Fusion of the low-light-level visible and infrared images for night-vision context enhancement

For better night-vision applications using the low-light-level visible and infrared imaging, a fusion framework for night-vision context enhancement（FNCE） method is proposed. An adaptive brightness stretching method is first proposed for enhancing the visible image. Then, a hybrid multi-scale decomposition with edge-preserving filtering is proposed to decompose the source images. Finally, the fused result is obtained via a combination of the decomposed images in three different rules. Experimental results demonstrate that the FNCE method has better performance on the details（edges）, the contrast, the sharpness, and the human visual perception. Therefore,better results for the night-vision context enhancement can be achieved.

Fish Density Estimation with Multi-Scale Context Enhanced Convolutional Neural Network

With the development of fishery industry,accurate estimation of the number of fish in aquaculture waters is of great importance to fish behavior analysis,bait feeding and fishery resource investigation.In this paper,we propose a method for fish density estimation based on the multi-scale context enhanced convolutional network,which could map a fish school image taken at any angle to a density map,and calculate the number of fish in the image finally.In order to eliminate the influence of camera perspective effect and image resolution on density estimation,multi-scale filters are utilized in a convolutional neural network to process fish image in parallel.And then,the context enhancement module is merged in the network structure to help the network understand the global context information of the image.Finally,different feature maps are merged together to construct the density map of fish school images,and finally get the number of fish in the image.In order to make the effectiveness of our method valid,we test the proposed method on DlouDataset.The results show that the proposed method has lower mean square error and mean absolute error,which is helpful to improve the accuracy of the fish counting in dense fish school images.

Detecting geo-relation phrases from web texts for triplet extraction of geographic knowledge:a context-enhanced method

As an effective organization form of geographic information,a geographic knowledge graph(GeoKG)facilitates numerous geography-related analyses and services.The completeness of triplets regarding geographic knowledge determines the quality of GeoKG,thus drawing considerable attention in the related domains.Mass unstructured geographic knowledge scattered in web texts has been regarded as a potential source for enriching the triplets in GeoKGs.The crux of triplet extraction from web texts lies in the detection of key phrases indicating the correct geo-relations between geo-entities.However,the current methods for key-phrase detection are ineffective because the sparseness of the terms in the web texts describing geo-relations results in an insufficient training corpus.In this study,an unsupervised context-enhanced method is proposed to detect geo-relation key phrases from web texts for extracting triplets.External semantic knowledge is introduced to relieve the influence of the sparseness of the georelation description terms in web texts.Specifically,the contexts of geo-entities are fused with category semantic knowledge and word semantic knowledge.Subsequently,an enhanced corpus is generated using frequency-based statistics.Finally,the geo-relation key phrases are detected from the enhanced contexts using the statistical lexical features from the enhanced corpus.Experiments are conducted with real web texts.In comparison with the well-known frequency-based methods,the proposed method improves the precision of detecting the key phrases of the geo-relation description by approximately 20%.Moreover,compared with the well-defined geo-relation properties in DBpedia,the proposed method provides quintuple key-phrases for indicating the geo-relations between geo-entities,which facilitate the generation of new triplets from web texts.

Enhanced context encoding for single image raindrop removal

Despite the great success achieved by convolutional neural networks in addressing the raindrop removal problem,the still relatively blurry results call for better problem formulations and network architectures.In this paper,we revisited the rainy-to-clean translation networks and identified the issue of imbalanced distribution between raindrops and varied background scenes.None of the existing raindrop removal networks consider this underlying issue,thus resulting in the learned representation biased towards modeling raindrop regions while paying less attention to the important contextual regions.With the aim of learning a more powerful raindrop removal model,we propose learning a soft mask map explicitly for mitigating the imbalanced distribution problem.Specifically,a two stage network is designed with the first stage generating the soft masks,which helps learn a context-enhanced representation in the second stage.To better model the heterogeneously distributed raindrops,a multi-scale dense residual block is designed to construct the hierarchical rainy-to-clean image translation network.Comprehensive experimental results demonstrate the significant superiority of the proposed models over state-of-the-art methods.