A pixel-based algorithm for multi-temporal Landsat (TM/ETM+/OLI/OLI-2) imagery between 1990 and 2022 monitored mangrove dynamics and detected their changes in the three provinces (i.e., Thai Binh, Nam Dinh and Hai Phong), which are located on the Northern coast of Vietnam, through the Google Earth Engine (GEE) cloud computing platform. Results showed that the mangrove area in the study area decreased from 2960 ha in 1990 to 2408 ha in 1995 and then significantly increased to 4435 ha in 2000 but later declined to 3502 ha in 2005. The mangrove areas experienced an increase from 4706 ha in 2010 to 10,125 ha in 2020 and reached a highest peak of 10,630 ha in 2022. In 2022, Hai Phong province had the largest area of mangrove (3934 ha), followed by Nam Dinh (3501 ha) and Thai Binh (3195 ha) provinces. The overall accuracies for 2020 and 2022 were 94.94% and 91.98%, while the Kappa coefficients were 0.90 and 0.84, respectively. The mangrove restoration programs and policies by the Vietnamese government and local governments are the key drivers of this increase in mangroves in the three provinces from 1990 to 2022. The results also demonstrated that the combination of Landsat time series images, a pixel-based algorithm, and the GEE platform has a high potential for monitoring long-term change of mangrove forests during 32 years in the tropics. Moreover, the obtained mangrove forest maps at a 30-m spatial resolution can serve as a useful and up-to-date dataset for sustainable management and conservation of these mangrove forests in the Red River Delta, Vietnam.

In this letter, we investigate fine-grained classification of arbitrary-oriented ships in very high resolution optical imagery using Bag of Word model with Spatial Pyramid (SP-BoW). Given that based on ‘spatial pyramid’ of the histogram of local features, the final feature vectors not only count the multiplicity of ‘words’ but also represent their spatial topology. We attempt to improve the performance of this model by introducing augmented data for training phase. Our aim is to make the dataset big enough to be able to capture holistic variation of ship orientation. Three data augmentation operations are used including random rotate by an angle of modulo 90°, random flip-left-right, and random flip-top-bottom. Through this procedure, our trained SP-BoW model is able to get better generalization. The proposed approach is validated on the High-Resolution Ship Collections 2016 (HRSC2016) ship dataset. The results indicate that training on augmented data can significantly improve the performance of SP-BoW. Beside, compared to other state-of-the-art convolutional neural network-based approaches, the approach proposed in this research has yielded competitive results and could make it a good baseline for evaluating more sophisticated CNN architecture in the future.