Visual Calibration

The Visual Calibration filter is a lightweight video analysis tool designed for computer vision applications. It provides crucial insights about your hardware setup and environment, including camera stability analysis, movement detection, and video quality assessment. Supports multi-topic processing - analyze multiple video streams independently with separate analysis states for each topic.

How It Works

The Visual Calibration filter uses a multi-stage analysis pipeline with per-topic processing:

1. Multi-Topic Input Processing: Reads video frames from various sources (files, RTSP streams, etc.) with different topics
2. Per-Topic Camera Stability Analysis: Detects camera shake and movement patterns independently for each topic
3. Per-Topic Video Properties Calculation: Analyzes frame properties, quality metrics, and technical specifications for each topic
4. Per-Topic Movement Detection: Optional movement tracking and analysis (when enabled) with independent state
5. Multi-Stream Real-time Visualization: Provides live analysis overlays via Webvis for multiple streams
6. Upstream Data Forwarding: Passes through data from upstream filters while preserving topic structure
7. Non-Image Frame Support: Forwards non-image frames as-is without processing

Features

Multi-Topic Processing

• Independent Analysis: Each topic processed with separate analysis state
• Per-Topic State Management: Camera stability, movement tracking, and video properties tracked independently
• Topic Preservation: Output data maintains same topic structure as input
• Cross-Contamination Prevention: No interference between different topic analyses
• Scalable Processing: Handle any number of topics simultaneously

Camera Stability Analysis (Per-Topic)

• Shake Detection: Identifies camera shake and vibration patterns for each topic
• Movement Tracking: Monitors camera movement and stability per topic
• Stability Categorization: Classifies video stability levels per topic
• Threshold-based Detection: Configurable sensitivity settings per topic

Video Properties Analysis (Per-Topic)

• Frame Analysis: Calculates frame dimensions, quality metrics per topic
• Technical Specifications: Analyzes video encoding, bitrate, etc. per topic
• Quality Assessment: Evaluates video clarity and consistency per topic
• Performance Metrics: Tracks processing performance per topic

Movement Detection (Per-Topic, Optional)

• Optical Flow Analysis: Tracks movement patterns using computer vision per topic
• Movement Thresholds: Configurable sensitivity for movement detection per topic
• ROI-based Analysis: Focuses analysis on specific regions of interest per topic
• Real-time Tracking: Continuous movement monitoring per topic

Flexible Configuration

• Environment Variable Driven: No command-line arguments needed
• Dynamic Output Naming: Automatically generates descriptive output filenames
• Multiple Input Sources: Supports files, RTSP streams, and web URLs
• Customizable Thresholds: Adjustable sensitivity for all detection algorithms
• Upstream Data Forwarding: Configurable forwarding of data from upstream filters

Real-time Visualization

• Live Analysis Overlays: Real-time stability and movement indicators per topic
• Webvis Integration: Browser-based visualization interface with multi-stream support
• Interactive Controls: Adjustable visualization parameters
• Multi-Stream Support: Handle multiple video sources simultaneously with independent analysis
• Topic-based URLs: Access different analysis streams via separate URLs

Configuration Options

Core Analysis Settings

Parameter	Type	Default	Description
FILTER_CALCULATE_CAMERA_STABILITY	boolean	true	Enable camera stability analysis
FILTER_CALCULATE_VIDEO_PROPERTIES	boolean	true	Enable video properties calculation
FILTER_CALCULATE_MOVEMENT	boolean	true	Enable movement detection
FILTER_SHAKE_THRESHOLD	integer	5	Camera shake detection threshold (lower = more sensitive)
FILTER_MOVEMENT_THRESHOLD	float	1.0	Movement detection threshold (lower = more sensitive)
FILTER_ROI	list	[]	Region of interest for analysis [x, y, width, height]

Output and Visualization Settings

Parameter	Type	Default	Description
FILTER_FORWARD_UPSTREAM_DATA	boolean	true	Forward data from upstream filters
FILTER_SHOW_TEXT_OVERLAYS	boolean	true	Show analysis overlays on video
FILTER_LOG_INTERVAL	integer	3	Log analysis results every N frames

Input/Output Settings

Parameter	Type	Default	Description
VIDEO_INPUT	string	./data/sample-video.mp4	Input video file path
OUTPUT_VIDEO_PATH	string	./output/{input_name}_vizcal_output.mp4	Output video file path
OUTPUT_FPS	integer	30	Output video frames per second
WEBVIS_PORT	integer	8002	Port for Webvis visualization

Usage Examples

Basic Multi-Topic Analysis

# Analyze a video file with multi-topic processing (same video, different analysis)
python scripts/filter_usage.py

High Sensitivity Analysis

# Detect more subtle camera movements across all topics
FILTER_SHAKE_THRESHOLD=2 FILTER_MOVEMENT_THRESHOLD=0.5 python scripts/filter_usage.py

Movement Detection Only

# Enable only movement detection, disable other analysis for all topics
FILTER_CALCULATE_CAMERA_STABILITY=false FILTER_CALCULATE_VIDEO_PROPERTIES=false python scripts/filter_usage.py

Custom Output Location

# Specify custom output paths
OUTPUT_VIDEO_PATH="/output/analysis.mp4" python scripts/filter_usage.py

RTSP Stream Analysis

# Analyze live RTSP stream
VIDEO_INPUT="rtsp://admin:[email protected]:554/stream" python scripts/filter_usage.py

Disable Text Overlays

# Run analysis without visual overlays
FILTER_SHOW_TEXT_OVERLAYS=false python scripts/filter_usage.py

Custom ROI Analysis

# Focus analysis on specific region for all topics
FILTER_ROI="[100, 100, 400, 300]" python scripts/filter_usage.py

Multi-Topic Pipeline Integration

# Example: Process multiple video streams with Visual Calibration
(Vizcal, {
    "id": "vizcal_multi",
    "sources": [
        "tcp://localhost:5552;main",      # Analyze main stream
        "tcp://localhost:5554;stream2",   # Analyze stream2
        "tcp://localhost:5556;stream3",   # Analyze stream3
    ],
    "outputs": "tcp://*:5580",
    "calculate_camera_stability": True,
    "calculate_video_properties": True,
    "calculate_movement": True,
    # Each topic gets independent analysis
}),

Output Data Structure

The filter provides real-time analysis data through frame metadata and visual overlays with per-topic analysis:

Multi-Topic Frame Metadata

Each processed frame includes analysis data in the frame's metadata, with independent analysis per topic:

{
  "main": {
    "camera_stability": {
      "shake_distance": 2.3,
      "is_stable": true,
      "stability_category": "Stable"
    },
    "video_properties": {
      "frame_width": 1920,
      "frame_height": 1080,
      "fps": 30
    },
    "movement_detection": {
      "movement_detected": false,
      "movement_distance": 0.5
    },
    "analysis_timestamp": "2024-01-15T10:30:45Z"
  },
  "stream2": {
    "camera_stability": {
      "shake_distance": 1.8,
      "is_stable": true,
      "stability_category": "Stable"
    },
    "video_properties": {
      "frame_width": 1920,
      "frame_height": 1080,
      "fps": 30
    },
    "movement_detection": {
      "movement_detected": true,
      "movement_distance": 2.1
    },
    "analysis_timestamp": "2024-01-15T10:30:45Z"
  }
}

Visual Overlays (Per-Topic)

When FILTER_SHOW_TEXT_OVERLAYS=true, the filter adds real-time analysis information directly to the video frames for each topic:

• Camera stability status per topic
• Movement detection alerts per topic
• Video properties per topic
• Analysis metrics per topic
• Independent overlays for each stream

When to Use

Perfect For:

• Multi-Camera Security Setup: Verify camera stability and positioning across multiple cameras
• Surveillance System Calibration: Ensure optimal camera placement with independent analysis
• Video Quality Assessment: Analyze video clarity and consistency across multiple streams
• Research Applications: Study camera behavior and environmental factors with per-topic analysis
• System Integration: Validate video input quality in production systems with multi-stream support
• Troubleshooting: Diagnose video quality issues across different video sources
• A/B Testing: Compare different video processing approaches with independent analysis
• Pipeline Branching: Analyze different processing paths independently

Consider Alternatives For:

• Real-time Processing: High-latency analysis may not suit real-time applications
• Simple Video Playback: Overkill for basic video viewing
• Batch Processing: Designed for continuous analysis, not batch operations

Performance Tuning

Shake Detection Sensitivity

• Lower threshold (1-3): More sensitive, detects subtle movements
• Higher threshold (10-20): Less sensitive, only major shake events
• Default (5): Balanced approach for most use cases

Movement Detection

• Lower threshold (0.1-0.5): Detects small movements and drift
• Higher threshold (2.0-5.0): Only significant camera movements
• Default (1.0): Good balance for typical surveillance scenarios

Analysis Features

• Camera Stability: Adds shake detection overhead but provides stability analysis
• Video Properties: Minimal overhead, provides technical video information
• Movement Detection: Adds optical flow processing overhead but provides movement analysis

Troubleshooting

Common Issues

High CPU Usage

• Disable movement detection if not needed (FILTER_CALCULATE_MOVEMENT=false)
• Disable video properties calculation if not needed (FILTER_CALCULATE_VIDEO_PROPERTIES=false)
• Use lower resolution input videos
• Increase FILTER_LOG_INTERVAL to reduce processing frequency

False Positive Shake Detection

• Increase FILTER_SHAKE_THRESHOLD value
• Check for environmental vibrations (fans, machinery)
• Verify camera mounting stability

Missing Analysis Data

• Ensure the filter is properly configured with required parameters
• Check that input video is accessible and readable
• Verify Webvis port is available

Poor Video Quality

• Check input video resolution and bitrate
• Ensure adequate lighting conditions
• Verify camera focus and positioning

Pro Tip

For best results, run Visual Calibration on a representative sample of your video content to establish optimal threshold values for your specific use case and environment.

Important

Ensure the input video file is accessible and the output directory has write permissions. The filter will automatically create output directories but requires proper file system access.