Ros package for converting 3D voxel maps generated by the UFOMap and the OctoMap mapping solution into 2D occupancy maps for unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGV)

Unmanned Aerial Vehicle (UAV) has created much significance in recent times. It is the booming technology which are mostly used in military applications, their use is rapidly expanding to commercial, scientific, agricultural and other applications. In this research, we proposed a model to estimate the density of crowd from aerial view. Our model used Gaussian Kernels to create the heat map (density map), which is used to estimate crowd in an image. We had deployed our model on raspberry pi to estimate crowd in real time from aerial view which has significant accuracy.

One of the newer application areas is autonomous vehicles, which include submersibles, land-based vehicles (small robots with wheels, cars or trucks), aerial vehicles, and unmanned aerial vehicles (UAV). The level of autonomy ranges from fully autonomous (unmanned) vehicles to vehicles where computer-vision-based systems support a driver or a pilot in various situations. Fully autonomous vehicles typically use computer vision for navigation, e.g. for knowing where it is, or for producing a map of its environment (SLAM) and for detecting obstacles. It can also be used for detecting certain task specific events, e.g., a UAV looking for forest fires. Examples of supporting systems are obstacle warning systems in cars, and systems for autonomous landing of aircraft. Several car manufacturers have demonstrated systems for autonomous driving of cars, but this technology has still not reached a level where it can be put on the market. There are ample examples of military autonomous vehicles ranging from advanced missiles to UAVs for recon missions or missile guidance. Space exploration is already being made with autonomous vehicles using computer vision, e.g., NASA's Curiosity and CNSA's Yutu-2 rover.

Aerial imaging is mainly done with small planes from several kilometers of altitude which involves high cost and manual support. An unmanned micro aerial vehicle with a camera mounted on it is an alternative airborne platform. This type of system allows fast aerial imaging of small areas with a higher level of detail and lower cost. A typical MAV with a bottom and frontal camera which can be communicated with the computer over WIFI opens up an opportunity to sense the environment and take decisions with the help of computer vision algorithms. It can be used for updates of geographical maps after natural phenomena, e.g. update of routes in the areas affected by floods. The two cameras help to reconstruct an unknown 3D environment of which the drone navigates. In this project, we used AR drone as a MAV and ROS as a platform to program it. We have developed a new package named ‘rkv_ardrone’ to traverse a specific area and capture images. The images are then stitched together to give a composite and high detailed view of the region.

Aerial Vehicle Mapper is an AI-powered tool that detects and maps vehicles from aerial images using the Detectron2 deep learning framework. It enables geospatial mapping of detected vehicles for applications in urban planning, surveillance, disaster management, and more.

YOLOv8 + SAHI aerial vehicle detection for Sri Lankan urban traffic | 54.4% mAP@0.5 | INT8 quantization | FastAPI + Docker

Real-time drone object detection & GPS georeferencing system using YOLOv8. Maps vehicles & pedestrians from aerial footage onto interactive maps with pixel-to-GPS transformation.

I created code and resources for performing drone mapping, a technique used to create detailed maps and 3D models of landscapes, infrastructure, and other areas of interest using unmanned aerial vehicles (UAVs).

AI powered solution that analyses aerial imagery to map affected areas, people, vehicles, detect terrain changes and generate 3D models for rescue planning and prioritisation. Provides real-time insights for emergency response teams

DroneCommand is an interactive control station for managing unmanned aerial vehicles (UAVs), featuring real-time map visualization, flight control, battery monitoring, no-fly zone management, and collision detection for seamless and safe drone operations.

DroneCommand3D is 3-dimensional implementation of the interactive control station for managing unmanned aerial vehicles (UAVs), featuring real-time map visualization, flight control, battery monitoring and collision detection for seamless and safe drone operations.

this is a recruitment task given to me by Unmanned Aerial Vehicle Society (UAS - DTU). a fire has broken out in a civilian area and our UAV is gathering images to map the location of houses and buildings in the affected zone

This includes the depth map, depth analysis, grayscale images of traffic video. The video contains an aerial view. I have created 3D boxes around each image and have labelled them with depth, speed and angular positions. The objects/ vehicles are being detected using coco dataset. Yolo v5, MiDAS, cv2 have been used here