Vol.9/No.2 (2017) INTERNETWORKING INDONESIA JOURNAL 33 ISSN: 1942-9703 / CC BY-NC-ND Abstract— Implementation of vision-based lane detection system in vehicle usually needs high detection rate for reliability. However the implementation can be limited by the calculation power available in the vehicle. This paper reports an improvement of our previous approach while keeping the processing power low. The algorithm proposed in this paper avoid heavy pixel-by-pixel image processing of the captured image, it works mostly on the detected line segments instead of the image. The proposed algorithm is also more global, capable in recognizing multiple lane markers and estimating the position of the vehicle on a multi-lane road. The algorithm is evaluated using miniature road and miniature vehicle for various poses. The recognition rate and the estimation accuracy are shown to be much higher with only 1% failure rate compared to the previous approach in [12] or [13]. Index Terms—Hough transform, inverse perspective mapping, multi-lane recognition, vehicle pose estimation. http://www.ieee.org/organizations/pubs/ani_prod/keywrd98.txt I. INTRODUCTION he future of automobiles is predicted to include a lot of automations [1]. In fact autonomous cars are under researched and tested in many countries by many institutions [2]-[3]. Some of the purposes of having autonomous car are to reduce traffic accidents caused by human error, to increase comfort, and to increase the capacity of the road. One important information for autonomous car is the pose of the Manuscript received October 2, 2015. This work was supported by research grant from the Indonesian Ministry of Technology Research and Higher Education. Sofyan Tan is with the Automotive and Robotics Engineering, BINUS ASO School of Engineering, Bina Nusantara University, Alam Sutera, Serpong, Indonesia (e-mail: sofyan@binus.edu). He received his B.S. degree in Computer Engineering from Bina Nusantara University, Jakarta, Indonesia, in 2002, and his M.Eng. degree in Electronics Engineering from the University of Tokyo, Tokyo, Japan, in 2008. Rudy Susanto is with the Computer Engineering Department, Bina Nusantara University, Jakarta, Indonesia, (phone: +6221-53696930; e-mail lierudys@gmail.com). He received his B.S. degree in Computer Engineering and M.S. degree in Computer Science both from Bina Nusantara University. vehicle relative to the road lane, which include the lateral position of the vehicle and the orientation of the vehicle relative to the lane. Many of the related research road for lane recognition system involve the Hough transform [4]-[8], Gabor Filter [5], [9], geometrical model [5], Fuzzy reasoning [8], or color features [10]. Some also use inverse or warp perspective mapping [6], [9], [11] to remove skewing caused by projection of 3D world to the 2D image in the camera. Most of these approaches require relatively high calculation cost that can limit their implementation in power limited and environment friendly vehicles. Our objective in this research is to develop a low cost algorithm for lane markers recognition and vehicle pose estimation on a multi-lane road. In this research we extend our low cost single-lane recognition algorithm in [12] and [13] to cover multiple lanes and to improve the lane recognition rate and estimation accuracy while maintaining the low cost nature of our algorithm. Instead of working directly on the image to recognize the lane markers, the algorithm uses the probabilistic Hough transform [14], to detect the lines in the source image. From this point forward the entire multi-lane recognition algorithm works only on the line’s coordinates which are much less than the number of pixel in the original image. II. IMAGE PRE-PROCESSING AND LINE DETECTION Assuming that the camera’s optical axis is parallel to the flat road, the vanishing point of the lane will be at the center row of the captured image. Therefore the information above the center row of the image can be neglected, and further image processing only applies to the lower half of the image to improve calculation speed. Canny edge detection is used to obtain the binary edge information from the captured gray image. The binary edge contains less number of non-zero pixels compared to the gray image. The lesser the pixels, the lesser the calculation cost to do Hough transformation of the edge image in order to detect lines in the image. Since the lane is assumed straight at up to an adequate distance, then most part of the lane markings will show up as straight lines. Lane markings at further distance from the camera will be projected to smaller number of pixels in the image. Hence Hough transform of the edge image will mainly detect lines near to the camera. Low Cost Vision-Based Multi-Lane Road Marker Recognition and Vehicle Pose Estimation Sofyan Tan and Rudy Susanto, Members, IEEE T