Road damage detection is essential to the maintenance and management of roads. The morphological road damage contains a large number of multi-scale features, which means that existing road damage detection algorithms are unable to effectively distinguish and fuse multiple features. In this paper, we propose a dense multiscale feature learning Transformer embedding cross-shaped attention for road damage detection (DMTC) network, which can segment the damage information in road images and improve the effectiveness of road damage detection. Our DMTC makes three contributions. Firstly, we adopt a cross-shaped attention mechanism to expand the perceptual field of feature extraction, and its global attention effectively improves the feature description of the network. Secondly, we use the dense multi-scale feature learning module to integrate local information at different scales, so that we are able to overcome the difficulty of detecting multiscale targets. Finally, we utilize a multi-layer convolutional segmentation head to generalize the previous feature learning and get a final detection result. Experimental results show that our DMTC network could segment pavement pothole patterns more accurately and effectively than other methods, achieving an F1 score of 79.39% as well as an OA score of 99.83% on the cracks-and-potholes-in-road-images-dataset (CPRID).

Geosynchronous equatorial orbit (GEO) satellite-derived AOD possesses huge advantages for monitoring atmospheric aerosol with high frequency; however, the data missing existing in the satellite-derived AOD products dramatically limits this expected advantage due to cloud obscuration and aerosol retrieval algorithm. In recent years, numerous AOD fusion algorithms have been proposed, while these algorithms are mostly developed to blend daily AOD products derived from low Earth orbit (LEO) satellites and generally neglect discrepancies from different categories of products. Therefore, a spatiotemporal fusion framework based on the Bayesian maximum entropy theorem, blending GEO with LEO satellite observations and incorporating data discrepancies (GL-BME), is developed to complementarily recover the Advanced Himawari-8 Imager (AHI) AOD products over East Asia. The results show that GL-BME significantly improves the average spatial completeness of AOD from 20.3% to 67.6% with ensured reliability, and the accuracy of merged AODs nearly maintains that of original AHI AODs. Moreover, a comparison of the monthly aerosol spatial distribution between the merged and original AHI AODs is conducted to evaluate the performance and significance of GL-BME, which indicates that GL-BME could further restore the real atmospheric aerosol situation to a certain extent on the basis of dramatic spatial coverage improvement.

The atmospheric boundary layer provides useful information about the accumulation and diffusion of pollutants. As a fast method, remote sensing techniques are used to retrieve the atmospheric boundary layer height (ABLH). Atmospheric detection lidar has been widely applied for retrieving the ABLH by providing information on the vertical distribution of aerosols. However, these previous algorithms that rely on gradient change are susceptible to residual layers. Contrary to the use of gradient change to retrieve ABLH, in this paper, we propose using a cluster analysis approach through multifunction lidar remote sensing techniques due to its increasing availability. The clustering algorithm for multi-wavelength lidar data can be divided into two parts: characteristic signal selection and selection of the classifier. First, since the separability of each type of signal is different, careful selection of the input characteristic signal is important. We propose using Fourier transform for all the observed signals; the most suitable characteristic signal can be determined based on the dispersion degree of the signal in the frequency domain. Then, the performances of four common classifiers (K-means method, Gaussian mixture model, hierarchical cluster method (HCM), and density-based spatial clustering of applications with noise) are evaluated by comparing with the radiosonde measurements from June 2015 to June 2016. The results show that the performance of the HCM classifier is the best under all states (R² = 0.84 and RMSE = 0.18 km). The findings obtained here offer insight into ABLH remote sensing technology.