A novel potential illegal construction (PIC) detection method by bitemporal unmanned aerial vehicle (UAV ) image comparison (change detection) within building roof areas is proposed. In this method, roofs are first extracted from UAV images using a depth-channel improved UNet model. A two-step change detection scheme is then implemented for PIC detection. In the change detection stage, roofs with appearance, disappearance, and shape changes are first extracted by morphological analysis. Subroof primitives are then obtained by roof-constrained image segmentation within the remaining roof areas, and object-based iteratively reweighted multivariate alteration detection (IR-MAD ) is implemented to extract the small PICs from the subroof primitives. The proposed method organically combines deep learning and object-based image analysis, which can identify entire roof changes and locate small object changes within the roofs. Experiments show that the proposed method has better accuracy compared with the other counterparts, including the original IR-MAD, change vector analysis, and principal components analysis-K-means.

The development of global urbanization has brought about a significant amount of solid waste. These untreated wastes may be dumped in any corner, causing serious pollution to the environment. Thus, it is necessary to accurately obtain their distribution locations and detailed edge information. In this study, a practical deep learning network for recognizing solid waste piles over extensive areas using unmanned aerial vehicle (UAV) imagery has been proposed and verified. Firstly, a high-resolution dataset serving to solid waste detection was created based on UAV aerial data. Then, a dual-branch solid waste semantic segmentation model was constructed to address the characteristics of the integration of solid waste distribution with the environment and the irregular edge morphology. The Context feature branch is responsible for extracting high-level semantic features, while the Spatial feature branch is designed to capture fine-grained spatial details. After information fusion, the model obtained more comprehensive feature representation and segmentation ability. The effectiveness of the improvement was verified through ablation experiments and compared with 13 commonly used semantic segmentation models, demonstrating the advantages of the method in solid waste segmentation tasks, with an overall accuracy of over 94%, and a recall rate of 88.6%—much better than the best performing baselines. Finally, a spatial distribution map of solid waste over Jiaxing district, China was generated by the model inference, which assisted the environmental protection department in completing environmental management. The proposed method provides a feasible approach for the accurately monitoring of solid waste, so as to provide policy support for environmental protection.

Change detection (CD) remains an important issue in remote sensing applications, especially for high spatial resolution (HSR) images, but it has yet to be fully resolved. This work proposes a novel object-based change detection (OBCD) method for HSR images that is based on region–line primitive association analysis and evidence fusion. In the proposed method, bitemporal images are separately segmented, and the segmentation results are overlapped to obtain the temporal region primitives (TRPs). The temporal line primitives (TLPs) are obtained by straight line detection on bitemporal images. In the initial CD stage, Dempster–Shafer evidence theory fuses the multiple items of evidence of the TRPs’ spectrum, edge, and gradient changes, and obtains the initial changed areas. In the refining CD stage, the association between the TRPs and their contacting TLPs in the unchanged areas is established on the basis of the region–line primitive association framework, and the TRPs’ main line directions (MLDs) are calculated. Some changed TRPs omitted in the initial CD stage are recovered by their MLD changes, thereby refining the initial CD results. Different from common OBCD methods, the proposed method considers the change evidence of TRPs’ internal and boundary information simultaneously via information complementation between TRPs and TLPs. The proposed method can significantly reduce missed alarms while maintaining a low level of false alarms in OBCD, thereby improving total accuracy. In our experiments, our method is superior to common CD methods, including change vector analysis (CVA), PCA-k-means, and iterative reweighted multivariate alteration detection (IRMAD), in terms of overall accuracy, missed alarms, and Kappa coefficient.

Hyperspectral unmixing plays an important role in hyperspectral imagery (HSI) processing. Numerous unmixing algorithms have been proposed over the last decades. In this letter, we focus on the blind source separation model, which has drawn much attention in the hyperspectral community. However, the nonconvexity of the blind unmixing method often suffers from undesired solution. Thus, additional assumptions and regularizations are required to advance the unmixing performance. In this work, we proposed a new weighted total variation regularized blind unmixing (wtvBU) for HSI. The nonconvex sparsity-inducing function log-exp was exploited to build the weight matrix, which promotes the smooth transitions in the abundance map while preserving the spatial contextual information of the image scene. The proposed algorithm was efficiently solved via the alternating direction method of multipliers. Experimental results on two benchmark HSIs demonstrated that wtvBU achieved competitive performance when compared with other state-of-the-art unmixing algorithms.

Change detection (CD) is in demand in satellite imagery processing. Inspired by the recent success of the combined transformer-CNN (convolutional neural network) model, TransCNN, originally designed for image recognition, in this paper, we present STDecoder-CD for change detection applications, which is a combination of the Siamese network (“S”), the TransCNN backbone (“T”), and three types of decoders (“Decoder”). The Type I model uses a UNet-like decoder, and the Type II decoder is defined by a combination of three modules: the difference detector, FPN (feature pyramid network), and FCN (fully convolutional network). The Type III model updates the change feature map by introducing a transformer decoder. The effectiveness and advantages of the proposed methods over the state-of-the-art alternatives were demonstrated on several CD datasets, and experimental results indicate that: (1) STDecoder-CD has excellent generalization ability and has strong robustness to pseudo-changes and noise. (2) An end-to-end CD network architecture cannot be completely free from the influence of the decoding strategy. In our case, the Type I decoder often obtained finer details than Types II and III due to its multi-scale design. (3) Using the ablation or replacing strategy to modify the three proposed decoder architectures had a limited impact on the CD performance of STDecoder-CD. To the best of our knowledge, we are the first to investigate the effect of different decoding strategies on CD tasks.

This study analyzed the applicability of five long-term precipitation datasets in the Hang-Jia-Hu Plain of eastern China based on meteorological observation data. The accuracy of each dataset at different time scales (yearly, monthly) was analyzed. Besides, their spatiotemporal distributions and differences in precipitation event frequency were also compared. The results indicate that the high-resolution (1 day, 1 km) and long-term (1961–2019) gridded dataset for temperature and precipitation across China (HRLT) exhibited the best overall performance at the annual scale, while the long-term, gauge-based gridded precipitation dataset for the Chinese mainland (CHM_PRE) performed the best at the monthly scale. The dataset of monthly precipitation with a resolution of 1 km in China from 1960 to 2020 (HHU) and the China Meteorological Forcing Dataset (CMFD) tend to overestimate the local precipitation amounts, while the other three products are characterized by an underestimation. The mean result of the five datasets indicates a slight, statistically insignificant rise in precipitation, by 4.19 mm annually, with an overall multi-year average of 1303.28 mm. The analysis of the five datasets successfully captures the spatial precipitation patterns across the Hang-Jia-Hu Plain, characterized by higher precipitation levels in the southwest and lower in the northeast. Although the interannual variability displays general consistency, there are significant discrepancies in the interannual growth rates and the spatial distribution of significance across different regions. This study can provide a reference for the accuracy of precipitation data in the fields of hydrology, meteorology, agriculture, and ecology, facilitating the analysis of uncertainties in related research.