Nx Witness AI Analytics Integration and Technical GuideLearn how to integrate AI analytics with Nx Witness VMS using RTSP/ONVIF without changing existing cameras. Technical details for smooth implementation.

On the Nx Witness platform, AI analytics can be performed without replacing cameras or NVR channels, utilizing standard protocols. Industrial image processing systems like Hype Vision provide brand-independent access through RTSP and ONVIF support.

This allows adding AI-based features to your existing infrastructure without installing new cameras or upgrading devices. However, factors such as camera image quality and network bandwidth usage can impact integration success.

• IP cameras with RTSP can connect directly to the AI server.
• ONVIF enables device discovery and channel access.
• Selecting sub-streams through NVR reduces bandwidth use.

The most common approach for AI analytics integration is obtaining RTSP streams directly from cameras or via Nx Witness/NVR devices. Typical RTSP URLs follow the format: rtsp://[username]:[password]@[ip_address]:554/streamX.

ONVIF protocol supports channel access through device interfaces and simplifies configuration via device discovery. Nx Witness VMS also allows streaming live and recorded video through its internal channels.

Using sub-streams is preferred to optimize resolution and bandwidth. For analytic purposes, a medium-quality stream like 720p is usually sufficient, reducing network load.

• RTSP URLs must be accurate and up to date.
• Device and channel identification occurs using ONVIF discovery.
• Sub-stream selection lowers bandwidth and processing demands.
• Encoder choices (H.264 or H.265) affect performance.
• Firmware updates improve connection stability and compatibility.

AI analytics with Nx Witness generally follows this architecture: camera video streams over IP networks to Nx Witness or NVR devices, then transfers to an edge server via RTSP/ONVIF.

On the edge server, Hype Vision’s Hype-Model family performs object detection, classification, pose/clothing tracking, OCR, and other analyses. Results trigger alarms on the Nx Witness panel, WhatsApp/SMS notifications, VMS pop-ups, or industrial controllers like PLCs/relays.

This structure is scalable according to wired or wireless network infrastructure and user permissions and is well suited for on-premise deployment.

• Camera → Video stream via Nx Witness/NVR
• Nx Witness → Edge server (AI processing unit)
• Edge server → Visual panel, ERP/MES, or alert systems
• Notifications: WhatsApp, SMS, VMS pop-up, PLC/relay
• Optional industrial cameras and lighting for quality control

Hype Vision’s Hype-Model family enables various AI analyses on video streams from Nx Witness. These include personal protective equipment (PPE) detection, restricted area violation alerts, forklift counting, people and vehicle counting, and quality control applications.

OCR modules are particularly useful for on-site label and barcode reading or quality assurance processes. Pose and hand tracking functions help monitor operator movements in sensitive workflows.

• PPE (helmet, jacket, gloves) compliance checks
• Restricted area entry/exit violation alerts
• Forklift and vehicle counting for traffic analysis
• Product counting and quality control (e.g., packaging errors)
• Text and barcode recognition via OCR
• Pose/hand tracking for workplace safety monitoring

User roles and access permissions must be properly configured in Nx Witness to prevent unnecessary access to AI modules. Network bandwidth is critical for the stability of RTSP and ONVIF streams.

H.264 is the most common codec, though H.265 offers bandwidth savings at the expense of higher processor load. Regular firmware updates are important to maintain device stability and protocol compatibility.

Camera positioning and lighting conditions must be optimized to enhance the performance of AI models.

• Nx Witness user and role-based authorization management
• Adequate bandwidth for RTSP sub-streams
• Video codec compatibility and network latency
• Up-to-date camera firmware versions
• Sufficient hardware resources for AI models
• Lighting and field conditions suitable for AI

Successful Nx Witness and AI analytics integration begins with analyzing camera and network infrastructure. Existing cameras’ RTSP/ONVIF support, resolution, and bandwidth are evaluated first.

Then, the edge server for AI analytics is installed and configured with proper channel routing inside Nx Witness. Model validation and testing follow. Finally, alarm and notification systems are integrated before starting the pilot phase.

The pilot phase typically lasts 2 to 4 weeks and includes performance monitoring, debugging, and parameter optimization.

• Verification of existing devices’ protocols (RTSP/ONVIF) and image quality
• Configuration of channels and user settings in Nx Witness
• Selection and installation of edge server hardware
• Testing RTSP or ONVIF connections
• Parameter tuning of AI models for target use cases
• Configuring alarm and notification channels
• Pilot operation and performance analysis