Using convolutional neural network to identify irregular segmentation objects from very high-resolution remote sensing imagery

Abstract

, "Using convolutional neural network to identify irregular segmentation objects from very high-resolution remote sensing imagery," J. Appl. Remote Sens. 12(2), 025010 (2018), doi: 10.1117/1.JRS.12.025010. Abstract. Convolutional neural network (CNN) has shown great success in computer vision tasks, but their application in land-use type classifications within the context of object-based image analysis has been rarely explored, especially in terms of the identification of irregular segmentation objects. Thus, a blocks-based object-based image classification (BOBIC) method was proposed to carry out end-to-end classification for segmentation objects using CNN. Specifically, BOBIC takes advantage of CNN to automatically extract complex features from the original image data, thereby avoiding the uncertainty caused by the manual extraction of features in OBIC. Additionally, OBIC compensates for the shortcomings of CNN whereby it is difficult to delineate a clear right boundary for ground objects at the pixel level. Using three high-resolution test images, the proposed BOBIC was compared with support vector machine (SVM) and random forest (RF) classifiers, and then, the effect of image blocks and mixed objects on classification accuracy was evaluated for the proposed BOBIC. Compared with conventional SVM and RF classifiers, the inclusion of CNN improved the OBIC classification performance substantially (5% to 10% increases in overall accuracy), and it also alleviated the effect derived from mixed objects.

Land Cover Maps Production with High Resolution Satellite Image Time Series and Convolutional Neural Networks: Adaptations and Limits for Operational Systems

“…In [20], a modified U-Net with residual connections and a guided filter were used in post-processing to straighten building prediction mask edges. Simpler methods have also been proposed, where CNN predictions are refined with object-based image analysis approaches [21]. Although these approaches improve the geometric accuracy of segmentation CNNs, they rely on the availability of dense annotations in order to learn how to predict edges or to regularize the segmentation.…”

Improved Winter Wheat Spatial Distribution Extraction from High-Resolution Remote Sensing Imagery Using Semantic Features and Statistical Analysis

“…Researchers used recurrent neural networks [62] and convolution [63] to improve the calculation efficiency. Reference [65] comprehensively used the pixel spatial distance information and category information as constraints for network training to improve the accuracy of image segmentation results.…”

“…Object-level information is an information category commonly used in post-processing methods; it includes object shape information [65] and position information [65,66]. Using object-level information to post-process the CNN segmentation results can improve the fineness of the edges.…”

“…Using object-level information to post-process the CNN segmentation results can improve the fineness of the edges. Multiresolution segmentation algorithms [67] and patch-based learning [65,68] have been used to successfully generate image object information. Classifiers are equally important; using more powerful classifiers such as decision trees, the results obtained are better than those obtained by simple linear classifiers [69].…”

Improved Winter Wheat Spatial Distribution Extraction Using A Convolutional Neural Network and Partly Connected Conditional Random Field

“…In addition, differences in crop growth within the planting area adversely affect the spectral feature extraction, thereby resulting in mis-segmented pixels that form the so-called "salt and pepper" phenomenon [23,24].As the spectral features only express the characteristic information of the pixels themselves, the effect is usually not ideal when applied to higher-spatial-resolution images [23]. There is more detailed information in higher-spatial-resolution remote sensing images, and the spatial correlation between pixels is significantly enhanced, but the spectral characteristics cannot express this correlation information, and therefore, in such cases, spectral features are ineffective [25,26]. To better express the spatial correlation information between pixels, previous studies have proposed a series of texture feature extraction methods, such as the wavelet transform [27,28], Gabor filter [29,30], and gray level co-occurrence matrix (GLCM) [31].…”