A real-time semantic segmentation algorithm for aerial imagery

Abstract

We propose a novel effective algorithm for real-time semantic segmentation of images that has the best accuracy in its class. Based on a comparative analysis of preliminary segmentation methods, methods for calculating attributes from image segments, as well as various algorithms of machine learning, the most effective methods in terms of their accuracy and performance are identified. Based on the research results, a modular near real-time algorithm of semantic segmentation is constructed. Training and testing is performed on the ISPRS Vaihingen collection of aerial photos of the visible and IR ranges, to which a pixel map of the terrain heights is attached. An original method for obtaining a normalized nDSM for the original DSM is proposed.

Optimal tuning of the contour analysis method to recognize aircraft on remote sensing imagery

“…a) b) Figure 7. Total relative measurement error (15) dependence from the C-130 (a) and P-3 (b) vectorcontour's items quantity.…”

“…(a) the minimal items' quantity of the certain class's instances to minimize the type II error; (b) the maximal items' quantity of the certain class's instances to minimize the type I error; (c) the recognized object's items quantity to retain its actual level of details. The table 5 represents the total relative measurement error (15) per each aircraft class along with the mean value for all the heuristic methods and the proposed one. The differences in results of heuristic methods (a-c) for certain classes emphasize the training dataset's irregularity.…”

“…Moreover, CNNs require transfer learning techniques [11] to operate on images with spectral parameters different from the training ones (for example, created by the sensor of another type). To the contour analysis list of disadvantages, we should write a much shorter range of applications, as the recognition process is driven by the only feature -the object's shape, together with the strong addiction to the segmentation method's quality used to extract the object from an underlying surface [12][13][14][15].…”

Applying doubly stochastic filters to evaluate the dynamics of object sizes on satellite image sequences

“…Firstly, these algorithms are largely incapable of taking into account the multi-zonal nature of remote sensing (RS), so each satellite image contains the results of the Earth's surface registration in different spectral ranges. Some works [4][5][6] suggest the possibility of processing hyperspectral images. Thus, the authors based their theory on criterion of uniformity for reception of connected areas of such hyperspectral image [4], modification and generalization of algorithm K-means [5] and use of physical properties of a satellite data [6].…”

“…Some works [4][5][6] suggest the possibility of processing hyperspectral images. Thus, the authors based their theory on criterion of uniformity for reception of connected areas of such hyperspectral image [4], modification and generalization of algorithm K-means [5] and use of physical properties of a satellite data [6]. Secondly, the existing approaches unable to use data on the observed territory received at previous points of time for image segmentation.…”