Traffic Light Detection Based on Multi-feature Segmentation and Online Selecting Scheme
04 Dec 2014-pp 204-209
TL;DR: A new simple method is proposed called edged-color image to segment candidate traffic light back board regions from even complex background, which is a way to enhance edge information in a color image substantially.
Abstract: This paper is concerned with vision-based traffic light detection by using multi-feature to segment one single image and an online selecting scheme. First, we propose a new simple method called edged-color image to segment candidate traffic light back board regions from even complex background, which is a way to enhance edge information in a color image substantially. Second, an online selecting scheme is used to calculate whether two or more candidate regions can be combined together. Those with faulty score closer to zero will be regarded as a traffic light. In addition, arrow light will be recognized from the traffic light. Applying the method above can mostly solve the problems as different light intensity, complex background, vehicle tail light, etc.
Citations
More filters
Proceedings Article•
01 Jan 2009
63 citations
TL;DR: A practical framework of hardware and software is proposed to reveal the external configuration and internal mechanism of an autonomous vehicle—a typical intelligent system and the performance of project cocktail is proven to be considerably better in terms of transmission delay and throughput.
Abstract: Architecture design is one of the most important problems for an intelligent system. In this paper, a practical framework of hardware and software is proposed to reveal the external configuration and internal mechanism of an autonomous vehicle—a typical intelligent system. The main contributions of this paper are as follows. First, we compare the advantages and disadvantages of three typical sensor plans and introduce a general autopilot for a vehicle. Second, we introduce a software architecture for an autonomous vehicle. The perception and planning performances are improved with the help of two inner loops of simultaneous localization and mapping. An algorithm to enlarge the detection range of the sensors is proposed by adding an inner loop to the perception system. A practical feedback to restrain mutations of two adjacent planning periods is also realized by the other inner loop. Third, a cross-platform virtual server (named project cocktail) for data transmission and exchange is presented in detail. Through comparisons with the robot operating system, the performance of project cocktail is proven to be considerably better in terms of transmission delay and throughput. Finally, a report on an autonomous driving test implemented using the proposed architecture is presented, which shows the effectiveness, flexibility, stability, and low-cost of the overall autonomous driving system.
44 citations
Cites background from "Traffic Light Detection Based on Mu..."
...15 is derived from an improved version of [50], mainly by adding a verification part to remove false positives and improving arrow light recognition....
[...]
TL;DR: An effective architecture that integrates a vision system with an accurate positioning system and an extended digital map for traffic light recognition is proposed and can recognize traffic lights with 98.68% precision, 92.73% recall, and 95.52% accuracy.
Abstract: Traffic light recognition is being intensively researched for the purpose of reducing traffic accidents at intersections and realizing autonomous driving. However, conventional vision-based approaches have several limitations due to full image scanning, always-on operation, various different types of traffic lights, and complex driving environments. In particular, it might be impossible to recognize a relevant traffic light among multiple traffic lights at multiple intersections. To overcome such limitations, we propose an effective architecture that integrates a vision system with an accurate positioning system and an extended digital map. The recognition process is divided into four stages and we suggest an extended methodology for each stage. These stages are: ROI generation, detection, classification, and tracking. The 3D positions of traffic lights and slope information obtained from an extended digital map enable ROIs to be generated accurately, even on slanted roads, while independent design and implementation of individual recognition modules for detection and classification allow for selection according to the type of traffic light face. Such a modular architecture gives the system simplicity, flexibility, and maintainable algorithms. In addition, adaptive tracking that exploits the distance to traffic lights allows for seamless state estimation through smooth data association when measurements change from long to short ranges. Evaluation of the proposed system occurred at six test sites and utilized two different types of traffic lights, seven states, sloped roads, and various environmental complexities. The experimental results show that the proposed system can recognize traffic lights with 98.68% precision, 92.73% recall, and 95.52% accuracy in the 10.02–81.21 m range.
29 citations
Additional excerpts
...…Yehu, zguner, Redmill, & Jilin, 2009; Ying, Chen, Gao, & Xiong, 2013; Yu t al., 2010; Zhang, Fu, Yang, & Wang, 2014; Zixing, Yi, & Mingqin, 012; Zong & Chen, 2014 ), but there has been less work on utilizng localization and map information for TLR ( Barnes et al., 2015; airfield & Urmson,…...
[...]
Patent•
11 Nov 2014
TL;DR: In this paper, an autonomous vehicle having a vehicle control system includes an image processing system that receives an image that includes a plurality of image portions and calculates a score for each image portion, indicating a level of confidence that a given image portion represents an illuminated component of a traffic light.
Abstract: The present disclosure is directed to an autonomous vehicle having a vehicle control system. The vehicle control system includes an image processing system. The image processing system receives an image that includes a plurality of image portions. The image processing system also calculates a score for each image portion. The score indicates a level of confidence that a given image portion represents an illuminated component of a traffic light. The image processing system further identifies one or more candidate portions from among the plurality of image portions. Additionally, the image processing system determines that a particular candidate portion represents an illuminated component of a traffic light using a classifier. Further, the image processing system provides instructions to control the autonomous vehicle based on the particular candidate portion representing an illuminated component of a traffic light.
22 citations
01 Oct 2017
TL;DR: This paper proposes a hybrid method, which combines the results of spotlight detection and color-shape model-based method for extracting candidates, and scales each candidate as a 10-by-10 image patch and uses its raw RGB pixel values as the input of a Support Vector Machine (SVM) classifier.
Abstract: In this paper, we consider the problem of recognizing circular traffic lights from an image. The traffic light recognition is divided into two stages: candidate region extraction and traffic light recognition. For extracting candidates, we propose a hybrid method, which combines the results of spotlight detection and color-shape model-based method. Instead of handcrafting a set of features for classification, we scale each candidate as a 10-by-10 image patch and use its raw RGB pixel values as the input of a Support Vector Machine (SVM) classifier. The classifiers are trained only using a Singapore dataset, and are tested on the US LISA dataset. The cross validation justifies the generalizability of our classifiers. The evaluation results show that our hybrid candidate extraction method lowers the chance of miss-detection and the proposed featureless classification approach has a high recognition precision. Our algorithm is robust and efficient, which can run at 30 fps for images with a resolution of 640∗480.
9 citations
Cites methods from "Traffic Light Detection Based on Mu..."
...Empirically determined thresholds are used for traffic light validation [4], [7], [9]–[11], [28]....
[...]
References
More filters
09 May 2011
TL;DR: This work introduces a convenient technique for mapping traffic light locations from recorded video data using tracking, back-projection, and triangulation, and is the first to account for multiple lights per intersection, which yields superior results by probabilistically combining evidence from all available lights.
Abstract: Detection of traffic light state is essential for autonomous driving in cities. Currently, the only reliable systems for determining traffic light state information are non-passive proofs of concept, requiring explicit communication between a traffic signal and vehicle. Here, we present a passive camera-based pipeline for traffic light state detection, using (imperfect) vehicle localization and assuming prior knowledge of traffic light location. First, we introduce a convenient technique for mapping traffic light locations from recorded video data using tracking, back-projection, and triangulation. In order to achieve robust real-time detection results in a variety of lighting conditions, we combine several probabilistic stages that explicitly account for the corresponding sources of sensor and data uncertainty. In addition, our approach is the first to account for multiple lights per intersection, which yields superior results by probabilistically combining evidence from all available lights. To evaluate the performance of our method, we present several results across a variety of lighting conditions in a real-world environment. The techniques described here have for the first time enabled our autonomous research vehicle to successfully navigate through traffic-light-controlled intersections in real traffic.
177 citations
"Traffic Light Detection Based on Mu..." refers background in this paper
...Google’s car just uses the first way [1]....
[...]
11 Sep 2009
TL;DR: A fast method of detecting a traffic light in a scene image by converting the color space from RGB to normalized RGB and a method based on the Hough transform is applied to detect an exact region.
Abstract: The concern of the intelligent transportation system rises and many driver support systems have been developed. In this paper, a fast method of detecting a traffic light in a scene image is proposed. By converting the color space from RGB to normalized RGB, some regions are selected as candidates of a traffic light. Then a method based on the Hough transform is applied to detect an exact region. Experimental results using images including a traffic light verifies the effectiveness of the proposed method.
118 citations
03 Dec 2010
TL;DR: A method for detecting a traffic light in a scene image based on the Hough transform is proposed and experimental results using images including a traffic lights taken by a digital camera through a windshield verify the effectiveness of the proposed method.
Abstract: Driver support systems using images are paid attention and various researches on recognizing and understanding the road environment have been done. If it is possible to detect and recognize traffic lights, it will give useful information to a driver to understand the road environment. In this paper, a method of detecting a traffic light in a scene image is proposed. Considering the structure of a traffic light, we propose a method for detecting a traffic light based on the Hough transform. Experimental results using images including a traffic light taken by a digital camera through a windshield verifies the effectiveness of the proposed method.
69 citations
"Traffic Light Detection Based on Mu..." refers methods in this paper
...proposed a method to detect traffic lights consider the circularity of a light region [2], but when the segmentation is not satisfying enough or it is an arrow light, then the detecting will fail....
[...]
Proceedings Article•
01 Jan 2009
63 citations
"Traffic Light Detection Based on Mu..." refers background in this paper
...In some researches like [4], the shape and structure of traffic lights were not taken into consideration, which will cause many wrong results such as tail lights and some traffic signs, etc....
[...]
03 Dec 2010
TL;DR: The algorithm of traffic lights detection which is applied in a vehicle driver assistance system is designed by using the image processing technology and the results of experiments show that the algorithm is stable and reliable.
Abstract: In order to reduce accident at traffic intersections during day and night, the algorithm of traffic lights detection which is applied in a vehicle driver assistance system is designed by using the image processing technology. The system of traffic light detection includes three parts: a CCD camera, an image acquisition card, and a PC. Based on RGB color space, the algorithm extracts red, green, and yellow objects in the image firstly; For the purpose of eliminating disturbance in the environment, the features of traffic lights are used to verify the object identity, and then the types of traffic signals are judged. The results of experiments show that the algorithm is stable and reliable.
50 citations