Dimitrios Chatziparaschis

I'm a PhD student at the ECE Department of the University of California, Riverside, under supervision of Prof. Konstantinos Karydis. Before that, I completed my MSc in Computer Science and Engineering and my Diploma in Electrical and Computer Engineering at the Technical University of Crete, where I was honored to work with Professors Panagiotis Partsinevelos and Michail G. Lagoudakis.

I would be happy to connect! Feel free to reach out via email or LinkedIn.

My research spans computer vision, machine learning, and robotics, toward real-time, perception-driven robotic autonomy in unstructured and dynamic outdoor environments. I focus in particular on multi-modal 3D detection in the field, state estimation, and learning-based monocular depth estimation.

Online annotation-to-detection framework for multi-modal detection and mapping in the field, coupling early-fusion cross-modal annotation with stage-incremental refinement, starting from minimal training labels.

Dense GICP scan-to-map matching with an adaptive mapping filter, rejecting motion-distorted and dynamic points, for high accuracy odometry in unstructured agricultural fields.

Online 6-DoF reaching and dynamic retargeting using Dynamic Neural Fields (DNFs) with Stochastic Optimal Control (SOC) based on visual feedback, with Kinova Gen3 Lite.

Real-time tree detection and global landmark association, with NDVI–LiDAR fusion, entropy-based candidate matching, and Kalman filter-based per-tree geometry and vegetation trait estimation.

Fully autonomous robotic platform for Stem Water Potential (SWP) in avocado fields, based on Gaussian Processes (GPs) for modeling sampling area uncertainty and onboard leaf detection to perform 6D pose estimation and leaf cutting.

Lightweight keypoint detector based on a centroid point distance criterion coupled with a multi-modal Non-Maximum Suppression (NMS) algorithm, to obtain salient and repeatable points in both colored and uncolored 3D point clouds.

Development of a semi-autonomous ground robotic platform for georeferenced soil ECa measurements, evaluated through sensor integration and terrain assessments, achieving >90% Pearson correlation with manual measurements.

Development of a real-time UAV surveying framework with a gimballed monocular camera and laser rangefinder for ground-target localization. Features a spatial filtering module enhancing multilateration through improved convergence and accuracy.

Development of a fully autonomous UAV framework for Search and Rescue, employing DDPG-based DRL for control and decision-making in a custom OpenAI Gym environment with an autoencoder for semantic segmentation.

UAV-assisted positioning system for GNSS-denied environments, leveraging clear-sky GNSS visibility from the UAV to support ground-based surveys. Theoretical evaluation of positioning error demonstrates the feasibility of the approach.

Aerial–ground robotic collaboration for Search and Rescue: (1) sensing-blind humanoid localization via a UAV using onboard stereo depth estimation, occupancy-grid mapping, and AMCL; and (2) navigation toward humans in need using a custom YOLO model and an R*-based footstep planner.

Credits to Jon Barron's amazing website.