End-to-End Neuromorphic Lane Detection

There have been many studies that developed approaches for real-time lane and vehicle detection. However, the majority of implementations that have been made with these approaches are focused on cars. On this paper, the existing approach of object detection is implemented for motorcycle detection. Lane detection is also incorporated for the validation of object's position. We employ Haar cascade classifier approach for motorcycle detection and edge detection approach for lane detection. In order to refine the detection result, angle-based elimination is applied to lines in lane detection. For the refinement of motorcycle detection, object candidates are eliminated by object's position on lane, frequency of occurrence, and variation of horizontal edges. Experimental results show that these approaches followed by the refinement method results provide better precision of motorcycle detection. This implementation is also able to operate in real-time.

<div class="section abstract"><div class="htmlview paragraph">Lane detection plays a critical role in autonomous vehicles for safe and reliable navigation. Lane detection is traditionally accomplished using a camera sensor and computer vision processing. The downside of this traditional technique is that it can be computationally intensive when high quality images at a fast frame rate are used and has reliability issues from occlusion such as, glare, shadows, active road construction, and more. This study addresses these issues by exploring alternative methods for lane detection in specific scenarios caused from road construction-induced lane shift and sun glare. Specifically, a U-Net, a convolutional network used for image segmentation, camera-based lane detection method is compared with a radar-based approach using a new type of sensor previously unused in the autonomous vehicle space: radar retro-reflectors. This evaluation is performed using ground truth data, obtained by measuring the lane positions and transforming them into pixel coordinates. The performance of each method is assessed using the statistical <i>R</i><sup>2</sup> score, indicating the correlation between the detected lane lines and the ground truth. The results show that the U-Net camera-based method exhibits limitations in accurately detecting and aligning the lane lines, particularly in challenging scenarios. However, the radar-based lane detection method demonstrates a strong correlation with the ground truth which implies that the use of this sensor may improve current reliability issues from conventional camera lane detection approach. Furthermore, the study highlights the limitations of the U-Net model for camera lane detection, especially in scenarios with sun glare. This study shows that infrastructure-based radar retro-reflectors can improve autonomous vehicle lane detection reliability. The integration of different sensor modalities and the development of advanced computer vision algorithms are crucial for improving the accuracy, reliability, and energy efficiency of lane detection systems. Addressing these challenges contributes to the advancement of autonomous vehicles and the realization of safer and more efficient transportation systems.</div></div>

Lane detection is a crucial module for today’s autonomous driving cars. Detecting road lanes is a challenging task as it varies in color, texture, boundaries and markings. Traditional lane detection techniques detect the lane by applying a model trained with centralized data. As roads vary in urban and rural areas, a more localized and decentralized training technique is desired for accurate and personalized lane detection. Federated learning has recently proved to be a promising technology that trains and prunes the model using local data. Applying federated learning-based lane detection improves the accuracy of detection and also ensures the security and privacy of autonomous cars. This paper proposes FedLANE, a federated learning-based lane detection technique. U-Net, U-Net long short-term memory (LSTM) and <br /> AU-Net architectures were explored using a federated learning approach. Experimental analysis using TuSimple and CuLane dataset shows that the FedLANE based lane detection performs similar to that of the traditional deep learning lane detection models.

Lane detection is a crucial task in the field of autonomous driving, as it enables vehicles to safely navigate on the road by interpreting the high-level semantics of traffic signs. Unfortunately, lane detection is a challenging problem due to factors such as low-light conditions, occlusions, and lane line blurring. These factors increase the perplexity and indeterminacy of the lane features, making them hard to distinguish and segment. To tackle these challenges, we propose a method called low-light enhancement fast lane detection (LLFLD) that integrates the automatic low-light scene enhancement network (ALLE) with the lane detection network to improve lane detection performance under low-light conditions. Specifically, we first utilize the ALLE network to enhance the input image’s brightness and contrast while reducing excessive noise and color distortion. Then, we introduce symmetric feature flipping module (SFFM) and channel fusion self-attention mechanism (CFSAT) to the model, which refine the low-level features and utilize more abundant global contextual information, respectively. Moreover, we devise a novel structural loss function that leverages the inherent prior geometric constraints of lanes to optimize the detection results. We evaluate our method on the CULane dataset, a public benchmark for lane detection in various lighting conditions. Our experiments show that our approach surpasses other state of the arts in both daytime and nighttime settings, especially in low-light scenarios.

Lane detection is crucial for driver assistance systems. However, road scenes are severely degraded in dense fog, which leads to the loss of robustness of many lane detection methods. For this problem, an end-to-end method combining polarimetric dehazing and lane detection is proposed in this paper. From images with dense fog captured by a vehicle-mounted monochrome polarization camera, the darkest and brightest images are synthesized. Then, the airlight degree of polarization is estimated from angle of polarization, and the airlight is optimized by guided filtering to facilitate lane detection. After dehazing, the lane detection is carried out by a Canny operator and Hough transform. Having helped achieve good lane detection results in dense fog, the proposed dehazing method is also adaptive and computationally efficient. In general, this paper provides a valuable reference for driving safety in dense fog.

Lane detection plays a crucial role in autonomous driving systems, enabling vehicles to navigate safely and efficiently in complex environment. Despite significant advancements in recent years, accurate lane detection remains a challenging task, particularly in scenarios with occlusions, ambiguous lane markings, and diverse lighting conditions. In this paper, we propose the Global Enhancement and Optimization Network (GEONet) for lane detection, which is designed to refine both feature extraction and global feature transmission. Traditional approaches typically depend on deep convolutional layer stacks for global feature extraction, a process that often compromises inference speed and the precision of global feature representation. In contrast, GEONet introduces a novel and more effective methodology. We present the Global Feature Extraction Module (GFEM), which is specifically engineered to capture comprehensive global features with higher accuracy. Additionally, we introduce the Top-Tier Supplementary Module (TTSM), which enhances these features through a bottom-up approach, improving overall lane detection accuracy. To further bolster our framework, we incorporate Whitening Batch Normalization (WBN) and Whitening Contrastive Learning (WCL), which enhance feature robustness and ensure better generalization. In addition to our novel network design, we propose two new loss functions to enhance lane detection accuracy. The Generalized Rectangular Intersection over Union (GRIoU) Loss extends the predicted points into rectangles, optimizing overlap and smoothness of lane predictions.The Angle Loss accounts for angular differences between predicted and ground truth lanes, improving alignment and continuity. Experimental results demonstrate that our proposed method significantly outperforms current state-of-the-art lane detection techniques.Editorial NotesThis article, which was published in Proceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI 2025), has been retracted by agreement between the authors and the journal.

Lane detection based on the visual sensor is of great significance for the environmental perception of the intelligent vehicle. Current mature lane detection algorithms are trained and implemented in good visual conditions. However, the low-light environment such as in the night is much more complex, easily causing misdetections and even perception failures, which are harmful to the downstream tasks such as behavior decision and control of ego-vehicle. To tackle this problem, we propose a new lane detection algorithm that introduces the multi-light information into lane detection task. The proposed algorithm adopts a multi-exposure image processing module, which generates and fuses multi-exposure information from the source image data. By integrating this module, mainstream lane detection models can jointly learn the extraction of lane features as well as the enhancement of low-exposed image, thus improving both the performance and robustness of lane detection in the night.

In recent years, many companies are working on the development of autonomous cars. Every citizen on the planet is concerned about their safety. In the framework of Advanced Driver Assistance Systems, one of the main goals is to improve safety and reduce road accidents, ultimately saving lives. One of the most difficult jobs in an autonomous driving system is detecting road lanes or road boundaries. Lane and object detection may be a crucial component of collision prevention in driving assistance systems. With the growth in traffic, there is a demand for more security and comfort when driving, it impose the development of new technology. Computer vision is one of the ways that may be utilized to assist the driver in difficult scenarios in order to improve his safety and comfort. The initial layer of autonomous vehicles' capabilities is lane tracking. Many sensors, including as lasers, radar, and vision sensors, are commonly employed for obstacle and lane detection. Computer vision is one of the primary method for detecting road limits and lanes using a vehicle's vision system. The technology uses a camera installed on the vehicle to capture the front view, then uses a few algorithms to detect lanes and objects. The lanes and objects are detected using a flexible algorithm. This paper focuses on a computer vision-based object detection technique by using CNN algorithm.

An advanced driver-assistance system (ADAS), based on lane detection technology, detects dangerous situations through various sensors and either warns the driver or takes over direct control of the vehicle. At present, cameras are commonly used for lane detection; however, their performance varies widely depending on the lighting conditions. Consequently, many studies have focused on using radar for lane detection. However, when using radar, it is difficult to distinguish between the plain road surface and painted lane markers, necessitating the use of radar reflectors for guidance. Previous studies have used long-range radars which may receive interference signals from various objects, including other vehicles, pedestrians, and buildings, thereby hampering lane detection. Therefore, we propose a lane detection method that uses an impulse radio ultra-wideband radar with high-range resolution and metal lane markers installed at regular intervals on the road. Lane detection and departure is realized upon using the periodically reflected signals as well as vehicle speed data as inputs. For verification, a field test was conducted by attaching radar to a vehicle and installing metal lane markers on the road. Experimental scenarios were established by varying the position and movement of the vehicle, and it was demonstrated that the proposed method enables lane detection based on the data measured.

Despite the recent advances in lane detection models, assuring the correct lane segmentation as output is nearly impossible because a deep learning network is a sort of black-box model. According to National Highway Traffic Safety Administration (NHTSA), traffic accidents caused 36,120 deaths in the U.S in 2019. Given the importance of the lane detection task on the autonomous driving system, the result of lane segmentation must be ensured. To provide the lane detection of high quality, we propose a novel lane segmentation framework by cascading the supervised version of the convolutional recurrent reconstructive network (CRRN) [1] to the existing lane detection network, ENet-SAD [2]. The supervised CRRN utilizes spatio-temporal information to reconstruct the lane segment that is detected by ENet-SAD. Experimental results for the CULane dataset confirm that our proposed framework improves the performance of the conventional lane detection network. Our framework achieves 0.435 of pixel-wise F1 measure, while vanilla ENet-SAD achieves 0.429.KeywordsLane detectionCRRNSpatio-temporal dataENet-SAD

Lane detection methods based on convolutional neural network have achieved excellent performance in recent years. Most of them treat lane detection as a semantic segmentation task which judges whether each pixel belongs to a lane. To make full use of the characteristics of lane shape, some researchers proposed to predict the whole lane. In this paper, we propose Lane Detection Combining Details and Integrity (LDCDI) which can explicitly leverage the advantages of both the segmentation-based methods and the regression-based methods. Specifically, we exploit an extra branch with regression-based methods as the auxiliary module after the main module. It not only maintains the advantages of the segmentation-based methods in lane detail segmentation, but also enables the model to have a sufficient understanding of the lane shape. Besides, the auxiliary module only takes part in the training, and there is no extra cost in the prediction. To further improve the quality of lane detection, we introduce a novel direction-sensitive block (DSB) based on ERFNet as the main module, which is more sensitive to the direction information of the image, so as to obtain better performance. Extensive experiments on the CULane dataset can demonstrate that our method outperforms other methods and achieves the state-of-the-art.

Nowadays, deep learning techniques are widely used for lane detection, but application in low-light conditions remains a challenge until this day. Although multi-task learning and contextual-information-based methods have been proposed to solve the problem, they either require additional manual annotations or introduce extra inference overhead respectively. In this paper, we propose a style-transfer-based data enhancement method, which uses Generative Adversarial Networks (GANs) to generate images in low-light conditions, that increases the environmental adaptability of the lane detector. Our solution consists of three parts: the proposed SIM-CycleGAN, light conditions style transfer and lane detection network. It does not require additional manual annotations nor extra inference overhead. We validated our methods on the lane detection benchmark CULane using ERFNet. Empirically, lane detection model trained using our method demonstrated adaptability in low-light conditions and robustness in complex scenarios. Our code for this paper will be publicly available.

CFI-Former: Efficient lane detection by multi-granularity perceptual query attention transformer.

Lane detection problem has been considered as an important computer vision task in autonomous driving. While it has received massive research attention in the literature, the problem is not yet fully solved. In this paper, we present a comprehensive literature review for lane detection, especially those with deep learning models. The latest collection of lane detection datasets is presented. We further fill the research gap by proposing a novel lane detection dataset named MudLane, which focuses on the lane detection task on suburban roads.

Lane departure warning system plays an important role in driver assistance systems. The proposed algorithm assumes that lanes are always the straight lines and whole algorithm is based on Hough transform. Due to the complexity of urban traffic scenes, false lane detections are highly caused by warning lines and signs whose shapes and colors are similar to the lane boundary. In this study, we improve the accuracy of the lane detection base on Hough, a score function based on the width between left and right lanes is proposed to obtain reliable lane detect results on urban traffic scene. Meanwhile, a list of candidate lanes is constructed at the least of execution time. Experiments under various scenes showed that the proposed lane detection method can work robustly in the real-time.KeywordsLane detectionIntelligent vehicleMachine visiondriving assistance