Method and Algorithm for Recognition Dynamic Objects from a Mobile Platform, from Images Obtained in Different Spectral Ranges and Lidar Data

Abstract

Purpose of research. In modern automatic information collection systems, autonomous mobile devices are increasingly used, data from which can be obtained in conditions hazardous to human health, from geographically remote places, in difficult meteorological conditions and in round-the-clock observation mode. For the autonomous operation of such devices, it is necessary to use methods and algorithms that allow you to build a map of the area, link a mobile platform to it, determine a route to a target point, highlight obstacles along the route and correct the route taking into account detected obstacles. Methods. The article proposes a method and an algorithm for the selection of dynamic objects from a mobile platform, based on the analysis of data obtained from a multispectral camera, which allows the selection of obstacles, such as water, plant origin, technogenic nature, etc. with reduced computational complexity. To improve the accuracy of determining the coordinates of detected objects, a laser rangefinder is used. Results. We consider the well-known methods of multispectral images recognition and present their comparative analysis. A method and an algorithm for recognition dynamic objects from a mobile platform, from images obtained in different spectral ranges and lidar data are proposed. Experimental studies were carried out to confirm the adequacy of the mathematical substantiation of the method, to reduce the error in calculating the coordinates of the object, at a distance of up to 100 meters to the object, RMSE - 0.447%, MAPE - 0.397, to increase the performance, it took 0, 04 seconds to select the object and determine its coordinates. Conclusion. The article analyzes modern methods for recognizing multispectral images, presents the principles on which each method is based, gives advantages and disadvantages. The authors have developed a method and an algorithm that make it possible to identify static and dynamic obstacles along the route of a mobile platform, based on a sequence of images obtained in different spectral ranges. In the course of experimental studies, the performance of the proposed solutions and compliance with the specified requirements for accuracy and reliability were confirmed.