Unsupervised change detection of land cover in multispectral satellite remote sensing images with a spatial resolution of 2–5 m has always been a challenging task. This paper presents a method of detecting land cover changes in high-spatial-resolution remote sensing imagery. This method has three characteristics: (1) Extended center-symmetric local binary pattern (XCS-LBP) is used to extract image features to emphasize spatial context information in initial change detection. Then, spectral information is combined to improve the accuracy of change detection. (2) The local histogram distance of XCS-LBP features is used as the change vector to improve the expression of change information. (3) A progressive Otsu method is developed for threshold segmentation of the change vector to reduce the false detection rate. Four datasets with different landscape complexities and seven state-of-the-art unsupervised change detection methods were used to test the performance of the proposed method. Quantitative results showed that the proposed method reduced the false detection rate and improved the accuracy of the detection of land cover changes. The F1 score achieved by the proposed method reached 0.8688, 0.8867, 0.7725, and 0.6634, respectively, which are higher than the highest corresponding F1 score achieved by the benchmark methods (0.8533, 0.8549, 0.6545, and 0.5895, respectively).

It has long been acknowledged that crimes of the same type tend to be committed at the same location or proximity in a short period. However, the investigation of whether this phenomenon exists across crime types remains limited. The spatial-temporal clustered patterns for two types of crimes in public areas (pocket-picking and vehicle/motor vehicle theft) are separately examined. Compared with existing research, this study contributes to current research from three aspects: (1) The repeat and near-repeat phenomenon exists in two types of crimes in a large Chinese city. (2) A significant spatial-temporal interaction between pocket-picking and vehicle/motor vehicle theft exists within a range of 100 m. Some cross-crime type interactions seem to have a stronger ability of prediction than does single-crime type interaction. (3) A risk-avoiding activity is identified after spatial-temporal hotspots of another crime type. The spatial extent with increased risk is limited to a certain distance from the previous hotspots. The experimental results are analyzed and interpreted with current criminology theories.

Repeat and near-repeat victimization are important concepts in the study of crime. The incidence of repeat offenses within a single type of crime has been confirmed. However, the study of the circumstances existing across crime types requires further investigation. This article investigates whether the phenomenon of near-repeat crime exists in different types of crime by studying the spread of crime risk within different crime types. Taking Suzhou City as the research area, a DBSCAN-based algorithm is proposed, which can detect a large number of important and stable hotspots through the multi-density self-adaptation of algorithm parameters. Pearson correlation is used to analyze the risk correlation between different types of crime. In different crime hotspots, the types of crime and the spread of crime risk among different types is also different. After a crime occurs, identifying the risk can aid crime prevention.

Empirical studies have focused on investigating the interactive relationships between crime pairs. However, many other types of crime patterns have not been extensively investigated. In this paper, we introduce three basic crime patterns in four combinations. Based on graph theory, the subgraphs for each pattern were constructed and analyzed using criminology theories. A Monte Carlo simulation was conducted to examine the significance of these patterns. Crime patterns were statistically significant and generated different levels of crime risk. Compared to the classical patterns, combined patterns create much higher risk levels. Among these patterns, “co-occurrence, repeat, and shift” generated the highest level of crime risk, while “repeat” generated much lower levels of crime risk. “Co-occurrence and shift” and “repeat and shift” showed undulated risk levels, while others showed a continuous decrease. These results outline the importance of proposed crime patterns and call for differentiated crime prevention strategies. This method can be extended to other research areas that use point events as research objects.

Effective predictive policing can guide police patrols and deter crime. Hourly crime prediction is expected to save police time. The selection of spatial resolution is important due to its strong relationship with the accuracy of crime prediction. In this paper, we propose an adaptive spatial resolution method to select the best spatial resolution for hourly crime prediction. The ST-ResNet model is applied to predict crime risk, with historical crime data and weather data as predictive variables. A prediction accuracy index (PAI) is used to evaluate the accuracy of the results. Data on property crimes committed in Suzhou, a big city in China, were selected as the research data. The experiment results indicate that a 2.4 km spatial resolution leads to the best performance for crime prediction. The adaptive spatial resolution method can be used to guide police deployment.