Projects | Nupur Nimbekar

Buoy Detection using Gaussian Mixture Models

Implementation of Gaussian Mixture Model (GMM) for detecting and segmenting buoys in underwater imagery. Uses probabilistic modeling to identify colored buoys in challenging aquatic environments with varying lighting conditions and water clarity. Demonstrates application of statistical learning for real-world robotic perception tasks.

                                    Method:
                                    GMM
                                

                                    Application:
                                    Object Detection
                                

                                    Domain:
                                    Underwater Vision
                                

Python GMM Computer Vision OpenCV Statistical Learning

Weighted Static Consensus Algorithm

Implementation of weighted static consensus algorithm for distributed systems and multi-agent coordination. Explores consensus theory where agents reach agreement through iterative weighted averaging of neighboring states. Applicable to robot swarm coordination, sensor network fusion, and distributed decision-making systems.

                                    Domain:
                                    Multi-Agent Systems
                                

                                    Method:
                                    Consensus Theory
                                

Python Distributed Systems Multi-Agent Graph Theory

AR Tag Detection and Pose Estimation

Computer vision system for detecting and decoding AR (Augmented Reality) tags from camera images. Implements homography-based pose estimation to determine 3D position and orientation of tags relative to the camera. Essential for robot localization, augmented reality applications, and camera calibration.

                                    Capability:
                                    6-DOF Pose
                                

                                    Application:
                                    Robot Localization
                                

                                    Method:
                                    Homography
                                

Python OpenCV AR Tags Pose Estimation Computer Vision

A* Path Planning for Rigid Robots

Implementation of A* algorithm for optimal path planning of rigid-body robots in 2D environments with obstacles. Incorporates robot geometry (radius and clearance) for collision-free path generation. Visualizes explored nodes, unexplored regions, and final optimal path using OpenCV. Demonstrates efficient search with heuristic-guided exploration for robot navigation.

                                    Algorithm:
                                    A*
                                

                                    Robot Type:
                                    Rigid Body
                                

                                    Environment:
                                    2D with Obstacles
                                

Python A* Algorithm Path Planning OpenCV Robotics

RANSAC Implementation for Robust Fitting

Implementation of RANSAC (Random Sample Consensus) algorithm for robust model fitting in presence of outliers. Demonstrates application to line fitting, plane detection in point clouds, and homography estimation. Critical technique for robust perception in robotics where sensor data contains noise and outliers. Used in SLAM, visual odometry, and 3D reconstruction pipelines.

                                    Method:
                                    RANSAC
                                

                                    Robustness:
                                    Outlier Rejection
                                

                                    Applications:
                                    Model Fitting
                                

Python RANSAC Computer Vision Robust Estimation Point Cloud