In the near future, the communication between autonomous cars will produce a network of sensors that will allow us to know the state of the roads in real time. Lidar technology, upon which most autonomous cars are based, allows the acquisition of 3D geometric information of the environment. The objective of this work is to use point clouds acquired by Mobile Laser Scanning (MLS) to segment the main elements of road environment (road surface, ditches, guardrails, fences, embankments, and borders) through the use of PointNet. Previously, the point cloud was automatically divided into sections in order for semantic segmentation to be scalable to different case studies, regardless of their shape or length. An overall accuracy of 92.5% has been obtained, but with large variations between classes. Elements with a greater number of points have been segmented more effectively than the other elements. In comparison with other point-by-point extraction and ANN-based classification techniques, the same success rates have been obtained for road surfaces and fences, and better results have been obtained for guardrails. Semantic segmentation with PointNet is suitable when segmenting the scene as a whole, however, if certain classes have more interest, there are other alternatives that do not need a high training cost.
Detailed environment perception is a crucial component of automated vehicles. However, to deal with the amount of perceived information, we also require segmentation strategies. Based on a grid map environment representation, well-suited for sensor fusion, free-space estimation and machine learning, we detect and classify objects using deep convolutional neural networks. As input for our networks we use a multi-layer grid map efficiently encoding 3D range sensor information. The inference output consists of a list of rotated bounding boxes with associated semantic classes. We conduct extensive ablation studies, highlight important design considerations when using grid maps and evaluate our models on the KITTI Bird's Eye View benchmark. Qualitative and quantitative benchmark results show that we achieve robust detection and state of the art accuracy solely using top-view grid maps from range sensor data.
20 December 2020 in International Journal of Remote Sensing
Macroalgae are a fundamental component of coastal ecosystems and play a key role in shaping community structure and functioning. Macroalgae are currently threatened by diverse stressors, particularly climate change and invasive species, but they do not all respond in the same way to the stressors. Effective methods of collecting qualitative and quantitative information are essential to enable better, more efficient management of macroalgae. Acquisition of high-resolution images, in which macroalgae can be distinguished on the basis of their texture and colour, and the automated processing of these images are thus essential. Although ground images are useful, labelling is tedious. This study focuses on the semantic segmentation of five macroalgal species in high-resolution ground images taken in 0.5 × 0.5 m quadrats placed along an intertidal rocky shore at low tide. The target species, Bifurcaria bifurcata, Cystoseira tamariscifolia, Sargassum muticum, Sacchoriza polyschides and Codium spp., which predominate on intertidal shores, belong to different morpho-functional groups. An explanation of how to convert vector-labelled data to raster-labelled data for adaptation to Convolutional Neural Network (CNN) input is provided. Three CNNs (MobileNetV2, Resnet18, Xception) were compared, and ResNet18 yielded the highest accuracy (91.9%). The macroalgae were correctly segmented, and the main confusion occurred at the borders between different macroalgal species, a problem derived from labelling errors. In addition, the interior and exterior of the quadrats were correctly delimited by the CNNs. The results were obtained from only one hundred labelled images and the method can be performed on personal computers, without the need to use external servers. The proposed method helps automation of the labelling process.
1 February 2019 in ISPRS Journal of Photogrammetry and Remote Sensing