Search

Objective: Deep convolutional neural network (DCNN) is widely used in automatic road extraction from high-resolution remote sensing images (HRSIs). However, the existing methods make it difficult to model the context relationship of pixels in the predicted results. To solve this problem, some studies have used fully connected conditional random field (FullCRF) to perform secondary optimization of semantic segmentation results combined with context information, but the discontinuity problem of road structure cannot be effectively improved. Methods: In order to improve the integrity of road structure, this paper proposes a short range conditional random filed (SRCRF) model combined with DCNN. SRCRF mainly includes a unary potential function based on road pre-segmentation, a binary potential function based on spectral spatial features, and a K-neighborhood mean field inference algorithm. First, the prior knowledge of road pre-segmentation results is obtained by using the powerful feature extraction capability of DCNN as the unary potential function of SRCRF. Then, the dependence of the binary potential function defined by the linear combination of Gaussian kernel functions on the surrounding nodes is modeled. The binary potential function enables the classification results to have local consistency,that is, adjacent pixels with similar spectral features have the same label. Finally, the K-neighborhood mean field inference algorithm based on the mean-field approximation inference algorithm optimizes the inference range to make full use of the spatial context information and spectral feature context information of the road, and then calculates the optimal label corresponding to each pixel based on the space and spectral feature to optimize the road accurately. The convolution method is adopted to control the inference range of SRCRF within the radius of K in order to improve the proportion of feature vectors. Results: The experimental results show that SRCRF not only alleviates the transition smoothness of FullCRF, but also alleviates the structural discontinuity and incompleteness in the road acquisition results of HRSIs. In Zimbawe-Roads dataset and Cheng-Roads dataset, F1-scores of SRCRF are increased by about 4.01% and 3.73% respectively compared with DCNN, and about 3.25% and 2.28% respectively compared with FullCRF. Conclusions: The proposed SRCRF combines the advantages of DCNN, optimizes the fully connected structure of traditional conditional random fields into the K-neighborhood structure, reduces the inference scope, and improves the proportion of feature vectors. Compared with FullCRF, SRCRF can make better use of image color features and spatial features to accurately optimize the road extraction results of deep learning output. The performance of SRCRF is improved compared with DCNN and FullCRF, and the time is shortened by one order of magnitude compared with FullCRF. In the future, we will further investigate the potential for learning Gaussian features and investigate more complex architecture of conditional random field to better capture global context information. Additionally, we are particularly interested in exploring the application potential of SRCRF in other fields, such as building extraction, vehicle extraction, lake extraction, etc. [ABSTRACT FROM AUTHOR]

Objectives: The registration of remote sensing image data and GIS ( geographic information system ) vector data is the basis of the integration of remote sensing and GIS, which is widely used in the fields of map data update, city monitoring, map change detection and so on. At present, the key to the registration of remote sensing images and vector data is the extraction of remote sensing image features. However, the existing remote sensing image feature extraction has problems such as incomplete feature extraction, which leads to registration failure or low accuracy. This paper proposes a registration method for remote sensing images and vector data based on Mask region -based convolutional neural network (Mask R -CNN). Methods: Firstly, we select the road intersection as the distinctive feature of the same name in the remote sensing image and vector data, and create a road intersection image data set to train Mask R -CNN model. Secondly, according to the geometric topological relationship, the vector data road intersections are selected as vector control points. And take the intersection control point of the vector data as the center, we use 400 × 400 pixels window to crop the remote sensing image data and input it into the Mask R-CNN model, extract the border of the road intersection in the image. The control points of the same name are determined according to the Euclidean distance between the remote sensing image and the vector data control points, and the control points of the same name are cleaned using the density-based spatial clustering of applications with noise algorithm. Finally, the affine transformation parameters are calculated according to the filtered control points of the same name to realize the registration of remote sensing image and vector data. The registration data of Shanghai vector data and Gaofen-2 image data were selected for registration experiment. Results: Experimental results show that the average deviations of the experimental data before were 15. 34 m and 1. 44 m, before and after registration based on Mask R -CNN. This method can correctly register remote sensing images and vector data, and the proposed method has better application prospects in urban data registration, and has the characteristics of strong robustness and high accuracy. Conclusions: The proposed method in this paper can automatically register remote sensing images and vector data from different sources in areas such as mountains, grasslands, and deserts. In the next step, the amount of sample data of remote sensing images will be increased, including types of remote sensing images and types of features. [ABSTRACT FROM AUTHOR]

Objectives: In order to solve the problems of the singleness of activity metric construction in fusion rules and the subjectivity of parameter setting of pulse-coupled neural network (PCNN), a remote sensing image fusion method combining low-level visual features and parameter adaptive pulse coupled neural network (PAPCNN) in the non-subsampled shearlet transform (NSST) domain is proposed in this paper. Methods: First, the panchromatic image and the luminance component Y in YUV color space of multi⁃ spectral image are decomposed by NSST to obtain the high and low frequency components. Second, a fusion rule based on low-level visual features is used to low-frequency components fusion, and a new activity measure is constructed by combining three low-level features, namely, local phase congruency, local abrupt measure and local energy information. Then, PAPCNN model is used to high-frequency components fusion, and the multi-scale morphological gradient is used as the external input signal of the model. Finally, the fused image is obtained through NSST inverse transform and YUV inverse transform in turn. Results: The experimental results show that this method has better performance in remote sensing images of different platforms and different ground features. Compared with other 11 methods, this method has ab⁃ solute advantages in all evaluation indexes. Conclusions: The proposed method can better preserve the spatial and spectral information in the original image, thus it can provide a fused image with complementary ad⁃ vantages. [ABSTRACT FROM AUTHOR]

Objectives: Change detection is a process of recognizing the state changes of ground surface by multiple observations. With the improvement of image data quality, it provides more possibilities for people to realize change detection. Traditional multi- temporal remote sensing image change detection is easily af- fected by the season in which the images were taken, solar altitude angle and shooting angle etc. In addi-tion, the traditional calculation methods of change index, variables which susceptible to the interference of change object are often introduced, such as mean value, median value and so on. To solve such problems, we proposed a change detection method based on vector data. Methods: Firstly, since traditional change de- tection method is always influenced by shooting season, shooting angle, etc, we use a vector-image method, which is different from the previous image-image method. With the application of this method, the change differences between the old and new images were calculated by using a similarity measurement method, and the change object will identified by a threshold value. The vector-image method using the vector data and the new time image data detect the changes. Secondly, the anomaly detection method in the data mining method can be introduced into the change detection method, hence the outlier has a fewer and different characteristics compared with the normal object. In this paper, we firstly got the object through incremental segmentation under the constraints of the previous vector image, then extracted its texture and spectral fea- tures to get the dataset by the principal component analysis transform. After that, using isolation forest method to calculate object's change index, we obtained the change threshold by Bayes method. Results: We took two experiments, and the effectiveness of the proposed method was verified by comparing image-image and vector-image change detection methods, as well as Markov distance and isolation forest change methods, among which, the accuracy rate of experiment 1 is 0.923 5 and that of experiment 2 is 0.931 8. Conclu- sions: By comparing the image - image method with the vector - image method, the experimental results show that the proposed method can not only improve the accuracy of change detection, but also improve the automation and intelligence of image-based change detection. [ABSTRACT FROM AUTHOR]

Remotely sensed images are often degraded due to the haze interference during the imaging pro- cess, which greatly reduces their utilization. In order to solve this problem, a spatial-spectral adaptive haze removal method for single remote sensing image is proposed in this paper. Based on the dark prior theory and haze image model, and taking into account the spatial and spectral varieties in the remotely sensed images, our proposed method effectively overcome the difficulties of over-correction on bright terrain and inadequate correction of haze in some wavebands．A bright object index (BOI) is constructed to extract the bright objects with the help of the density segmentation method, and an adaptive correction function is then introduced to refine the misestimated transmittance. Given the influences of atmospheric scattering are wavelength dependent among visible channels, two empirical accommodation coefficients are applied to de- rive the transmittance of the different channels, achieving the adaptive adjustment of processing intensity in different wavebands. Experimental results show that our proposed method can remove the haze completely and yield visually haze-free images, comparing with the other existing methods. [ABSTRACT FROM AUTHOR]

BM3D (block matching and 3D filtering) method is an effective Gaussian noise removing method, but it shows limitation when removing Gaussian and impulse mixed noise in remote sensing images. The integrated method is a more effective way to remove mixed noise. In order to make full use of the advantages of BM3D method, an improved integration method (integrated BM3D method) has been developed. And a modified method M-Liu method is raised and used for estimating the priori noise level as an input parameter for the integrated BM3D method. According to algorithm validation, this paper draws the conclusion that the integrated BM3D method? taking the removal of image noise and the retention of detail signals into account, has good denoising performance. The method can be used to provide a new method for image denoising, and it is useful for the later application of remote sensing image. [ABSTRACT FROM AUTHOR]

This paper presents a new method that combines PCA with HSI to differentiate and extract construction shadows from high resolution remote sensing images and then calculates each block floor area ratio by using regression analysis. The proposed method reduces the complexity and instability when extracting floor area ratios using the traditional method. Taking downtown Fuzhou as an example, we extracted the construction shadow and floor area ratio by the new method, to compare and validate it against the traditional method. The results show that the shadow index from the new method is more reliable and precise than results obtained from the traditional approach. [ABSTRACT FROM AUTHOR]