Ubiquitous trajectory data provides new opportunities for production and update of the road network. A number of methods have been proposed for road network construction and update based on trajectory data. However, existing methods were mainly focused on reconstruction of the existing road network, and the update of newly added roads was not given much attention. Besides, most of existing methods were designed for high sampling rate trajectory data, while the commonly available GPS trajectory data are usually low-quality data with noise, low sampling rates, and uneven spatial distributions. In this paper, we present an automatic method for detection and update of newly added roads based on the common low-quality trajectory data. First, additive changes (i.e., newly added roads) are detected using a point-to-segment matching algorithm. Then, the geometric structures of new roads are constructed based on a newly developed decomposition-combination map generation algorithm. Finally, the detected new roads are refined and combined with the original road network. Seven trajectory data were used to test the proposed method. Experiments show that the proposed method can successfully detect the additive changes and generate a road network which updates efficiently.

Landslide susceptibility mapping (LSM) constitutes a valuable analytical instrument for estimating the likelihood of landslide occurrence, thereby furnishing a scientific foundation for the prevention of natural hazards, land-use planning, and economic development in landslide-prone areas. Existing LSM methods are predominantly data-driven, allowing for significantly enhanced monitoring accuracy. However, these methods often overlook the consideration of landslide mechanisms and uncertainties associated with non-landslide samples, resulting in lower model reliability. To effectively address this issue, a knowledge-guided landslide susceptibility assessment framework is proposed in this study to enhance the interpretability and monitoring accuracy of LSM. First, a landslide knowledge graph is constructed to model the relationships between landslide entities and summarize landslide susceptibility rules. Next, combining the obtained landslide rules with geographic similarity principles, high-confidence non-landslide samples are selected to optimize the quality of the samples. Subsequently, a Landslide Knowledge Fusion Cell (LKF-Cell) is utilized to couple landslide data with landslide knowledge, resulting in the acquisition of informative and semantically rich landslide event features. Finally, a precise and credible landslide susceptibility assessment model is built based on a convolutional neural network (CNN), and landslide susceptibility spatial distribution levels are mapped. The research findings indicate that the CNN-based model outperforms traditional machine learning algorithms in predicting landslide probability; in particular, the Area Under the Curve (AUC) of the model was improved by 3–6% after sample optimization, and the AUC value of the LKF-Cell method was 6–11% higher than the baseline method.

Few-shot remote sensing scene classification (FSRSSC) has been used for new class recognition in the presence of a limited number of labeled samples. The representation vector (prototype) of categories obtained using images only confronts some challenges, such as insufficient generalization when the number of samples is too small. To address this problem, we propose a new FSRSSC method based on prototype networks, named CNSPN, which combines semantic information of class names (name of the scene categories, such as aircraft, harbor, and bridge). First, CNSPN extracts semantics for class names using a pre-trained word-embedding model, which enriches the feature representation ability of the category at the source. Then, an enhanced fusion prototype is generated by fusing the semantic information of text and visual information in the image through a multimodal prototype fusion module (MPFM). Finally, the query image is classified by measuring the distance between the query sample and the visual prototype, and between the query sample and the fusion prototype. Comparative experiments on the NWPU-RESISC45 and RSD46-WHU datasets show that the proposed method significantly improves FSRSSC performance. Code is available at https://github.com/RS-CSU/CNSPN.git.

As a hot research topic in urban geography, spatiotemporal interaction analysis has been used to detect the hotspot mobility patterns of crowds and urban structures based on the origin-destination (OD) flow data, which provide useful information for urban planning and traffic management applications. However, existing methods mainly focus on the detection of explicit spatial interaction patterns (such as spatial flow clusters) in OD flow data, with less attention to the discovery of underlying crowd travel demands. Therefore, this paper proposes a framework to discover the crowd travel demands by associating the dynamic spatiotemporal interaction patterns and the contextual semantic features of the geographical environment. With urban functional zones (UFZs) as the basic units of human mobility in urban spaces, this paper gives a case study in Wuhan, China, to detect and interpret the human mobility patterns based on the characteristics of spatiotemporal interaction between UFZs. Firstly, we build the spatiotemporal interaction matrix based on the OD flows of different UFZs and analyze the characteristics of the interaction matrix. Then, hotspot poles, defined as the local areas where people gather significantly, are extracted using the Gi-statistic-based spatial hotspot detection algorithm. Next, we develop a frequent interaction pattern mining method to detect the frequent interaction patterns of the hotspot poles. Finally, based on the detected frequent interaction patterns, we discover the travel demands of crowds with semantic features of corresponding urban functional zones. The characteristics of crowd travel distance and travel time are further discussed. Experiments with floating car data, road networks, and POIs in Wuhan were conducted, and results show that the underlying travel demands can be better discovered and interpreted by the proposed framework and methods in this paper. This study helps to understand the characteristics of human movement and can provide support for applications such as urban planning and facility optimization.

Hotspot detection is an important exploratory technique to identify areas with high concentrations of crime and help deploy crime-reduction resources. Although a variety of methods have been developed to detect crime hotspots, few studies have systematically evaluated the performance of various methods, especially in terms of the ability to detect complex-shaped crime hotspots. Therefore, in this study, a comparative study of hotspot detection approaches while simultaneously considering the concentration and shape characteristics was conducted. Firstly, we established a framework for quantitatively evaluating the performance of hotspot detection for cases with or without the ”ground truth”. Secondly, accounting for the concentration and shape characteristics of the hotspot, we additionally defined two evaluation indicators, which can be used as a supplement to existing evaluation indicators. Finally, four classical hotspot-detection methods were quantitatively compared on the synthetic and real crime data. Results show that the proposed evaluation framework and indicators can describe the size, concentration and shape characteristics of the detected hotspots, thus supporting the quantitative comparison of different methods. From the selected methods, the AMOEBA (A Multidirectional Optimal Ecotope-Based Algorithm) method was more accurate in describing the concentration and shape characteristics and was powerful in discovering complex hotspots.

The detection of anomalous paths from floating vehicle trajectories plays an increasingly important role in dynamic path planning because of its ability to identify fraudulent transport behaviors, unexpected traffic accidents, and traffic restriction areas. Existing studies mostly emphasize the significant deviations based on global shape metrics and cannot manage trajectories with missing segments. To address these problems, this study proposes a new sequential pattern mining based method for detecting anomalous paths (SPDAP) in floating vehicle trajectories. First, discrete trajectory points were transformed into consecutive road segment sequences using map matching and essential empirical information. Focusing on these segment sequences, this method constructs a directed graph and calculates the conditional occurrence probabilities of multi-order sub-paths on the graph. On this basis, we designed a sequential pattern mining algorithm by estimating kernel density distributions to adaptively extract multi-order frequent sub-paths. Finally, these anomalous paths can be identified by matching the operations with frequent paths. Comparative experiments on simulated trajectories demonstrated that SPDAP could accurately, adequately, and adaptively detect anomalous paths. Additionally, using real-life taxi trajectories, we conducted semantic analysis by considering travel time and charges to further derive four patterns from the detected anomalous paths. These patterns are expected to support the identification of fraudulent taxi trips and the discovery of candidate driving routes with a preference for saving time or charge. In the future, we will focus on tracing the intrinsic causes of anomalous paths by introducing multisource information and performing real-time anomalous path detection using adaptive time intervals.