Real-time Corridor Recognition for Autonomous Vehicle

Abstract

Recognition of a working environment is critical for an autonomous vehicle such as a mobile robot to guide it along corridor and to confirm its possible intelligence. Therefore it is necessary to equip a recognition system with sensor that collect environmental information. As an effective sensor a CCD camera is generally useful for all kinds of mobile robots. However, it is hard to use the CCD camera for visual feedback since it requires to acquire information in real-time, and moreover to be robust against lighting condition varieties. This research presents a corridor recognition method using unprocessed gray-scale image, termed a raw image, and a genetic algorithm (GA), without any image information conversion, to conduct the recognition process in real-time. To achieve robustness concerning lighting condition varieties, we propose a model-matching method using a representative object model designated here as surface-strips model. The robustness of the method against noise in the environment, including lighting conditions variations, and the effectiveness of the method for real-time recognition have been verified using real corridor images.

Real‐time recognition of multiple pedestrians using car‐mounted camera

“…Specifically, the surface model evaluates the fitness for the facial region (circular skin-colored region) which is the object of extraction, and the strips model evaluates it for the background (region not skin-colored). In a previous study by the authors it was verified that the surface-strips model is robust to noise other than the object, and to changes of the illumination environment, since the evaluation is based on the difference between the object and the background [12]. This property is utilized in this study.…”

“…In addition, the surface-strips model is used as a shape evaluation model in this study; this model can estimate the position more accurately by combining an evaluation for the object and an evaluation for the background. By a previous study by the authors [11,12], it was verified that the surface-strips model is robust to noise and changes of the environment, since the evaluation is based on difference information between the object and the background. By the above processing, the problem of recognition of multiple pedestrians in an input image is replaced by an optimization problem of searching for multiple local maxima of the evaluation function [13].…”

“…Consequently, this study uses the Step-GA search in order to perform the search efficiently in real time [12,13]. The Step-GA search has the feature that the generation change procedure is not repeated for each input image frame until the convergence of the GA.…”

Evolutionary recognition of marker using adaptive model-based matching

“…An evolutionary recognition process for dynamic image is that the GA is applied only one time to the newly input raw image. The difference from the GA recognition method for static image is simply that the image input process is included in the GA iteration loop, therefore every input image is evaluated only one time, and we named it as "1-step GA" (6) .…”

“…Considering the varied length of the markers, an adaptive model with variable length is presented to solve the problem. In order to avoid scanning the whole search region, a genetic algorithm (GA) [6] is utilized in the research. The position of target to be identified in the whole search region is determined by exploring the optimum solution for object function.…”