Multi-class object detection in remote sensing image based on context information and regularized convolutional network

Abstract A U-Net convolutional network structure is fully capable of completing the end-to-end training with extremely little data, and can achieve better results. When the convolutional network has a short link between a near input layer and a near output layer, it can implement training in a deeper, more accurate and effective way. This paper mainly proposes a high-resolution remote sensing image change detection algorithm based on dense convolutional channel attention mechanism. The detection algorithm uses U-Net network module as the basic network to extract features, combines Dense-Net dense module to enhance U-Net, and introduces dense convolution channel attention mechanism into the basic convolution unit to highlight important features, thus completing semantic segmentation of dense convolutional remote sensing images. Simulation results have verified the effectiveness and robustness of this study.

Download Full-text

A Segmentation Model for Extracting Farmland and Woodland from Remote Sensing Image

10.20944/preprints201712.0192.v1 ◽

2017 ◽

Author(s):

Chengming Zhang ◽

Shujing Wan ◽

Shuai Gao ◽

Fan Yu ◽

Qingdi Wei ◽

...

Keyword(s):

Remote Sensing ◽

Remote Sensing Image ◽

Remote Sensing Images ◽

Convolutional Network ◽

Coverage Area ◽

Plant Coverage ◽

Training Stage ◽

Convolution Kernels ◽

Neural Network Structure ◽

Segmentation Of Images

It is very difficult to accurately divide farmland and woodland in Gaofen 2 (GF-2) remote sensing image, because their single plant coverage is very small, and their spectra are very similar. The ratio of spatial resolution and one plant’s coverage area must be fully taken into account when designing the Convolutional Neural Network structure for extracting them from GF-2 image. We establish a Convolutional Encode Neural Networks model (CENN), The first layer has two sets of convolution kernels to learn the characteristics of farmland and woodland respectively, while the second layer is the encoder to encode the characteristics by transfer function, which can map the results to the corresponding category number. In the training stage, samples of farmland, woodland, and other categories are categorically used to train CENN, as soon as training is accomplished, CENN would acquire enough ability to accurately extract farmland and woodland from remote sensing images. The final extraction result is obtained by implementing per-pixel segmentation of images used to train the CENN. CENN is compared and analyzed with others such as Deep Belief Network (DBN), Full Convolutional Network (FCN), Deeplab Model. The results of experiments show that CENN can more accurately mine the characteristics of farmland and woodland, and it achieves its goal of extracting farmland and woodland with high precision from GF-2 images.

Download Full-text

A Review of Remote Sensing Image Object Detection Algorithms Based on Deep Learning

2020 IEEE 5th International Conference on Image, Vision and Computing (ICIVC) ◽

10.1109/icivc50857.2020.9177453 ◽

2020 ◽

Author(s):

Zhe Zheng ◽

Lin Lei ◽

Hao Sun ◽

Gangyao Kuang

Keyword(s):

Remote Sensing ◽

Deep Learning ◽

Object Detection ◽

Remote Sensing Image ◽

Detection Algorithms ◽

Image Object Detection ◽

Image Object

Download Full-text

Robust Building Extraction for High Spatial Resolution Remote Sensing Images with Self-Attention Network

Sensors ◽

10.3390/s20247241 ◽

2020 ◽

Vol 20 (24) ◽

pp. 7241

Author(s):

Dengji Zhou ◽

Guizhou Wang ◽

Guojin He ◽

Tengfei Long ◽

Ranyu Yin ◽

...

Keyword(s):

Remote Sensing ◽

Spatial Resolution ◽

High Spatial Resolution ◽

Remote Sensing Image ◽

Context Information ◽

Building Extraction ◽

Remote Sensing Images ◽

Global Features ◽

Long Distance ◽

Attention Network

Building extraction from high spatial resolution remote sensing images is a hot spot in the field of remote sensing applications and computer vision. This paper presents a semantic segmentation model, which is a supervised method, named Pyramid Self-Attention Network (PISANet). Its structure is simple, because it contains only two parts: one is the backbone of the network, which is used to learn the local features (short distance context information around the pixel) of buildings from the image; the other part is the pyramid self-attention module, which is used to obtain the global features (long distance context information with other pixels in the image) and the comprehensive features (includes color, texture, geometric and high-level semantic feature) of the building. The network is an end-to-end approach. In the training stage, the input is the remote sensing image and corresponding label, and the output is probability map (the probability that each pixel is or is not building). In the prediction stage, the input is the remote sensing image, and the output is the extraction result of the building. The complexity of the network structure was reduced so that it is easy to implement. The proposed PISANet was tested on two datasets. The result shows that the overall accuracy reached 94.50 and 96.15%, the intersection-over-union reached 77.45 and 87.97%, and F1 index reached 87.27 and 93.55%, respectively. In experiments on different datasets, PISANet obtained high overall accuracy, low error rate and improved integrity of individual buildings.

Download Full-text