IndustryBAT BangladeshWork period · 2025Manufacturing

BAT AI Leaf SOP Checker

Multiprocess RTSP analytics system that checks six leaf-buying SOP stages: bale opening, barcode scan, moisture inspection, layer-by-layer inspection, weighing step-back, and unwanted activities.

BAT AI Leaf SOP Checker project workflow and system architecture
Project workflow and system architecture overview. Open full-size diagram

Workflow at a glance

  1. Read camera feeds

    Ingest video from leaf-buying work areas.

  2. Detect workflow cues

    Identify observations relevant to SOP stages.

  3. Track SOP state

    Evaluate six workflow checks over time.

  4. Deliver evidence

    Store snapshots and send stage-level events.

Project facts & reported results

6
Workflow Checks

Implementation fact

Multiprocess RTSP pipelines
Architecture

Implementation fact

Evidence + API events
Output

Implementation fact

Project Overview

Problem Statement

Manual SOP review is difficult to apply consistently across repeated production workflows and provides limited searchable evidence for later review.

Approach & Methodology

Implemented multiprocess RTSP capture and threaded analysis for leaf-buying workflow checks. Detection and region logic update persistent stage state, while per-check senders deliver evidence and events.

Outcome & Scope

Produces structured, time-stamped compliance events and snapshots through independent processing and sender services while maintaining persistent workflow state.

Technical Challenges & Solutions

1

Workflow state across frames

SOP stages span many observations rather than a single detection.

Solution

Maintain task-specific persistent states and region-based transitions.

2

Keeping capture responsive

Slow analysis or delivery can interfere with video capture.

Solution

Separate capture, threaded processing, and per-check sending.

System Architecture

Independent video processing and persistent SOP event delivery.

Key Components:

RTSP readers

Acquire video through separate capture processes.

Task pipelines

Analyze configured workflow observations and regions.

Stage state

Maintain counters and workflow state across observations.

Evidence senders

Deliver snapshots and task events through independent senders.

Key Features & Capabilities

Bale-opening check

Barcode-scan workflow check

Moisture-inspection check

Layer-by-layer inspection check

Weighing step-back check

Unwanted-activity checks

Persistent stage state and per-check evidence delivery

Current Scope & Limitations

This case study describes SOP observation support, not a certification that every physical procedure was completed.

Synchronized cross-camera bale handoff is not claimed.

Technologies & Tools

Python
YOLO11
OpenCV
RTSP
Multiprocessing
State machines
Event delivery