identification results indicate that the developed string matching method performs better than the previous matching methods including dorsal ratio, curvature, and curve matching. The developed computer-assisted system can help marine mammalogists in their identification of dolphins, since it allows them to examine only a handful of candidate images instead of the currently used manual searching of the entire database.

An efficient recursive algorithm for time-varying Fourier transform

An efficient recursive algorithm for computing the time-varying Fourier transform (TVFT) or short-time Fourier transform (STFT) of a time sequence is presented. In this approach, instead of excluding the old samples, their importance is diminished by using all-pole moving windows. This recursive algorithm requires about one half of the computation and storage of the Amins algorithm. The resulting TVFT does not possess any sidelobes. The performance of the algorithm is illustrated by two numerical examples. >

A transportable neural-network approach to autonomous vehicle following

This paper presents the development and testing of a neural-network module for autonomous vehicle following. Autonomous vehicle following is defined as a vehicle changing its own steering and speed while following a lead vehicle. The strength of the developed controller is that no characterization of the vehicle dynamics is needed to achieve autonomous operation. As a result, it can be transported to any vehicle regardless of the nonlinear and often unobservable dynamics. Data for the range and heading angle of the lead vehicle were collected for various paths while a human driver performed the vehicle following control function. The data was collected for different driving maneuvers including straight paths, lane changing, and right/left turns. Two time-delay backpropagation neural networks were then trained based on the data collected under manual control-one network for speed control and the other for steering control. After training, live vehicle following runs were done under the neural-network control. The results obtained indicate that it is feasible to employ neural networks to perform autonomous vehicle following.