A Deep Learning Approach on Building Detection from Unmanned Aerial Vehicle-Based Images in Riverbank Monitoring.

Buildings along riverbanks are likely to be affected by rising water levels, therefore the acquisition of accurate building information has great importance not only for riverbank environmental protection but also for dealing with emergency cases like flooding. UAV-based photographs are flexible and cloud-free compared to satellite images and can provide very high-resolution images up to centimeter level, while there exist great challenges in quickly and accurately detecting and extracting building from UAV images because there are usually too many details and distortions on UAV images. In this paper, a deep learning (DL)-based approach is proposed for more accurately extracting building information, in which the network architecture, SegNet, is used in the semantic segmentation after the network training on a completely labeled UAV image dataset covering multi-dimension urban settlement appearances along a riverbank area in Chongqing. The experiment results show that an excellent performance has been obtained in the detection of buildings from untrained locations with an average overall accuracy more than 90%. To verify the generality and advantage of the proposed method, the procedure is further evaluated by training and testing with another two open standard datasets which have a variety of building patterns and styles, and the final overall accuracies of building extraction are more than 93% and 95%, respectively.

WDHD1 is over-expressed in nasopharyngeal carcinoma and may control the expression of ITGAV.

Nasopharyngeal carcinoma (NPC) is a highly metastatic and invasive malignant tumor that originates in the nasopharynx. The DNA-binding protein WD repeat and HMG-box DNA-binding protein 1 (WDHD1) are highly expressed in a variety of tumours, but its expression and mechanism of action in NPC have not been reported to date. To investigate the involvement of WDHD1 in NPC, we first mined databases for the gene expression profile of NPC. Immunohistochemistry (IHC) was performed on 338 cases of NPC and 112 non-NPC samples to verify the results. We report that the expression of WDHD1 is significantly elevated in NPC. ChIP-seq was used to show that integrin alpha V (ITGAV) and WDHD1 exhibit a significant binding peak in the promoter region of the ITGAV gene. The expression levels of ITGAV and WDHD1 exhibit a significant positive correlation, and IHC was performed to show that ITGAV is highly expressed in NPC. Expression of ITGAV increased after overexpression of WDHD1, suggesting that ITGAV may be a potential target gene of WDHD1. Pathway analysis showed that both genes were closely related to the cell cycle, and flow cytometry was used to further confirm that decreased expression of WDHD1 significantly increased the number of apoptotic cells. In conclusion, our results suggest that expression of WDHD1 is increased in NPC and is likely to be associated with the NPC cell cycle; thus, we propose that WDHD1 may have the potential as a target gene for primary screening and treatment of NPC.

[An object-oriented remote sensing image segmentation approach based on edge detection].

Satellite sensor technology endorsed better discrimination of various landscape objects. Image segmentation approaches to extracting conceptual objects and patterns hence have been explored and a wide variety of such algorithms abound. To this end, in order to effectively utilize edge and topological information in high resolution remote sensing imagery, an object-oriented algorithm combining edge detection and region merging is proposed. Susan edge filter is firstly applied to the panchromatic band of Quickbird imagery with spatial resolution of 0.61 m to obtain the edge map. Thanks to the resulting edge map, a two-phrase region-based segmentation method operates on the fusion image from panchromatic and multispectral Quickbird images to get the final partition result. In the first phase, a quad tree grid consisting of squares with sides parallel to the image left and top borders agglomerates the square subsets recursively where the uniform measure is satisfied to derive image object primitives. Before the merger of the second phrase, the contextual and spatial information, (e. g., neighbor relationship, boundary coding) of the resulting squares are retrieved efficiently by means of the quad tree structure. Then a region merging operation is performed with those primitives, during which the criterion for region merging integrates edge map and region-based features. This approach has been tested on the QuickBird images of some site in Sanxia area and the result is compared with those of ENVI Zoom Definiens. In addition, quantitative evaluation of the quality of segmentation results is also presented. Experiment results demonstrate stable convergence and efficiency.