Semantic segmentation is a critical task in computer vision that aims to assign each pixel in an image a corresponding label on the basis of its semantic content. This task is commonly referred to as dense labeling because it requires pixel-level classification of the image. The research area of semantic segmentation is vast and has achieved critical advances in recent years. Deep learning architectures in particular have shown remarkable performance in generating high-level, hierarchical, and semantic features from images. Among these architectures, convolutional neural networks have been widely used to address semantic segmentation problems. This work aims to review and analyze recent technological developments in image semantic segmentation. It provides an overview of traditional and deep-learning-based approaches and analyzes their structural characteristics, strengths, and limitations. Specifically, it focuses on technical developments in deep-learning-based 2D semantic segmentation methods proposed over the past decade and discusses current challenges in semantic segmentation. The future development direction of semantic segmentation and the potential research areas that need further exploration are also examined.

To deal with real-world complexities, such as varying road conditions, noise interference, and lane line occlusions, we present a highly efficient and robust algorithm for detecting and tracking lane lines using camera sensors. By leveraging the accessibility and affordability of camera sensors over other sensing technologies like Light Detection and Ranging (LiDAR), our approach provides a cost-effective solution for applications. We refine the density-based spatial clustering of applications with noise through a fusion strategy for clustering to address the crucial challenges of noise interference and detection accuracy in this domain. This strategy, which relies on angle and centroid calculations, significantly mitigates noise interference and bolsters the detection accuracy of solid and dashed lane lines. Such enhancement promotes comprehensive lane line detection and reinforces its stability under diverse road conditions. We also restructure the inlier selection logic of random sample consensus to enrich the model selection mechanism and strengthen the anti-noise mechanism. Such improvements increase the algorithm’s robustness and adaptability. To augment the tracking capability of our algorithm, we integrated the Kalman filter and optical flow computation. The incorporation of tracking logic rooted in lane line features and movement direction during optical flow computation efficiently resolves detection issues during lane line occlusion. Our method processes a single frame image on the datasets within only 0.1045 s, reducing overall runtime by roughly 82.4% compared with conventional algorithms.