A5014887553- Remote-Sensing Image Classification
- Advanced Image and Video Retrieval Techniques
- Remote Sensing and Land Use
- Advanced Image Fusion Techniques
- Advanced Neural Network Applications
- Sparse and Compressive Sensing Techniques
- Advanced Chemical Sensor Technologies
- Image and Object Detection Techniques
- Image Processing Techniques and Applications
- Infrared Target Detection Methodologies
- Image and Signal Denoising Methods
- Advanced SAR Imaging Techniques
- Synthetic Aperture Radar (SAR) Applications and Techniques
Xidian University
2022-2023
Deep learning has attracted increasing attention across a number of disciplines in recent years. In the field remote sensing, ship detection based on deep for synthetic aperture radar (SAR) imagery is replacing traditional methods as mainstream research method. The multiple scales objects make targets challenging task SAR images. This paper proposes new methodology better multi-scale images, which YOLOv5 with small model size (YOLOv5s), namely network (MSSDNet). We construct two modules...
Hyperspectral anomaly detection (HAD) is a challenging task in hyperspectral image processing, which to capture the by spectral and spatial information without prior knowledge. Recently, some isolation forest (IF) methods HAD are proposed achieve good accuracy. However, these build trees global pixels single band partition, way limits utilization of spectral-spatial information, resulting suffering from poor performance detecting hard anomalies. To this end, novel two-stream based on deep...
Sparse representation (SR)-based approaches and collaborative (CR)-based methods are proved to be effective detect the anomalies in a hyperspectral image (HSI). Nevertheless, existing for achieving anomaly detection (HAD) generally only consider one of them, failing comprehensively exploit them further promote performance. To address issue, novel HAD method, which integrates both SR CR, is proposed this article. specific, an model, whose overcomplete dictionary generated by means...
The low rank and sparse representation (LRSR) technique has attracted increasing attention for hyperspectral anomaly detection (HAD). Although a large quantity of research based on LRSR HAD is proposed, the performance still limited, due to unsatisfactory dictionary construction insufficient consideration global local characteristics. To tackle above-mentioned concern, novel method, termed dual collaborative constraints regularized low-rank via robust dictionaries construction, proposed in...
At present, in arbitrary-oriented object detection, the angular periodicity problem of rotated bounding box described by angle causes an to have different numerical representations, which leads uncertainty regression. To eliminate problem, this paper, we propose a novel and simple ellipse parameters representation method for object, hides focal vector avoid direct prediction. Moreover, proposed can enable only one representation, is beneficial alleviate In order adapt method, adopt 2D...
The imaging process of real-world images is inevitably polluted by noise, which affects the visual quality and subsequent processing images. How to restore image details while removing noise has always been a challenging problem. existing complementary learning strategies combine advantages both denoised have good effects. However, these methods that are based on single generative adversarial network (GAN) suffer from complex structure, difficulty in training, further improvement. Therefore,...
Recently, the isolation forest (IF) methods have received increasing attention for their promising performance in hyperspectral anomaly detection (HAD). However, limited by ability of exploiting spatial-spectral information, existing IF-based suffer from a lot false alarms and disappointing detecting local anomalies. To overcome two problems, multiscale superpixel guided discriminative method is proposed HAD. First, segmentation employed to generate some homogeneous regions, it can...
ABSTRACTThe objects in remote sensing images usually have complex background and appear anywhere any direction, which poses challenges for deep learning-based object detection. In order to solve the above problems, we embed local binary pattern (LBP) into detection facilitate performance improvements. Specifically, implement thep LBP via depthwise separable convolution without learnable parameters, makes highlight object-related regions of feature map instead original images. The...
Arbitrarily oriented object detection is one of the most-popular research fields in remote sensing image processing. In this paper, we propose an approach to predict angles indirectly, thereby avoiding issues related angular periodicity and boundary discontinuity. Our method involves representing long edge angle as a vector, which then decompose into horizontal vertical components. By predicting two components can obtain information indirectly. To facilitate transformation between...