Algorithms and software for detecting and identifying small-sized objects on stereoscopic images of transport systems infrastructure
In the modern world, special attention is paid to the tasks of digital image processing, since their importance for practical application is increasing. Naturalistic driving datasets and data analytics is very important for decision making. This is especially important when drivers make decisions in a traffic situation. It is also necessary to consider and take into account various infrastructure and transport systems. To improve the quality of digital image processing, it is necessary to develop and improve algorithms for detecting and identifying small-sized objects on stereoscopic images and their software implementation. The authors emphasize the high accuracy of detection, localization and identification of an object in real time, even if the object is partially lost. The algorithm contains a procedure for calculating the geometric characteristics and the relative location of objects in space and allows you to determine their coordinates and characterize the appearance of the object. This is possible without increasing the set of sensors used. The authors solved the problem of developing and implementing algorithms for detecting and identifying small-sized objects on stereoscopic images. Algorithms and methods for constructing stereo images for a given set of object classes in transport systems have been developed.
3D surround view; ADAS; Augmented reality; Camera auto-calibration; Detectors; Panoramic view; Sensors; Transport systems infrastructure