Search

Objectives: Shadows in high-resolution remote sensing images will cause objects information loss and image quality decline, which is not beneficial for relative applications. Current shadow compensa- tion methods often take advantage of non-shadow information around the shadow area to increase the bright- ness, whereas there is an unsolved problem that the contrast cannot be enhanced well enough and self-adap- tively. Wallis filter principle has been used in image dodging. However, when it is used in shadow compen-sation, contrast improvement is not as good as other methods, leading to poor compensation results. There- fore, an improved Wallis model compensation method is proposed in this paper to enhance the brightness and contrast better to restore the shaded information. Methods: First, by adding compensation strength and stretch parameters, an improved Wallis model is designed. The strength parameter is positive to the bright-ness and contrast, and the stretch parameter is sensitive to the contrast. Therefore, the improved Wallis model is more efficient to adjust brightness and contrast. Moreover, an automatic parameters calculation strategy is further explored to customize a suitable compensation model for each shadow area. On one hand, the brightness average and deviation of the adjacent non-shadow region are calculated and used as the compensation target values. On the other hand, based on searching the same kinds of points around shadow boundaries, a series of non-shadow and shadow feature points are matched. Assumed the feature value of the non-shadow point is the approximate value of its responding shadow point, they can be used to calculate strength and stretch parameters automatically. Lastly, the brightness of each pixel in different shadow re-gions will be compensated by customized models to cover the shaded information self-adaptively. Results: In this paper, three sets of comparative experiments are set up to analyze and compare the part compensa- tion method's algorithm, original Wallis method in three images with ordinary object shadows and cloud shadows, respectively. The differences in brightness average and gradient average between the compensated value and the non-shadow target value are used to evaluate the compensation quality. The experimental re- sults show that: (1) Original Wallis model is useful to enhance brightness to some extent, while it is insuffi- cient to improve contrast. As a consequence, the visual compensated results are not acceptable as the other two methods. (2) Part compensation method can improve brightness and contrast effectively. However, it cannot adjust self-adaptively according to each shadow region's condition, leading to some over-compensated and some insufficient compensation results in the same image. (3) The proposed method results indicate the best compensation quality in different shadow regions because the improved Wallis model is more pointed to improve contrast and the automatically calculated parameters values are suitable to customize the com-pensation model for each shadow region. Conclusions: Aiming at the contrast improvement of the current algorithm of automatic shadow compensation, an automatic compensation method based on an improved Wallis model is proposed in this paper. The experimental results show that the newly designed model is more effective to enhance brightness and contrast, which is useful for finding suitable parameters values. Combining with the strategy of automatic parameters calculation, its customized model can adaptively com-pensate each shadow. However, it should be pointed out that there are still some limitations about the shadow border compensation and the internal difference in one shadow area, which need to be further studied. [ABSTRACT FROM AUTHOR]

Objectives: A method of deep learning change detection with domain knowledge as an optimization strategy was proposed to improve the change detection precision of high-resolution remote sensing images. Methods: The improved change vector analysis algorithm and grey-level co-occurrence matrix algorithm were used to obtain the spectral and texture changes of images, and reasonable thresholds were set to divide the changed samples from the unchanged samples based on the spectral and texture change intensity maps. The pattern shape index and spectral knowledge in domain knowledge were introduced as an optimization strategy to filter the changed samples for obtaining high-quality training samples. The deep belief net⁃ work model was constructed and trained, and the results of deep learning change detection were optimized by the optimization strategy to reduce the influence of “salt and pepper noise” and false change zones on the detection accuracy. Results: The Results of change detection experiments show that the accuracies of Gaofen⁃2 and IKONOS images were increased by 7.58% and 14.69% and the recall by 17.08% and 23.87%, respectively, while the false alarms and were decreased by 30.22% and 23.30% and the missing alarms by 17.08% and 23.87%, respectively. Conclusions: Compared with the method before the optimization strategy was adopted, the proposed method in this paper can effectively improve the precision of change detection, and it provides a new way of using remote sensing images to improve the precision of deep learning change detection. [ABSTRACT FROM AUTHOR]

In view of the existing machine learning methods in the field of high -resolution remote sensing image building extraction and the like, which requires positive and negative training samples to participate at the same time, a one- class building change detection algorithm based on one -class samples without the need for negative samples is proposed. Firstly, it extracts the morphological building index features of the image, and fuse multifeatures with the spectral features. Secondly, based on the one - class classification method proposed in this paper, from the object-based perspective, it gets the object-level building change detection results. Finally, it constructs a new shape feature which is refined to obtain the final building change detection result. Through experiments on multi-source high-resolution remote sensing images, it is verified that our proposed algorithm is robust and has better detection accuracy than existing building change detection algorithms. [ABSTRACT FROM AUTHOR]

Objective: The building collapse is the main cause of casualties and economic losses in all kinds of disasters. Accurate assessment of the collapse of buildings is an important basis for disaster rescue and government decision-making, and plays an important role in the proper placement of the victims, the protection of the lives and property of the victims. Therefore, the rapid assessment of building collapse after disaster is a research hotspot in the field of remote sensing and civil engineering. The main Objective: of this study was to apply the deep learning technology to automatically identify and assess building collapse and structural damage with on-site images. Methods: The building collapse assessment method used in this paper is ResNet (residual network). A ResNet-50 architecture was implemented in TensorFlow, which consists of 5 stages each with a convolution and identity block, 3 convolution layers in each convolution block and 3 convolution layers in each identity block. According to the building damage assessment standard developed by the China Earthquake Administration, the evaluation output of this model is adjusted to three types: Completely collapsed, semi-collapsed, and non-collapse. In the data augmentation process, geometry transformation and blur processing were used to simulate different shooting environments, to reduce the overfitting of the classifier model by enhancing the training image. Through data expansion, the number of training data has been increased by at least 10 times on the basis of PEER Hub ImageNet Challenge open dataset. In the process of transfer learning, the gradient descent algorithm was used to adjust and optimize the parameters of the neural network. The Results: of this method were evaluated by using accuracy and Kappa coefficient. Results: Finally, in the building collapse assessment experiment, the classification accuracy of the assessment model reached 77.23% and the Kappa coefficient was 0.65; in the structural damage assessment experiment, the classification accuracy of the assessment model reached 77% and the Kappa coefficient was 0.693.The proposed collapse classification successfully attains a relatively high rate for true-positives. Overall, this method performed well. This method shows great promise in supporting on-site decision-making for the rapid accurate assessment of building collapse after disaster. Conclusions: A novel automatic classification method for post-disaster building collapse evaluation is proposed,which meets the building damage evaluation standard. The method is demonstrated on a specific example classification focused on collapse classification and structural damage type identification.The experiment Results: show great promise in the rapid accurate assessment of building collapse after disaster. In the future we plan to improve the model to enhance the ability to extract subtle features in the image. [ABSTRACT FROM AUTHOR]

The generation of initial seamline network is one key step in image mosaicking. The quality of generation has a massive impact on subsequent local optimization. In this paper, a novel method for generating initial seamline network is designed. The overlapping information of effective regions of images is encapsulated in the form of a bit array and assigns to the geometric elements such as points, lines and faces. The design contributes to excavate the spatial relation between overlapping areas and to facilitate the calculation of joint priorities of overlapping areas. The main consideration of seam line generation is no longer subject to the specific shape of overlapping areas, but only the degree of overlap. Therefore, it greatly improves the robustness of seam line generation. In addition, the simple bitwise operation reduces the complexity of algorithm and thus improves the computational efficiency. This method is compared with the area voronoi diagram with overlap method. The experiments show that the proposed method has advantages in robustness and efficiency of seam line generation. [ABSTRACT FROM AUTHOR]