Airborne laser scanning (ALS) point cloud has been widely used in the fields of ground powerline surveying, forest monitoring, urban modeling, and so on because of the great convenience it brings to people’s daily life. However, the sparsity and uneven distribution of point clouds increases the difficulty of setting uniform parameters for semantic classification. The PointNet++ network is an end-to-end learning network for irregular point data and highly robust to small perturbations of input points along with corruption. It eliminates the need to calculate costly handcrafted features and provides a new paradigm for 3D understanding. However, each local region in the output is abstracted by its centroid and local feature that encodes the centroid’s neighborhood. The feature learned on the centroid point may not contain relevant information of itself for random sampling, especially in large-scale neighborhood balls. Moreover, the centroid point’s global-level information in each sample layer is also not marked. Therefore, this study proposed a modified PointNet++ network architecture which concentrates the point-level and global features on the centroid point towards the local features to facilitate classification. The proposed approach also utilizes a modified Focal Loss function to solve the extremely uneven category distribution on ALS point clouds. An elevation- and distance-based interpolation method is also proposed for the objects in ALS point clouds which exhibit discrepancies in elevation distributions. The experiments on the Vaihingen dataset of the International Society for Photogrammetry and Remote Sensing and the GML(B) 3D dataset demonstrate that the proposed method which provides additional contextual information to support classification achieves high accuracy with simple discriminative models and new state-of-the-art performance in power line categories.

Currently, deep neural networks have gained impressive performance for object detection when plenty of labeled training data are available. However, a substantial quantity of labeled data is typically rare and time-consuming to obtain in deep space exploration. To possibly alleviate this problem, we use domain adaptation (DA) to effectively detect unannotated real data samples in the crater detection problem with the help of autoannotated synthetic data samples only. Specifically, we present a novel network, namely, CraterDANet, which unifies image- and feature-level adversarial DAs to bridge the domain gap between synthetic and real data. To evaluate our method, we present a new lunar crater dataset that contains nearly 20 000 small-scale craters captured from different Lunar Reconnaissance Orbital (LRO) Narrow Angle Camera (NAC) grayscale images that have large variations in shape, size, overlap, degradation, and illumination conditions. To the best of our knowledge, this is the first lunar crater dataset with diversification of illumination conditions, which provides bounding box annotations. Through experiments, the proposed CraterDANet, which was trained on labeled synthetic data without using any additional real labeled data, achieved an $F1$ -score of 80.27%, which is slightly better than traditional supervised methods trained on real data. Although the performance still falls short of the state-of-the-art Faster R-convolutional neural network (CNN), it achieves an $F1$ -score of +5.98% performance improvement compared with Faster R-CNN trained on synthetic data only. This result demonstrates the effectiveness of the proposed method for reducing the domain gap between synthetic and real data.

Unmanned aerial vehicles (UAV) have been widely used to capture and down-link real-time videos/images. However, their role as a low-cost airborne platform for capturing high-resolution, geo-referenced still imagery has not been fully utilized. The images obtained from UAV are advantageous over remote sensing images as they can be obtained at a low cost and potentially no risk to human life. However, these images are distorted due to the noise generated by the rotary wings which limits the usefulness of such images. One potential application of such images is to detect changes between the images of the same area which are collected over time. Change detection is of widespread interest due to a large number of applications, including surveillance and civil infrastructure. Although UAVs can provide images with high resolution in a portable and easy way, such images only cover small parts of the entire field of interest and are often with high deformation. Until now, there is not much application of change detection for UAV images. Also the traditional pixel-based change detection method does not give satisfactory results for such images. In this paper, we have proposed a novel object-based method for change detection using UAV images which can overcome the effect of deformation and can fully utilize the high resolution capability of UAV images. The developed method can be divided into five main blocks: pre-processing, image matching, image segmentation and feature extraction, change detection and accuracy evaluation. The pre-processing step is further divided into two sub-steps: the first sub-step is to geometrically correct the bi-temporal image based on the geo-reference information (GPS/INS) installed on the UAV system, and the second sub-step is the radiometric normalization using a histogram method. The image matching block uses the well-known scale-invariant feature transform (SIFT) algorithm to match the same areas in the images and then resample them. The image segmentation and feature extraction block is used to separate the images to different meaningful regions by the mean shift method, extract the textural features and contextual features (polygon,etc.) , Based on the features extracted above as well as the SIFT features of the area, the optimization result is achieved by considering the neighbourhood information .The proposed method is being tested by using multi-temporal images acquired by UAV. The results confirm the effectiveness of the proposed approach.

Structural health monitoring with wireless sensor networks has been increasingly popular in recent years because of the convenience. In this paper, a real-time monitoring system for cracks on the surface of reactor containment buildings is presented. Customized wireless sensor networks platforms are designed and implemented with sensors especially for crack monitoring, which include crackmeters and temperature detectors. Software protocols like route discovery, time synchronization and data transfer are developed to satisfy the requirements of the monitoring system and stay simple at the same time. Simulation tests have been made to evaluate the performance of the system before full scale deployment. The real-life deployment of the crack monitoring system is carried out on the surface of reactor containment building in Daya Bay Nuclear Power Station during the in-service pressure test with 30 wireless sensor nodes.

As an external defect, cracks cannot be neglected, and crack detection is crucial for the safety monitoring of nuclear containments. Because of the low continuity and tiny size of cracks, traditional crack-detection methods sometimes fail to detect cracks in complex backgrounds. This paper proposes a crack-detection method featuring a template that can adaptively determine the crack width. The search starts from the local minimum point of the gray level, and the crack width is determined by comparing the gray difference between the seed region and its neighborhoods on a pixel basis. Then, the block-iterative algorithm is used to intersect with the results of the template method followed by the morphological method for further noise removal. Compared with the traditional methods, the developed method can detect tiny cracks with a higher recognition ratio. The applicability of this method was verified by experiments from a nuclear containment in southern China.

Deep learning in 3-D classification tasks focuses on the designs of comprehensive local aggregation operators. Airborne laser scanning (ALS) point clouds have their own characteristics: 1) the object overlaps in the vertical direction; 2) the spatial density is uneven; and 3) the objects present large-scale variations between different categories. However, the dynamic graph convolutional neural network (DGCNN) ignores the inherent properties of ALS point clouds. This study modifies the DGCNN network and proposes a new neural network module called Webconv, which densely connects point pairs in multiscale local regions to learn contextual information. One modified cross entropy loss function with variable weight is proposed to solve the problem of uneven category distributions in ALS points. Because preprocessing such as block partition ignores the context information of the whole region, the conditional random field is used to refine the point cloud. Our approach achieves state-of-the-art performance on the dataset of the 2019 IEEE GRSS Data Fusion Contest 3-D Point Cloud Classification Challenge and can be widely used for ALS point classification.