We research solutions to enable intelligence, surveillance and reconnaissance by means of near-real-time target recognition, identification and classification. The cloud services, intelligence, surveillance, target acquisition and reconnaissance are of importance for the C4+iSTAR systems. These platforms are expected to ensure high-level cognitive autonomy to accomplish complex missions and tasks in rapidly-changing adverse environments. We research and apply deep learning concepts and algorithms to enable high-confidence awareness and advance situational analysis. We examine engineering solutions for object identification and classification. Our findings ensure sufficient level of fidelity on the object recognition and classification likelihood with high identification probability, processing latency on low-power ARM CPUs, and, integration capabilities. Advanced concepts on moving target recognition and object classification using fly data are researched with low-fidelity experimental substantiations. We modify the YOLOv3 object detection method to detect bounding boxes for arbitrary orientation angle, angle of view and corner shapes which are of importance in aerial applications. Our results demonstrate adequate detection capability while maintaining fast computational performance of the original YOLOv3 architecture. The proposed algorithms and computing schemes are supported by codes in C++.