Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks

By Martin Längkvist, Andrey Kiselev, Marjan Alirezaie, Amy Loutfi. 2016

313 citations recorded by the source corpus.

The availability of high-resolution remote sensing (HRRS) data has opened up the possibility for new interesting applications, such as per-pixel classification of individual objects in greater detail. This paper shows how a convolutional neural network (CNN) can be applied to multispectral orthoimagery and a digital surface model (DSM) of a small city for a full, fast and accurate per-pixel classification. The predicted low-level pixel classes are then used to improve the high-level segmentation. Various design choices of the CNN architecture are evaluated and analyzed. The investigated land area is fully manually labeled into five categories (vegetation, ground, roads, buildings and water), and the classification accuracy is compared to other per-pixel classification works on other land areas that have a similar choice of categories. The results of the full classification and segmentation on selected segments of the map show that CNNs are a viable tool for solving both the segmentation and object recognition task for remote sensing data.

Subfields: Environmental Engineering, Media Technology, Ocean Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Environmental Engineering

In Media Technology

In Ocean Engineering

Citations in the corpus

Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community (2017, methodology): “Längkvist et al.(176) applied a CNN to orthorectified multispectral imagery (MSI) and a digital surface model of a small city for a full, fast, and accurate per-pixel classification.”
A review of deep learning methods for semantic segmentation of remote sensing imagery (2021, methodology): “L ¨ angkvist et al., 2016 smoothed the segmentation map using the pre-generated superpixels from SLIC.”
ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data (2019, background): “, building, road, or certain vegetation types [30, 11, 58], or multiple classes all together [44, 28, 33, 37].”
Classification With an Edge: Improving Semantic Image Segmentation with Boundary Detection (2016, background): “On the contrary, the segnet encoder-decoder network in conjunction with a hed style boundary extractor matches an ensemble of VGG-style models, although the latter is a lot heavier in memory usage and training time.”
Learning Ground Traversability From Simulations (2017, background): “vector machines or random forests [8]), whereas others adopt deep learning techniques such as CNNs [1], [9], which operate on raw image data and learn meaningful problem-specific visual features.”
Evaluation of Different Machine Learning Methods and Deep-Learning Convolutional Neural Networks for Landslide Detection (2019, methodology): “CNNs have also been used for VHR image classification and segmentation [32,33], semantic segmentation [34], scene annotation using different high spatial resolution and aerial images [35], and object detection [36].”
End-to-End Change Detection for High Resolution Satellite Images Using Improved UNet++ (2019, methodology): “Based on a standard CNN, a patch-based CNN approach [59] was proposed, where the label of each pixel is generated by the patch enclosing it.”
MARE: Self-Supervised Multi-Attention REsu-Net for Semantic Segmentation in Remote Sensing (2021, background): “It is unanimously recognized that DL methods have high generalization capabilities, succeeding in extracting robust and efficient features [8,9].”
Classification for High Resolution Remote Sensing Imagery Using a Fully Convolutional Network (2017, methodology): “The most typical one is the patch-based CNN approach [27,28].”
Understanding Deep Learning Techniques for Image Segmentation (2019, background): “Segmentation algorithms have significant usage in satellite images for various geo-statistical analysis [109].”