The dissertation analyzed means and technologies of image collection using UAVs, existing systems of a similar purpose, methods and means of image processing of objects and their recognition. In the work, methods of collecting, recognizing and processing images obtained using UAVs for the detection of specified objects have been developed, which improve the efficiency of collection, accuracy of recognition and processing of images obtained using UAVs, as well as appropriate tools have been developed, experiments have been set up and carried out with the developed tools experimental studies.
The object of the study is the process of collecting, recognizing and processing images obtained using UAVs to detect given objects.
The subject of the study is methods and algorithms for ensuring the collection, recognition and processing of images obtained using UAVs for the detection of given objects.
The aim of the dissertation research is to improve the collection efficiency, accuracy of recognition and processing of images obtained with the use of UAVs for the detection of given objects.
The scientific novelty of the obtained results is as follows:
1) a new method of constructing UAVs routes according to self-learning technologies has been developed, which makes it possible to improve movement and synchronization between a group of drones or one drone within the working segment and due to this increase in the amount of processed data;
2) a new method of dynamically obtaining images of given structural objects in three-dimensional space using several drones has been developed, which improves the coordination between different drones and achieving the movement of the entire group of drones from the given starting points to the end points of the program mission autonomously;
3) a new method of synchronizing video streams in real time has been developed, which makes it possible to perform a comparison of the received current results with the past ones in real time, and this ensures the prompt obtaining of results and the detection of structural objects that were missed in the process of past program missions;
4) the method of detecting given structural objects in images has been improved, which, unlike the original YOLOv5 architecture, consists in the fact that the neural network focusing module has been modified, the convolutional layer combining the input feature map with the concatenation operation has been removed, the visual attention mechanism for feature extraction has been updated, which made it possible to improve the accuracy of detection and reduce the training time of the neural network.
The practical significance of the results obtained. The developed automated system is designed to detect and count the number of apples in an orchard in real time. The advantage of the developed system over analogues is that it receives multiple video frames in real time from the cameras of several UAVs and synchronizes these video frames with each other into one informational data structure, which will later be transformed into a continuous image. In addition, the use of image quality optimization functions allows for the most efficient detection of structural during UAV operational missions in the operational environment. The use of such a transformation tool enabled the system to receive a continuous flow of data to all subsequent software components of the automated system. As a result of the experimental studies, the effectiveness of the developed automated system was proven, which is confirmed by a high average value of 82.69% of the reliability indicator of detecting and calculating the number of fruit fruits and a low average level of errors I (14.67%) and II (18.33%) genus.
The theoretical and practical results of the research were implemented in LLC “UKS++” (Khmelnytskyi), SE “NOVATOR” (Khmelnytskyi), PE “NOLT TECHNOLOGY” (Khmelnytskyi), LLC “AGROTECHSERVIS” (Bohdanivtsi), and in the educational process of Khmelnytskyi National University in teaching disciplines at the Department of Computer Engineering and Information Systems for the specialty 126 Information Systems and Technologies, 123 Computer Engineering and the Department of Computer Science for the specialty 122 Computer Science, as well as in the implementation of the state budget theme of Khmelnytskyi National University “Development of information technology for making human-controlled critical and security decisions on mental and formal models”.
