Content uploaded by Jaerock Kwon
Author content
All content in this area was uploaded by Jaerock Kwon on May 28, 2018
Content may be subject to copyright.
A preview of the PDF is not available
Lane following and obstacle detection techniques in autonomous driving vehicles
... There are various image pre-processing techniques like region of interest (ROI) selection, colour space conversions, distortion correction, and inverse perspective mapping (IPM) (Oliveira et al., 2015). Next, the lane detection is the most crucial step that involves feature extraction using Canny Edge Detector (Muthalagu et al., 2021) and model fitting using colour space exploration but Hough Transform algorithm, is commonly used for lane detection (Amaradi et al., 2016). There are many works reported in the area of lane detection using both conventional and deep learning techniques. ...
... In the field of computer vision, Convolutional Neural Network (CNN) methods are primarily used for feature extraction (Oliveira et al., 2015;Muthalagu et al., 2021) and semantic segmentation to segment the shape and the location of lanes (Xing et al., 2018;Wang et al., 2018). In (Amaradi et al., 2016), weakly-supervised adversarial consisting of three deep neural networks domain adaptation is implemented to improve the segmentation performance from synthetic data to real scenes and the object detection and prediction of segmentation map uses Detection and segmentation (DS) model. Though some of the previous works can detect the lanes accurately, it has some limitations as this framework needs pre-processing step making the algorithm computationally expensive. ...
Ego lane detection is one of the key techniques in Ego Lane Analysis System (ELAS) implemented in smart autonomous driving cars for lane detection in roads. This technique has been extensively studied in recent years because of its accurate and robust detection of shape and location of lanes. The conventional methods are less robust and computationally expensive since they have several challenges in localization of lanes due to presence of occlusions on roads. So to avoid these issues, this paper uses a novel 2-stage lane detection method using deep convolutional neural network to detect the lanes and its key-points by optimally fit a curve to the lane to compensate on above mentioned issues. The proposed methodology for lane detection predicts the key-points accurately and it robust under various weather conditions and highway driving scenarios. In terms of performance, this technique is fast and robust with low computational cost and has high performance when deployed on autonomous vehicle-based systems.
... Automatic driving has numerous advancements within the domain of lane-keeping, distance maintenance, cruise control, and lane departure, etc. Such improvements were vital in ensuring safety and comfort but unfortunately, several challenges are still associated with autonomous vehicles [1][2][3][4]. The most significant amongst them is decision making. ...
... In addition, the distance between 'L 1 ' and 'L 2 ' is lower enough so that 'C' cannot return to driving lane before overtaking both vehicles. Therefore, under such circumstances, a new location of marker is added 1s ahead of 'L 1 ' which is given as 'M 3 ...
Intelligent and safe overtaking maneuvering is always a challenging task for autonomous vehicles. This paper proposes and experimentally implements a novel approach of overtaking maneuvering using modified form of Rendezvous Guidance (RG) algorithm for trajectory planning and obstacle avoidance, considering driver safety and comfort during autonomous overtaking. The simulations for all possible scenarios are conducted to ensure the effectiveness of proposed modified RG algorithm. These scenarios involved presence and absence of obstacle vehicle in overtaking lane alongside leading vehicle in driving lane. In addition, the enhanced performance of modified RG algorithm is established over conventional RG algorithm by comparative analysis. The results indicate that overtaking maneuvering period could be decreased by 10% using a modified RG algorithm and vehicle will cover less distance to complete overtaking. The efficacy of proposed method is justified by performing experiments using mobile robots. The experimental results and simulation results of modified RG algorithm are compared, and their plots are almost identical.
... The PHD filters are used to reduce the computational difficulty of Bayes filters, resulting in higher cost on time complexity. Amaradi P. et. al. [38] presents the Hough transform technique of lane tracking and obstacle detection using LIDAR sensors to measure the distance drifted from the center of the lane to be able to detect obstacles. Kim ZW [39] presented the RANdom SAmple Consensus (RANSAC) algorithm to find the lane boundary hypothesis in real-time. ...
This paper presents an intelligent autonomous document mobile delivery robot using a deep learning approach. The robot is built as a prototype for document delivery service for use in offices in which it can adaptively move across different surfaces, such as terrazzo, canvas, and wooden. In this work, we introduce a convolutional neural network (CNN) to recognize the traffic lanes and the stop signs with the assumption that all surfaces have identical traffic lanes. We train the model using a custom indoor traffic lane and stop sign dataset with the label of motion directions. CNN extracts a direction-of-motion feature to estimate the robot's direction and to stop the robot based on an input image monocular camera view. These predictions are used to adjust the robot's direction and speed. The experimental results show that this robot can move across different surfaces along with the same structured traffic lanes, achieving the model accuracy of 96.31%. The proposed robot helps to facilitate document delivery for office workers, allowing them to work on other tasks more efficiently.
... The distance between the MAS and the acquired object varies between 5 and 20 meters. This variation in distance respects the range indicated by the manufacturer, which guarantees an operating range between 0 and 80 meters [46]. In the acquisition of cloud points of people and cars, the distance variation was between 5 and 10 meters. ...
Works that use point cloud avoid wasting time and cost of collection, using simulators and datasets available in the literature. In this way, there is access to an unlimited and organized amount of point clouds, an ideal setting for deep learning networks and Vehicular ad hoc networks (VANETs). However, models trained with synthetic data present problems when applied to real-world data.This work proposes the use of deep learning in the recognition of 3D objects captured with a Light Detection and Ranging (LIDAR), including a pre-processing stage. In addition, it is proposed two datasets, a real-world and a syntetic; each dataset includes three classes. A method of pre-processing is proposed to circumvent the distribution discrepancies of the proposed datasets and the existing datasets from literature, such as ModelNet. We use deep learning with the PointNet method, as it supports raw data from point clouds as input to the network. We performed three evaluation approaches: training and testing steps with the proposed datasets using (1) Lidar3DNetV1, which is a proposed network in this paper, (2) PointNet, and (3) classification of ModelNet datasets using Lidar3DNetV1. The proposed network achieved 98.33% of accuracy and a testing time of 88μs in the synthetic dataset, while in the real-world dataset, the network reached 98.48% and 145μs in accuracy and testing time, respectively.
The greatest number of fatal traffic accidents occurs on curved roads at nighttime. In most cases, the late recognition of objects in the traffic zone plays an important role. This highlights to the importance of the role of automobile forward lighting systems. This paper developed a proto-type auto adjustable headlamp and mirror tilt to improve cost and reliability. Also, an adaptive mirror is implemented to remove the blind spots while taking turns. The methodology used here adaptive headlamps and mirrors are developed using Raspberry Pi3 as hardware and Python is used as programming language. Machine learning algorithm “Linear regression” is used for computing the output. Machine Learning Linear regression is considered here as it simple and efficient algorithm in terms of implementation and memory usage. Easily available components like Raspberry Pi3, LDR Sensor, ADXL Gyroscope are used and the design is developed to provide the steering mechanism for the headlamps and mirror which are actuated along with the steering of the front wheels. Around 15% increase in the illuminated area on road and 20% increase in the side mirror view is achieved.
With the development of high-speed computing devices and advanced machine learning theories such as deep learning, end-to-end detection algorithms can be used to solve the problem of lane detection in a more efficient way. However, the key challenge for lane detection systems is to adapt to the demands of high reliability and diverse road conditions. An efficient way to construct a robust and accurate advanced lane detection system is to fuse multimodal sensors and integrate the lane detection system with other object detection systems. In this chapter, we briefly review traditional computer vision solutions and mainly focus on deep learning-based solutions for lane detection. Additionally, we also present a one-lane detection evaluation system, including offline and online systems. Finally, we use one lane detection algorithm and code to show how lane detection works in an autonomous driving system.
The advances in lane detection technologies and computer vision enabled the evolution of lane-keeping systems, driver assistance and lane departure warning in traffic management for road safety. However, it is very challenging to identify and track the lane lines due to improper marking of lane lines and blind turns on the road. The present work proposes effective and efficient vision-based real-time lane markings and tracking lane detection methods for straight and curved lane lines. That can adapt to various environmental conditions. Further, the Hough transform optimization is performed to identify lane lines accurately, and the Kalman filter is employed to track lane lines detected in the ROI by the Sobel operator. The proposed approaches show their significance by achieving real-time response and high accuracy for a vehicle in lane change assistant system on highways. While comparing, the proposed methods show better results in terms of detection rate and processing time for straight lanes and detection accuracy, precision, recall and F1-Score for the curved lanes. The result of processing time and accuracy rate for straight lane detection is 16.7 fps, 96.3%, respectively, and the accuracy, precision, recall and F1-scores for curved lane detection are 97.74%, 98.15%, 97.35% and 97.75% in videos, respectively.
In this work we developed a novel and fast traffic sign recognition system, a
very important part for advanced driver assistance system and for autonomous
driving. Traffic signs play a very vital role in safe driving and avoiding
accident. We have used image processing and topic discovery model pLSA to
tackle this challenging multiclass classification problem. Our algorithm is
consist of two parts, shape classification and sign classification for improved
accuracy. For processing and representation of image we have used bag of
features model with SIFT local descriptor. Where a visual vocabulary of size
300 words are formed using k-means codebook formation algorithm. We exploited
the concept that every image is a collection of visual topics and images having
same topics will belong to same category. Our algorithm is tested on German
traffic sign recognition benchmark (GTSRB) and gives very promising result near
to existing state of the art techniques.
We present a system for detecting and tracking the lanes of high-speed roads, in order to warn the driver of accidental lane departures. The proposed method introduces a novel variant of the classic Hough transform, better equipped to detect and locate linear road markings with a common vanishing point. This is combined with a simple model of the lane and an Extended Kalman Filter to make detection and tracking more robust. This allows detection of lane changes, resistant to visual interference by traffic and irrelevant road markings.
Lane detection algorithms constitute a basis for intelligent vehicle systems such as lane tracking and involuntary lane departure detection. In this paper, we propose a simple and video-based lane detection algorithm that uses a fast vanishing point estimation method. The first step of the algorithm is to extract and validate the line segments from the image with a recently proposed line detection algorithm. In the next step, an angle based elimination of line segments is done according to the perspective characteristics of lane markings. This basic operation removes many line segments that belong to irrelevant details on the scene and greatly reduces the number of features to be processed afterwards. Remaining line segments are extrapolated and superimposed to detect the image location where majority of the linear edge features converge. The location found by this efficient operation is assumed to be the vanishing point. Subsequently, an orientation-based removal is done by eliminating the line segments whose extensions do not intersect the vanishing point. The final step is clustering the remaining line segments such that each cluster represents a lane marking or a boundary of the road (i.e. sidewalks, barriers or shoulders). The properties of the line segments that constitute the clusters are fused to represent each cluster with a single line. The nearest two clusters to the vehicle are chosen as the lines that bound the lane that is being driven on. The proposed algorithm works in an average of 12 milliseconds for each frame with 640×480 resolution on a 2.20 GHz Intel CPU. This performance metric shows that the algorithm can be deployed on minimal hardware and still provide real-time performance.
This paper proposes an improved lane departure algorithm based on region of interest (ROI) segmentation. ROI of a test image is extracted first. Based on statistical parameters a departure warning is generated. Hough transform is used to detect the lines in the segmented ROI. Distance between Hough origin and lane-line midpoint is estimated. Based on the difference between these distances, a lane departure decision is made. A typical road may have multiple lanes which may cause difficulty in generating appropriate departure decision with good accuracy and in short time. Experiments show that the proposed algorithm eliminates this problem. It is simple to implement and robust, and can detect the lane markings accurately and quickly.
In this paper, an integrated driving safety system, which combines map navigation, lane detection, vehicle detection and speed camera warning, is presented. Lane detection and vehicle detection are work together for providing accurate recognition result. The system is porting on PAPAGO P3 driving recorder to verify the recognition rate and performance through field testing. With the knowledge of road model and vehicle appearance, either Lane Departure Warning System and Front Collision Warning System provides advanced assistance to drivers. Experimental results show that the proposed method is robust and the performance is capable for practical applications.
This paper describes a vision-based lane detection system with the optimized Hough Transform circuit. The Hough Transform is a popular method to find the line features in an image. This is very robust to noises and changes in the illumination level, but it requires long computation time and large data storage for calculation. It needs large logic gates for implementation. It is difficult to apply in products that require real-time performance. In this paper, we propose the optimized Hough Transform circuit architecture and a lane departure warning system using vision device. We suggest the Hough Transform architecture to minimize the size of logic and the number of cycle time. Our implemented Hough Transform circuit show the good performance than other circuit architecture. We tested the Hough Transform circuit and lane departure warning system on the Xilinx FPGA board.
The vision based Lane Departure Warning System (LDWS)is a effective way to prevent Single Vehicle Road Departure accident. In practice, a variety of complex noise make it very difficult to detect lanes fast and accurately, so to establish a set of image processing method which can give results fast and accurately in the non-ideal conditions is the primary work. In this paper, a lane detection method for Lane Departure Warning System is proposed. The Canny algorithm is determined as the edge detection method according to the experiment comparison, and Hough transform is selected as the efficient way to detect the beeline. To meet the real-time requirement, the region of interest (ROI) is defined to reduce noise for rising accurately, and to enhance the processing speed. At last, experiment results indicate that this lane detection method can extract lane information from road images acquired efficiently and accurately.