Conveyor Counting, Parking Occupancy & Threat Analytics
Modular computer vision solutions for smart infrastructure: conveyor package directional counting with CentroidTracker, stall-level smart parking occupancy with polygon hysteresis, pyzbar catalog verification, and open-vocabulary fire/smoke detection.
Internal R&D / Personal Engineering Project at Bondstein Technologies Ltd.
Module of Bondstein Vision Suite; presented as a separate case study.
Workflow at a glance
Read facility video
Acquire imagery for the configured task.
Detect and track
Locate objects or occupied regions.
Apply module rules
Evaluate counts, occupancy, or event cues.
Review outputs
Inspect task-specific results and evidence.
Project facts & reported results
Implementation fact
Implementation fact
Implementation fact
Project Overview
Problem Statement
Facility teams often need several focused analytics—counting, parking occupancy, label validation, and fire or smoke screening—without separate, incompatible runtimes.
Approach & Methodology
Engineered lightweight Vision Suite modules sharing a unified runtime (`vision_core.py`). Implemented single-pass line-crossing package counting with CentroidTracker, per-stall polygon parking occupancy with hysteresis, pyzbar catalog verification, and open-vocabulary fire/smoke detection with zone gating.
Outcome & Scope
Implemented modular, config-driven R&D components that emit structured snapshots and events for dashboard integration.
System Architecture
Unified vision microservice engine for facility and logistics automation
Key Components:
YOLO + CentroidTracker directional line crossing deduplicator
Per-stall polygon containment with temporal hysteresis smoothing
pyzbar barcode decoder with catalog regex fuzzy matching
Open-vocab flame detection with temporal voting and zone gating
Key Features & Capabilities
Conveyor package directional line-crossing counting with single-pass track deduplication
Per-stall polygon parking occupancy detection with occupied/free hysteresis smoothing
Label & barcode verification with pyzbar and regex catalog validation
Open-vocabulary fire, smoke, and flame detection with temporal voting and zone gating
Periodic JSON snapshot & event telemetry for dashboard integration