경량화된 임베디드 시스템에서 역 원근 변환 및 머신 러닝 기반 차선 검출

홍성훈,박대진 대한임베디드공학회 2022 대한임베디드공학회논문지 Vol.17 No.1

This paper proposes a novel lane detection algorithm based on inverse perspective transformation and machine learning in lightweight embedded system. The inverse perspective transformation method is presented for obtaining a bird’s-eye view of the scene from a perspective image to remove perspective effects. This method requires only the internal and external parameters of the camera without a homography matrix with 8 degrees of freedom (DoF) that maps the points in one image to the corresponding points in the other image. To improve the accuracy and speed of lane detection in complex road environments, machine learning algorithm that has passed the first classifier is used. Before using machine learning, we apply a meaningful first classifier to the lane detection to improve the detection speed. The first classifier is applied in the bird’s-eye view image to determine lane regions. A lane region passed the first classifier is detected more accurately through machine learning. The system has been tested through the driving video of the vehicle in embedded system. The experimental results show that the proposed method works well in various road environments and meet the real-time requirements. As a result, its lane detection speed is about 3.85 times faster than edge-based lane detection, and its detection accuracy is better than edge-based lane detection. This paper proposes a novel lane detection algorithm based on inverse perspective transformation and machine learning in lightweight embedded system. The inverse perspective transformation method is presented for obtaining a bird’s-eye view of the scene from a perspective image to remove perspective effects. This method requires only the internal and external parameters of the camera without a homography matrix with 8 degrees of freedom (DoF) that maps the points in one image to the corresponding points in the other image. To improve the accuracy and speed of lane detection in complex road environments, machine learning algorithm that has passed the first classifier is used. Before using machine learning, we apply a meaningful first classifier to the lane detection to improve the detection speed. The first classifier is applied in the bird’s-eye view image to determine lane regions. A lane region passed the first classifier is detected more accurately through machine learning. The system has been tested through the driving video of the vehicle in embedded system. The experimental results show that the proposed method works well in various road environments and meet the real-time requirements. As a result, its lane detection speed is about 3.85 times faster than edge-based lane detection, and its detection accuracy is better than edge-based lane detection.

형태학과 색상 정보를 이용한 차선 인식 알고리즘

배찬수(Bae Chansu),이종화(Jong-Hwa Lee),조상복(Sang-Bock Cho) 大韓電子工學會 2011 電子工學會論文誌-SD (Semiconductor and devices) Vol.48 No.6

지능형 자동차 시스템에 대한 인식이 높아지면서 차선 획득 알고리즘에 대해 많이 연구되고 있다. 일반적인 차선 인식에서 사용하는 경계선 추출을 사용하는 방법은 도로에서의 차선 검출에 좋은 결과를 가져 올 수 있다. 하지만 도로에 그림자, 혹은 가로 선 같은 다른 경계선이 검출 될 수 있다. 본 논문에서는 이와 같은 문제를 해결하기 위해 형태학적 연산을 적용하여 차선에 대한 정보를 추출하였다. 또한 HSV(Hue, Saturation, Value) 칼라 모델을 적용하여 색상에 대한 정보를 이용함으로써 한번 더 차선의 정보를 추출하였다. 추출된 차선의 후보들을 이용하여 Hough 변환을 통해 차선이 존재할 가능성이 높은 차선 검출 영역을 설정하고, 이러한 차선 검출 영역 내에서 차선을 추출하는 방식을 사용함으로써 효과적으로 차선을 검출할 수 있었다. As increase awareness of intelligent vehicle systems, many kinds of lane detection algorithm have been proposed. General boundary extraction method can bring good result in detection of lane on the road. But a shadow on the road, or other boundaries, such as horizontal lines can be detected. The method using morphological operations was used to extract information about Lane. By applying HSV color model for color information of lane, the candidate of the lane can be extracted. In this paper, the lane detection region was set by Hough transformation using the candidate of the lane. By extracting lane markings on the lane detection region, lane detection method can bring good result.

SENSOR FUSION-BASED LANE DETECTION FOR LKS+ACC SYSTEM

정호기,이윤희,강형진,김재희 한국자동차공학회 2009 International journal of automotive technology Vol.10 No.2

This paper discusses the market trends and advantages of a safety system integrating LKS (Lane Keeping System) and ACC (Adaptive Cruise Control), referred to as the LKS+ACC system, and proposes a method utilizing the range data from ACC for the sake of lane detection. The overall structure of lane detection is the same as the conventional method using monocular vision: EDF (Edge Distribution Function)-based initialization, sub-ROI (Region Of Interest) for left/right and distance-based layers, steerable filter-based feature extraction, and model fitting in each sub-ROI. The proposed method adds only the system for confining lane detection ROI to free space that is established by range data. Experimental results indicate that such a simple adaptive ROI can overcome occlusion of lane markings and disturbance of neighboring vehicles. This paper discusses the market trends and advantages of a safety system integrating LKS (Lane Keeping System) and ACC (Adaptive Cruise Control), referred to as the LKS+ACC system, and proposes a method utilizing the range data from ACC for the sake of lane detection. The overall structure of lane detection is the same as the conventional method using monocular vision: EDF (Edge Distribution Function)-based initialization, sub-ROI (Region Of Interest) for left/right and distance-based layers, steerable filter-based feature extraction, and model fitting in each sub-ROI. The proposed method adds only the system for confining lane detection ROI to free space that is established by range data. Experimental results indicate that such a simple adaptive ROI can overcome occlusion of lane markings and disturbance of neighboring vehicles.

도로 환경 변화에 강인한 차선 검출 방법

김병수(Byeoung-su Kim),김회율(Whoi-Yul Kim) 대한전자공학회 2012 電子工學會論文誌-SC (System and control) Vol.49 No.1

자동차 기술의 발전으로 카메라를 이용하여 차선을 검출하는 운전자 보조 시스템에 대한 연구가 활발히 진행되고 있다. 하지만 비가 오거나 차선이 노후화된 경우 차선 검출이 어려운 문제가 있다. 본 논문에서는 도로 환경 변화에 강인한 차선 검출 방법을 제안한다. 제안하는 방법은 밝기 값과 차선의 평균적인 폭 정보를 이용하여 후보 영역을 추출한다. 추출된 후보 영역을 기준으로 허프 변환을 이용하여 구간별 직선을 추출하고, B-Snake 방법을 사용하여 자연스러운 차선을 검출하게 된다. 노후화 되거나 손실된 차선을 검출하기 위하여, 기존에 검출된 차선 정보를 이용하여 다음 프레임에서 차선이 위치할 경로를 계산하고, 계산된 경로를 기준으로 차선 영역에서 검출되는 후보 영역에 대한 가중치를 부여한다. 실험 결과 제안하는 방법은 노후화되거나 비가 내려 차선의 밝기가 낮은 경우에도 효과적으로 차선을 검출하였다. Lane detection methods using camera, which are part of the driver assistance system, have been developed due to the growth of the vehicle technologies. However, lane detection methods are often failed by varying road conditions such as rainy weather and degraded lanes. This paper proposes a method for lane detection which is robust in varying road condition. Lane candidates are extracted by intensity comparison and lane detection filter. Hough transform is applied to compute the lane pair using lane candidates which is straight line in image. Then, a curved lane is calculated by using B-Snake algorithm. Also, weighting value is computed using previous lane detection result to detect the lanes even in varying road conditions such as degraded/missed lanes. Experimental results proved that the proposed method can detect the lane even in challenging road conditions because of weighting process.

카메라와 도로평면의 기하관계를 이용한 모델 기반 곡선 차선 검출

장호진(Ho-Jin Jang),백승해(Seung-Hae Baek),박순용(Soon-Yong Park) 제어로봇시스템학회 2015 제어·로봇·시스템학회 논문지 Vol.21 No.2

In this paper, we propose a robust curved lane marking detection method. Several lane detection methods have been proposed, however most of them have considered only straight lanes. Compared to the number of straight lane detection researches, less number of curved-lane detection researches has been investigated. This paper proposes a new curved lane detection and tracking method which is robust to various illumination conditions. First, the proposed methods detect straight lanes using a robust road feature image. Using the geometric relation between a vehicle camera and the road plane, several circle models are generated, which are later projected as curved lane models on the camera images. On the top of the detected straight lanes, the curved lane models are superimposed to match with the road feature image. Then, each curve model is voted based on the distribution of road features. Finally, the curve model with highest votes is selected as the true curve model. The performance and efficiency of the proposed algorithm are shown in experimental results.

Vision-based Fusion of Robust Lane Tracking and Forward Vehicle Detection in a Real Driving Environment

최현철,J.-M. PARK,W.-S. CHOI,S.-Y. OH 한국자동차공학회 2012 International journal of automotive technology Vol.13 No.4

With the goal of developing an accurate and fast lane tracking system for the purpose of driver assistance, this paper proposes a vision-based fusion technique for lane tracking and forward vehicle detection to handle challenging conditions, i.e., lane occlusion by a forward vehicle, lane change, varying illumination, road traffic signs, and pitch motion,all of which often occur in real driving environments. First, our algorithm uses random sample consensus (RANSAC) and Kalman filtering to calculate the lane equation from the lane candidates found by template matching. Simple template matching and a combination of RANSAC and Kalman filtering makes calculating the lane equation as a hyperbola pair very quick and robust against varying illumination and discontinuities in the lane. Second, our algorithm uses a state transfer technique to maintain lane tracking continuously in spite of the lane changing situation. This reduces the computational time when dealing with the lane change because lane detection, which takes much more time than lane tracking, is not necessary with this algorithm. Third, false lane candidates from occlusions by frontal vehicles are eliminated using accurate regions of the forward vehicles from our improved forward vehicle detector. Fourth, our proposed method achieved robustness against road traffic signs and pitch motion using the adaptive region of interest and a constraint on the position of the vanishing point. Our algorithm was tested with image sequences from a real driving situation and demonstrated its robustness.

A New Adaptive Region of Interest Extraction Method for Two-Lane Detection

Chen Yingfo,Wong Pak Kin,Yang Zhi-Xin 한국자동차공학회 2021 International journal of automotive technology Vol.22 No.6

As a key environment perception technology of autonomous driving or driver assistance systems, lane detection is to ensure vehicles to drive safely in corresponding lane. However, existing lane detection algorithms for two-lane detection focus on using various filtering methods to reduce the impact of useless information, resulting in low accuracy and low efficiency. In this paper, a novel Adaptive Region of Interest (A-ROI) extraction method is proposed to improve the accuracy and real-time performance of the two-lane detection algorithm. Three key technologies are introduced to solve the problems. First, A-ROI, which only focuses on the lane where the vehicle is located, is applied to the Bird’s-Eye-View image obtained by using Inverse Perspective Mapping (IPM). Next, based on Bayesian framework and Likelihood models, a lane feature extraction method with a lane-like feature filter is used for edge detection. Finally, an improved Random Sample Consensus (RANSAC) algorithm is introduced by using a filter that can remove noisy lane data. The performance of the proposed A-ROI method together with the improved lane detection method is evaluated via simulation of various scenarios. Experimental results show the proposed method has better accuracy and real-time performance than the traditional lane detection methods.

정밀 지도 제작을 위한 LAS 데이터로부터 차선 끝점 검출 방법

장은석(Eun Seok Jang),서재규(Jae Kyu Suhr),정호기(Ho Gi Jung) 한국자동차공학회 2018 한국자동차공학회 학술대회 및 전시회 Vol.2018 No.11

Recently, precise map including various landmark information has been used in many studies of vehicle localization. Generally, precise maps are built on LAS data acquired from MMS(Mobile Mapping System) using LiDAR(Light Detection and Ranging). As we are developing lane-end-point-based vehicle localization, precise map of lane-end-point is needed for both learning and test. However, as the existing maps only provide lane-level maps, we have to manually input the position of lane-end-point referring to LAS data. This process has the disadvantages of requiring a lot of labor and time. To address the problem, this paper proposes a lane-end-point detection from LAS data for building precise map more efficiently. The proposed method consists of three steps. First, a lane-level map is used to detect the area of dashed-lane from LAS data. Second, the profile of dashed-lane is generated based on intensity values of LAS data. Third, lane-end-points are detected by applying clipping, median filter and local peak detection. Experimental results show that the proposed method automatically detects 5,672 lane-end-points in the 40km length highway. As the total number of lane-end-points is 10,042, it can be recognized that the proposed method can detect 56. 5% of lane-end-points automatically.

단안 카메라를 이용한 차선 인식 방법

이윤희(YunHee Lee) 한국자동차공학회 2010 한국자동차공학회 학술대회 및 전시회 Vol.2010 No.11

This paper proposes a method of mono camera based lane detection. The proposed method extracts lane marking features using steerable filter. The proposed method uses EDF(Edge Distribution Function) method to detect the input angle of steerable filter in first frame. The input angle of steerable filter in first frame is maximum value of histogram accumulated by each pixels orientation. Since second frame, the input angle of steerable filter is detected from the slope of lane markings in previous frame. The features from steerable filer are candidates of lane markings. The lane markings are selected from candidates using the conditions of lane marking width and lane width. Also, the proposed method uses Kalman filter tracking to reduce false detection on account of road markings, neighbor vehicle and so on.

Morphology와 다중 ROI를 이용한 차선 인식 및 추적

오동언(Dong-eon Oh),이민채(Minchae Lee),선우명호(Myoungho Sunwoo) 한국자동차공학회 2010 한국자동차공학회 부문종합 학술대회 Vol.2010 No.5

Lane detection algorithm have been used for passenger safety systems of luxury vehicle such as the lane keeping assist system (LKAS) and the lane departure warning system (LDWS). In order to enhance the performance of passenger safety systems, robustness and low computation load is required for an effective lane detection algorithm. In previous studies, edge detection, pattern recognition, and probabilistic method have been applied for lane detection. However these approaches have some limitations such as high sensitivity to noise, non-uniform illumination, and high computation load. In this study, we proposed a robust lane detection algorithm using morphology and tracking methods. Morphology is used as a preprocessor for lane detection to reduce the influence of cracked road surfaces and shadows. Moreover, we applied tracking method using multiple regions of interest (ROI) windows to reduce the computation time. In particular, the sizes of multiple ROI were determined by considering the geometric scale effect of the lane mark width. The proposed lane detection and tracking algorithm was evaluated on various road conditions and large scale data. As a result, the proposed algorithm proved to be robust and fast enough to apply to real-time safety-critical systems.