3D Vision & AI Inference for Industrial Textile Sorting

The Challenge: High-Volume Textile Classification

Industrial laundry and textile handling facilities need to identify garments, foreign objects, and handling risks on fast-moving conveyors. The environment also includes heat, humidity, vibration, and lint that can shorten the life of standard workstations.

The system needed to:

1. Process Visual Data Locally: Classify textiles and objects without sending every frame to a remote server.
2. Operate in a Harsh Environment: Handle heat, humidity, vibration, and textile fibers.
3. Support 3D Data: Feed point-cloud and camera data into a machine vision model with predictable latency.

The Solution: YX-E 500 Vision Node

Yantronic YX-E 500 was selected as the compute platform for the sorting robot.

Engineering Advantages of YX-E 500

• GPU Expansion: The PCIe expansion path supports GPU-accelerated inference for camera and 3D vision workloads.
• Rugged Mechanical Design: Industrial enclosure and mounting options help the system remain stable near moving machinery.
• Thermal Management: Heat-pipe cooling and airflow planning support sustained AI workloads in warm production environments.

Technical Architecture

The system integrates sensing, edge processing, and machine control:

• Sensors: 3D cameras provide spatial data for object and textile classification.
• Edge Processing: The YX-E 500 runs the inference workload close to the conveyor.
• Execution: The control layer sends low-latency signals to sorting actuators and downstream handling equipment.

Impact & Results

The deployment improved the reliability and maintainability of the classification line:

• Throughput Support: Local inference helped keep classification close to conveyor timing.
• Safety Support: Vision-assisted sorting reduced reliance on manual inspection in higher-risk handling areas.
• Maintenance Fit: Industrial thermal and mechanical design reduced the need to adapt consumer workstations for the production floor.