Developed a real-time stereo vision system for accurate depth estimation and 3D reconstruction, addressing challenges in autonomous navigation and robotics applications.
Engineered a comprehensive stereo vision pipeline utilizing block matching algorithms (StereoSBM and StereoSGBM) for disparity mapping. The system processes synchronized stereo camera feeds to generate precise depth maps in real-time.
Noise Reduction: Low-cost cameras introduced significant noise affecting algorithm accuracy. Implemented advanced filtering techniques and iterative calibration processes to achieve sub-pixel accuracy alignment. Currently developing Kalman filter integration for enhanced noise suppression.
Successfully generated accurate depth maps with applications in autonomous vehicle navigation, drone obstacle avoidance, and real-world 3D scene reconstruction. The system demonstrates real-time processing capabilities suitable for embedded robotics applications.
I'm actively exploring a multi-lens, multi–focal length camera arrangement to replicate the perception system used in Tesla's vision-only self-driving stack. The goal is to combine overlapping fields of view at different focal lengths to improve long-range object detection while maintaining accurate short-range depth through stereo disparity. This includes synchronized capture, cross-camera calibration, and fusion strategies to align outputs across lenses with different intrinsics.
Designing and building a fully autonomous drone system from scratch, integrating custom flight controller hardware with advanced computer vision for autonomous navigation in complex environments.
Sensor Fusion: Implementing Kalman filtering algorithms to combine IMU, GPS, and barometric data while mitigating vibration noise and sensor drift. Developing custom noise models specific to rotorcraft dynamics.
Hardware Complete: Flight controller assembly finished, currently in PID tuning phase for initial flight testing. Software integration of vision system with flight control in progress.
Next Phase: System integration testing, flight envelope characterization, and autonomous mission validation.
Developed a smart garage parking assistant using sensors and real-time feedback to help drivers park safely and efficiently.