3D lidar point-cloud projection operator and transfer machine learning for effective road surface features detection and segmentation

Li, Heyang; Todd, Zachary; Bielski, Nikolas; Carroll, Felix

doi:10.1007/s00371-021-02103-8

Cited by 24 publications

(8 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The introduced system performance was compared with the top-performing algorithms on the KITTI road benchmark. Li et al [25] applied the Mask regionbased convolutional neural network R-CNN algorithm for road surface object segmentation. The model inputs were a 2D image generated by projecting the point cloud encoding the intensity and height information.…”

Section: Related Workmentioning

confidence: 99%

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

Dowajy,

Somogyi,

Barsi

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Automatic road segmentation from three-dimensional point cloud data has gained increasing interest recently. However, it is still challenging to do this task automatically due to the wide variations of roads and complex environments, especially in non-urban areas. This research proposed a comprehensive approach for using shallow neural networks to segment non-urban road point clouds to support autonomous driving applications. The proposed approach involves converting raw point cloud data into a regular grid of cells or partial clouds. Initially, a shallow neural network based on cells' properties (cell plane fitting error, cell average intensity, cell elevation range, and cell weighted density) was employed to extract road cells from raw point cloud data. The road cells were refined and segmented into inside-road and road border point clouds based on morphologic operations. A point-wise shallow neural network was used to extract road points from the border point clouds based on intensity and geometric features (roughness, curvature, and change rate of the normal). A precise road surface point cloud is obtained by merging the inside-road and filtered border point clouds. The method performance was evaluated for two datasets captured using a mobile laser scanner (MLS). In the first dataset, the road points were extracted at average completeness, correctness, quality, and overall accuracy of 98.40%, 99.13%, 97.56%, and 98.47%, respectively. Similarly, the method achieved high scores for the second dataset with 97.22% completeness, 99.02% correctness, 96.29% quality, and 98.71% overall accuracy. The method performance demonstrates an advancement when compared to various state-of-the-art methods and also confirms its adaptability to different road environments.INDEX TERMS road segmentation, 3D point cloud, shallow neural network, cell point cloud, plan fitting, weighted density, geometric features.

show abstract

Section: Related Workmentioning

confidence: 99%

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

Dowajy,

Somogyi,

Barsi

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For reducing the computational power the authors aggregate the point cloud data into a 2D-image space before using the affine transformation. Experimental results displayed their model’s efficacy during detecting lanes [ 43 ]. Haris et al proposed an asymmetric kernel convolution (AK-CNN) for detecting lanes under complex traffic conditions.…”

Section: Literature Reviewmentioning

confidence: 99%

LLDNet: A Lightweight Lane Detection Approach for Autonomous Cars Using Deep Learning

Khan

Haque

Hasan

et al. 2022

Sensors

View full text Add to dashboard Cite

Lane detection plays a vital role in making the idea of the autonomous car a reality. Traditional lane detection methods need extensive hand-crafted features and post-processing techniques, which make the models specific feature-oriented, and susceptible to instability for the variations on road scenes. In recent years, Deep Learning (DL) models, especially Convolutional Neural Network (CNN) models have been proposed and utilized to perform pixel-level lane segmentation. However, most of the methods focus on achieving high accuracy while considering structured roads and good weather conditions and do not put emphasis on testing their models on defected roads, especially ones with blurry lane lines, no lane lines, and cracked pavements, which are predominant in the real world. Moreover, many of these CNN-based models have complex structures and require high-end systems to operate, which makes them quite unsuitable for being implemented in embedded devices. Considering these shortcomings, in this paper, we have introduced a novel CNN model named LLDNet based on an encoder–decoder architecture that is lightweight and has been tested in adverse weather as well as road conditions. A channel attention and spatial attention module are integrated into the designed architecture to refine the feature maps for achieving outstanding results with a lower number of parameters. We have used a hybrid dataset to train our model, which was created by combining two separate datasets, and have compared the model with a few state-of-the-art encoder–decoder architectures. Numerical results on the utilized dataset show that our model surpasses the compared methods in terms of dice coefficient, IoU, and the size of the models. Moreover, we carried out extensive experiments on the videos of different roads in Bangladesh. The visualization results exhibit that our model can detect the lanes accurately in both structured and defected roads and adverse weather conditions. Experimental results elicit that our designed method is capable of detecting lanes accurately and is ready for practical implementation.

show abstract

“…The method firstly divides the 3D point cloud on the x–y plane to form a uniform point cloud column, converts it into a sparse virtual image through a learnable encoder and finally performs a convolution operation on it to extract the features. To reduce the required computing resources, Li et al [ 18 ] proposed a transformation function, which aggregates point clouds into 2D image space before using affine transformation for transformation. Then, the convolution algorithm is applied to the transformed image space.…”

Section: Related Workmentioning

confidence: 99%

GC-MLP: Graph Convolution MLP for Point Cloud Analysis

Wang

Geng

Zhou

et al. 2022

Sensors

View full text Add to dashboard Cite

With the objective of addressing the problem of the fixed convolutional kernel of a standard convolution neural network and the isotropy of features making 3D point cloud data ineffective in feature learning, this paper proposes a point cloud processing method based on graph convolution multilayer perceptron, named GC-MLP. Unlike traditional local aggregation operations, the algorithm generates an adaptive kernel through the dynamic learning features of points, so that it can dynamically adapt to the structure of the object, i.e., the algorithm first adaptively assigns different weights to adjacent points according to the different relationships between the different points captured. Furthermore, local information interaction is then performed with the convolutional layers through a weight-sharing multilayer perceptron. Experimental results show that, under different task benchmark datasets (including ModelNet40 dataset, ShapeNet Part dataset, S3DIS dataset), our proposed algorithm achieves state-of-the-art for both point cloud classification and segmentation tasks.

show abstract

3D lidar point-cloud projection operator and transfer machine learning for effective road surface features detection and segmentation

Cited by 24 publications

References 27 publications

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

LLDNet: A Lightweight Lane Detection Approach for Autonomous Cars Using Deep Learning

GC-MLP: Graph Convolution MLP for Point Cloud Analysis

Contact Info

Product

Resources

About