DroneTechVision — AI-Powered Airspace Awareness

Origin

Built on proven airborne detection.

DroneTechVision began as the SkyGuard project — an open-source raptor alert system protecting backyard poultry. That same AI pipeline — real-time object segmentation, species classification, and instant alerting — now targets a far larger threat.

The original SkyGuard system (now DroneTechVision) was developed as a Texas A&M capstone project to solve a real problem: 49% of chicken owners cite aerial predators as their top concern, yet no affordable detection solution existed.

We built a full-stack detection platform: YOLO-based segmentation, custom species classification via transfer learning on 53,000+ images, edge deployment on Raspberry Pi, and real-time alerting — all for under $150 per node.

That foundation — detecting small, fast-moving airborne objects against variable sky conditions on constrained hardware — is precisely the problem set that drone detection demands. The pivot is natural. The technology is ready.

Real-Time Object Detection

YOLO v11 segmentation at 25 FPS on edge hardware

Species Classification

Custom CNN via transfer learning — 400+ species, 85%+ top-1 accuracy

Sub-2-second response — audio, SMS, email, push

Instant Alert Pipeline

Edge-First Architecture

Full operation on $150 Raspberry Pi 5 node

Field Validated

Multi-site pilot deployment in operational farm environments

Capability Roadmap

Four phases to full threat awareness.

Each phase builds on proven SkyGuard-derived technology, extending detection into classification, reporting, and threat assessment.

Phase 1 Detection

Distinguish drones from birds, aircraft, and environmental clutter in real-time using edge-deployed computer vision. Leverage existing YOLO segmentation pipeline retrained on drone imagery datasets.

Binary drone/non-drone classification

Sub-2-second detection latency

Multi-weather, multi-lighting robustness

False positive suppression via temporal filtering

Edge inference on Raspberry Pi 5 / Jetson

Tiling (SAHI) for small-object detection

Phase 2 Classification

Determine the nature and type of detected drones — categorize by platform class, size profile, and flight characteristics using a multi-model classification pipeline mirroring the proven species-identification architecture.

Platform type identification (multirotor, fixed-wing, hybrid)

Size and altitude estimation via visual geometry

Manufacturer/model fingerprinting where possible

Behavioral signature analysis (hover, transit, orbit)

Transfer learning from drone image datasets

Confidence scoring with multi-class probability

III

Phase 3 Reporting & Tracking

Transmit detection metadata, flight vector, and classification data to a central command and control system. Enable multi-node sensor fusion for coordinated airspace awareness across distributed sites.

REST API for centralized data ingestion

Flight vector estimation (heading, speed, altitude)

Multi-node sensor fusion and correlation

Geospatial event mapping and track history

Real-time dashboard with detection feeds

Encrypted, authenticated data transmission

Phase 4 Threat Assessment

Classify threat probability based on drone type, behavior pattern, proximity to protected assets, and contextual intelligence. Provide actionable threat levels to enable proportional response decisions.

Multi-factor threat scoring algorithm

Behavioral anomaly detection (loitering, approach vectors)

Geofence violation alerting

Tiered threat levels (LOW / MEDIUM / HIGH / CRITICAL)

Integration hooks for countermeasure systems

Operator decision-support dashboards

Technology

Edge-native. AI-first. Field-proven.

Every component was selected for reliability, cost, and deployability in austere environments.

Computer Vision

YOLOv11 real-time object detection with custom-trained classification models. SAHI tiling for small-object detection at distance.

Transfer Learning

Purpose-built CNNs trained on 53K+ images. Architecture supports rapid retraining on drone-specific datasets with minimal compute.

Edge Compute

Raspberry Pi 5 and Jetson Nano deployment. Full detection pipeline at ~$150/node with quantized models for real-time inference.

API

REST Architecture

Flask-based web portal with REST API for monitoring, configuration, and centralized data reporting across distributed nodes.

Event Storage

SQLite metadata with configurable retention. Annotated detection images, track histories, and audit-ready event logs.

Open Architecture

Python/PyTorch stack. Modular, extensible codebase at 121K+ lines. Designed for integration with existing C2 infrastructure.

Leadership

Built by engineers. Proven in the field.

John Daughtridge

Co-Founder & Technical Lead

M.Eng. Technical Management candidate at Texas A&M University. Designed, built, and field-deployed the SkyGuard detection platform (now DroneTechVision) — 121K+ lines of code, custom-trained AI models, and multi-site pilot operations. Bridging applied AI research with practical edge deployment.

→ LinkedIn Profile

Computer Vision Edge AI YOLOv11 PyTorch Systems Engineering Texas A&M

Joseph McCoy

Co-Founder & AI Strategy Lead

Microsoft Modern Workplace Architect with deep expertise in cloud infrastructure, AI-driven enterprise solutions, and large-scale technology deployment. Texas A&M graduate bringing years of experience managing complex technical projects from design through delivery — ensuring DroneTechVision scales from prototype to production.