AI Based Real-Time Detection of Damaged Road and Lane Detection for Autonomous Vehicle

Abstract

AI Based Real-Time Detection of Damaged Road and Lane Detection for Autonomous Vehicle ABSTRACT: A lot of research is being done in the field of autonomous vehicles. Road lane detection is one of the primary worries that many engineers working on self-driving cars have when it comes to the deployment of traditional computing systems. Damage detection on roads is extremely useful in a wide variety of contexts, particularly those that are concerned with identifying deviations from the typical patterns of roads and surfaces. This allows for timely and appropriate actions to be taken against the damage. In the framework of advanced driver assistance systems, one of the most important goals is to improve road safety and reduce the number of incidents that occur on the roads, which will ultimately save lives. Road lane detection or road boundary detection appears to be one of the jobs that future road vehicles will be expected to perform, which is a difficult and challenging process. The process is based on lane detection, which involves the localisation of the road, the determination of the relative location between the vehicle and the road, and the analysis of the heading direction of the vehicle. The utilisation of a visual system installed on the vehicle is one of the primary methods for identifying the limits of the road and the lanes. Lane detection, on the other hand, is a challenging problem to solve because of the wide range of road conditions that one could experience when driving.

Specification

Description:DESCRIPTIONS:
Under most driving conditions, humans can effectively detect lane markings when fully attentive to the task. Computers are not inherently adept at performing identical tasks. The research aims to train a model to attain human-level proficiency in lane line identification. This will enable the vehicle to assume this responsibility from the human operator. The project's issue lies in identifying a line appropriate for both the left and right lane markings. Road transport is among the most prevalent and economical methods for linking a source point to its destination. The degradation of road surfaces, leading to cracks, potholes, or irregularities, is attributed to the breakdown of the road or surface, primarily due to traffic load and insufficient maintenance. The study aims to assess the condition of the road using eye inspection and surveying devices, along with surface condition analysis. The objective of lane detection is to ascertain the lanes on the roadway and provide the exact location and geometry of each lane. To facilitate contemporary assisted and automated driving systems, this technology is among the most crucial currently obtainable. Conversely, the detection systems face challenges due to various unique properties of lanes. Lane detection algorithms often become perplexed by nearby objects that possess identical visual attributes due to the lack of distinguishing features. The performance is additionally impeded by the diversity of lane line patterns, including solid, broken, single, double, merging, and dividing lines, along with the variable number of lanes on a road. This paper aims to introduce a method utilising deep neural networks, referred to as LaneNet. The methodology is structured to partition the lane identification procedure into two phases: lane edge suggestion and lane line localisation. A lane edge proposal network is employed in the initial stage for pixel-wise lane edge categorisation. The lane line localisation network employed in the second stage recognises lane lines based on the lane edge proposals. Our LaneNet, developed exclusively for lane line identification, encounters further difficulties in mitigating erroneous detections of other road markings, such as arrows and characters. Please be cognisant of this fact. Our lane recognition method is proven to be resilient in both highway and urban environments, eliminating the necessity to depend on assumptions about lane numbers or lane line patterns. This occurs notwithstanding the numerous hurdles involved. Our LaneNet may be deployed on vehicle-based systems because of its high operational speed and cheap computing cost. The outcomes of our trials indicate that LaneNet consistently delivers exceptional performance when utilised in real-world traffic scenarios. The World Health Organisation (WHO) reports that over 1.19 million individuals have perished, with the primary cause of death from injuries sustained in automobile accidents being that more than half of a million people aged 5 to 29 are involved in these incidents. Interest in the development of fully automated vehicles and, ultimately, autonomous driving skills has markedly risen in recent years. This is undertaken to diminish traffic volume and the incidence of accidents. The safety of these automated actions is a critical factor that must be taken into account. Identifying traffic patterns around vehicles is a crucial phase in the autonomous driving process. Road markings, especially lane boundaries, are crucial for autonomous driving systems to identify crowded traffic circumstances. The modules involved in decision-making and path planning necessitate this knowledge. Human drivers predominantly depend on their vision for operating vehicles, and the market pricing for vision sensors are quite modest. Vision-based techniques have been enhanced due to recent advancements in computer vision and machine learning. Consequently, the vision-based lane detection algorithm has emerged as the predominant method for lane recognition. Camera-based lane identification is a crucial element of environmental awareness as it allows vehicles to operate inside the boundaries of the lane. This concept serves as the basis for most lane departure warning and lane-keeping systems. This article advocated the development of a real-time lane recognition system with video sequences collected from a vehicle moving on a highway. The method employs various images. A selection of the diverse frames employed in the lane detecting approach, referred to as the F.H.D. algorithm, is presented from this series. This technique performs image segmentation and removes road shadows. We regard the lanes as straight lines within an acceptable range to ensure vehicle safety, as they are typically lengthy and curve smoothly.
, Claims:CLAIMS:
1. The implementation of damage detection and lane detection systems that are powered by artificial intelligence is the goal of this project, which aims to develop roadway management solutions for autonomous vehicles.
2. The development of algorithms that are capable of reliably identifying road damage and detecting lane markings in real time is the key focusses of this project.
3. The purpose of this system is to improve both the safety and the efficiency of roadways by utilising technology such as machine learning and computer vision software.
4. Autonomous vehicles will be able to more successfully manoeuvre around barriers and dangers with the assistance of the damage detection component, while the lane recognition element will guarantee that vehicles are positioned precisely within the lanes that have been allotted for them.
5. In general, the purpose of this integrated strategy is to enhance the capabilities of autonomous cars, which will ultimately lead to transportation networks that are safer and more reliable.

Documents