Chapter 8: WIA-SEMI-013 Standard Specification

Overview

The WIA-SEMI-013 standard provides comprehensive specifications for 3D image sensor technologies, encompassing ToF, structured light, and stereo vision systems. By establishing common data formats, API interfaces, communication protocols, and certification criteria, the standard enables interoperability and accelerates adoption of 3D sensing across industries.

Four-Phase Architecture

WIA-SEMI-013 follows the WIA four-phase standardization framework, progressively addressing data representation, programmatic interfaces, communication protocols, and system integration. All four phase specifications are version 1.0 (2025-01-15).

Phase 1: Data Format

Defines standardized representations for 3D sensor data and specifications.

Sensor Specification Schema: Defined in JSON Schema (draft-07), with standard, version, and sensor required. Within sensor, technology (iToF, dToF, structured_light, stereo, hybrid), resolution, frameRate (1-240), and depthRange are required; accuracy (value, units, at_distance), field of view, illumination wavelength (400-1550nm), and illumination type are optional.
Depth Map Format: A binary format in which a 64-byte header is followed by depth values (uint16, uint32, or float32) and an optional confidence map (1 byte per pixel, 0-255). A CSV format is also defined for debugging and interchange.
Point Cloud Format: An extended PCD v0.7 format with x, y, z, rgb, intensity, confidence, and timestamp fields, plus a JSON format for small datasets.
Calibration Data: Defines intrinsic parameters (fx, fy, cx, cy and Brown-Conrady distortion coefficients k1, k2, p1, p2, k3), extrinsic parameters (quaternion rotation and translation in meters), and a temperature compensation table holding per-temperature depth scale and baseline.

The following is the sensor specification example included in Phase 1.

{
  "standard": "WIA-SEMI-013",
  "version": "1.0",
  "sensor": {
    "technology": "iToF",
    "manufacturer": "Sony",
    "model": "IMX556PLR",
    "resolution": {"width": 640, "height": 480},
    "frameRate": 30,
    "depthRange": {"min": 0.2, "max": 5.0, "units": "meters"},
    "accuracy": {"value": 5, "units": "mm", "at_distance": 2.0},
    "fieldOfView": {"horizontal": 87, "vertical": 58, "units": "degrees"},
    "wavelength": 940,
    "illumination": "active"
  }
}

Depth Map Binary Header

Offset	Size (bytes)	Field	Description
0	4	`magic`	0x57494133 ("WIA3")
4	2	`version`	Format version (major.minor)
6	2	`encoding`	0=uint16, 1=uint32, 2=float32
8	4	`width`	Image width (pixels)
12	4	`height`	Image height (pixels)
16	4	`timestamp_sec`	Unix timestamp (seconds)
20	4	`timestamp_usec`	Microseconds
24	4	`depth_scale`	Scale factor (depth_mm = value × scale)
28	4	`min_depth`	Minimum valid depth (mm)
32	4	`max_depth`	Maximum valid depth (mm)
36	4	`confidence_offset`	Confidence map offset (0 if absent)
40	24	`reserved`	Reserved for future use

File extensions are .wia-depth (binary) and .wia-depth.csv (text) for depth maps, .pcd for point clouds, .wia-calib.json for calibration data, and .wia-sensor.json for sensor specifications. Depth is measured along the Z axis (perpendicular to the sensor plane), not as Euclidean distance from the optical center. Version 1.x formats are backward compatible, and parsers must ignore unknown fields. Phase 1 also references ISO 8601:2019 for timestamps, IEEE 754-2019 for floating point, and IEEE 1588-2019 (PTPv2) and RFC 5905 (NTPv4) for network time synchronization.

Phase 2: API Interface

Provides RESTful APIs and SDKs for sensor control, data acquisition, and processing.

Device Management: GET /api/v1/devices discovers devices, and GET /api/v1/devices/{device_id} returns a device's specifications, capabilities, and current configuration.
Configuration: PUT .../config sets resolution, frame rate, depth quality, temporal and spatial filters, and power mode; PUT .../preset applies the face_authentication, gesture_recognition, navigation, inspection, or long_range preset.
Data Streaming: POST .../stream/start requests depth, rgb, and confidence streams and returns a WebSocket URL. Each binary message begins with a 32-byte header carrying frame type, sequence number, microsecond timestamp, width, height, and encoding.
Calibration Access: GET .../calibration retrieves intrinsic and extrinsic parameters and temperature compensation, and PUT .../calibration/user applies user calibration values.
Processing API: Provides endpoints for depth-to-point-cloud conversion and point cloud filtering (statistical outlier removal, voxel downsampling).

SDKs are provided for TypeScript/JavaScript (@wia/semi-013-sdk), Python (wia-semi-013), C++ (CMake package WIA_SEMI_013), and as a ROS 2 package; the ROS 2 node publishes the /wia/depth/image, /wia/depth/camera_info, /wia/points, and /wia/rgb/image topics. For conformance, Phase 2 requires an OpenAPI 3.1.0 description, RFC 9457 Problem Details error responses, at minimum bearerAuth authentication, and rate-limit headers on rate-limited endpoints. TLS 1.3 is recommended for transport, with TLS 1.2 allowed as the minimum.

Phase 3: Protocol

Defines communication protocols for networked 3D sensing systems.

WIA-PC-Stream: A point cloud streaming protocol over UDP (with an optional reliability layer), serialized with Protocol Buffers v3, compressed with LZ4 or Zstandard, and optionally encrypted with AES-256-GCM. It has four latency modes: ultra_low (under 10ms), low (under 50ms), balanced (under 100ms), and reliable (under 500ms).
WIA-Depth-Stream: Differentially encodes depth maps using periodic keyframes and delta frames carrying the difference from the previous frame, and can transmit only a region of interest (ROI).
Synchronization: Hardware synchronization captures on the rising edge of a GPIO trigger and keeps inter-sensor jitter under 100μs. Software synchronization uses NTP- or PTP-based timestamps, targeting error under 1ms on a LAN and under 10ms over a WAN.
Security: Specifies JWT-based device authentication with ECDSA signatures, ECDH key exchange with AES-256-GCM stream encryption (rekeying every 1GB or every hour), and secure boot (TPM 2.0 attestation) for biometric and safety-critical applications.
Error Recovery: Defines Reed-Solomon forward error correction (FEC), NACK-based selective retransmission, and an exponential-backoff reconnection policy.

Protocol conformance has three levels. Level 1 (Basic) requires single-sensor streaming and software synchronization; Level 2 (Standard) requires multi-sensor synchronization, TLS encryption, and QoS support; Level 3 (Advanced) requires hardware synchronization, end-to-end encryption, and real-time FEC. These conformance levels are a separate scheme from the Bronze, Silver, and Gold performance certification below.

Phase 4: Integration

Provides guidelines and reference implementations for system integration.

Multi-Sensor Fusion: Distinguishes homogeneous fusion, which combines sensors of the same technology, from heterogeneous fusion, which combines different technologies such as ToF and stereo, and presents weighted-average, confidence-based selection, and Kalman filter fusion algorithms.
SLAM Integration: Provides reference implementations of a visual-inertial odometry pipeline combining depth and RGB frames with IMU data, and of loop closure detection.
Application Templates: Provides requirements and implementation examples for face authentication, gesture recognition (21-joint hand tracking), and robot navigation (obstacle detection).
Testing and Certification: Defines accuracy, repeatability, multi-path, and environmental test procedures and the Bronze, Silver, and Gold certification criteria.
Alignment with Related Standards: For safety-related integration, requires declaring the functional safety standards applicable to the domain, such as ISO 13849-1, IEC 61508, ISO 10218, ISO/TS 15066, ISO 26262, and IEC 60601-1.

Certification Levels

WIA-SEMI-013 Phase 4 defines three certification levels based on performance criteria.

Parameter	Bronze	Silver	Gold
Depth Accuracy	±5mm @ 2m	±2mm @ 2m	±1mm @ 2m
Minimum Resolution	320x240	640x480	1280x960
Minimum Frame Rate	15 fps	30 fps	60 fps
Operating Environment	Indoor	Indoor + Outdoor	Full sunlight (100,000 lux)
Multi-path Handling	Basic processing	Advanced correction	AI-based processing
Temperature Range	-20°C to +60°C	-40°C to +85°C	-40°C to +85°C

Compliance Testing

Certification requires passing the following standardized tests.

Accuracy Test: Mount the sensor on a stable tripod and place a calibration target at 0.5m, 1m, 2m, 3m, and 5m. Capture 100 frames at each distance, compute the mean error and standard deviation, and verify that accuracy is within the specified limit at every distance.
Repeatability Test: Collect 1000 frames of a static scene, compute the per-pixel standard deviation, and verify that 95% of pixels are within the precision specification.
Multi-Path Test: In a corner retroreflector configuration, measure depth in the corner region, compare it with ground truth, and verify that the error is within 2x the normal accuracy specification.
Environmental Test: Verify operation under temperature (per the table above), humidity 10-90% RH (non-condensing), ambient light of 100,000 lux (Gold), vibration of 10-2000 Hz at 2g, and electromagnetic immunity per IEC 61000-4-3 (10 V/m).
API Compliance Test: Verify against the published OpenAPI description that each endpoint follows its declared schema.

A certification application requires a test report from an approved laboratory, a calibration certificate, a user manual stating the accuracy specifications, a software SDK compliant with WIA-SEMI-013, and a declaration of conformity.

Reference Implementation

The following is the TypeScript SDK example from Phase 2.

import { WIA3DImageSensor, DepthFrame } from '@wia/semi-013-sdk';

// Initialize sensor
const sensor = new WIA3DImageSensor({
  deviceId: 'sensor-001',
  resolution: { width: 640, height: 480 },
  frameRate: 30
});

await sensor.start();

// Stream depth frames
sensor.on('depth-frame', (frame: DepthFrame) => {
  // Depth at the center pixel
  const depth = frame.getDepth(320, 240);

  // Convert to point cloud
  const pointCloud = frame.toPointCloud({
    removeInvalid: true,
    voxelSize: 0.01  // 1cm downsampling
  });

  // Export as PCD
  const pcdData = pointCloud.exportPCD();
});

// Configure depth quality and temporal filter
await sensor.configure({
  depthQuality: 'high',
  filters: {
    temporal: { enabled: true, frames: 3 }  // 3-frame temporal filter
  }
});

await sensor.stop();

Adoption and Implementation Strategy

For Sensor Manufacturers

Implementing WIA-SEMI-013 compliance phase by phase yields a clear benefit at each step.

Implementation Phase	Actions Required	Benefits
Phase 1: Data Format	Support sensor specification JSON, WIA depth map, and calibration formats	Interoperability with WIA tools
Phase 2: API	Implement REST API and SDK interfaces	Developer ecosystem compatibility
Phase 3: Protocol	Support WIA-PC-Stream and WIA-Depth-Stream, synchronization, and security	Multi-vendor system integration
Phase 4: Integration and Certification	Apply integration guidelines, apply for performance certification testing	Objective performance evidence, meeting procurement requirements

For Application Developers

Developing against WIA-SEMI-013 reduces integration effort and improves application portability:

Sensor Abstraction: Code against standard interfaces, swap sensors without rewriting application logic.
Library Ecosystem: Leverage shared tools (visualization, calibration, processing) across projects.
Quality Assurance: Because certification levels and test conditions are published, the scope of in-house field testing can be narrowed.
Backward Compatibility: Version 1.x formats are backward compatible and parsers ignore unknown fields, so existing applications keep working as the specification is revised.

For System Integrators

Building systems with WIA-compliant components simplifies integration and ongoing support. Multi-vendor sourcing reduces supply chain risk. Standardized interfaces enable component upgrades without system redesign. To claim Phase 4 conformance, however, subsystems must comply with Phases 1-3, the applicable functional safety standards must be declared, and a risk analysis appropriate to the application domain must be performed and documented.

Future Roadmap

The Phase 4 specification (v1.0, dated 2025-01-15) sets out the following revision plan. The dates below are those stated in the v1.0 specification; check the current WIA release status for the latest schedule.

Version 1.1 (planned for Q2 2026): Event-based 3D sensors, neural radiance field (NeRF) integration, enhanced multi-sensor calibration.
Version 2.0 (planned for 2027): Compressed neural representations, 5G edge computing integration, AR Cloud for persistent maps.

Phase 1 lists NeRF representations, compressed neural point clouds, multi-spectral depth data, and event-based depth streams as candidates that may be introduced in data format version 2.0.

Conclusion

The WIA-SEMI-013 standard provides a comprehensive framework for 3D image sensor technology, from low-level data formats through high-level application integration. By following this standard, manufacturers ensure interoperability, developers gain consistent APIs across sensors, and end users benefit from certified performance levels.

As 3D sensing becomes ubiquitous in smartphones, automobiles, robots, and AR/VR systems, standardization becomes increasingly critical. WIA-SEMI-013 accelerates innovation by providing a stable foundation upon which the entire ecosystem can build.

Summary

            Key Takeaways:
            Four-Phase Framework: WIA-SEMI-013 progressively addresses data formats, API interfaces,
                    communication protocols, and integration (including certification), enabling phased adoption and backward compatibility.
Interoperability First: The standard's primary goal is enabling sensors, software, and
                    applications from different vendors to work together seamlessly.
Technology Agnostic: Specifications cover ToF, structured light, and stereo vision equally,
                    avoiding favoritism toward particular technologies.
Open and Accessible: SDKs, tools, and reference implementations are open source, lowering
                    barriers to adoption for startups and established companies alike.
Certification Provides Trust: Bronze, Silver, and Gold tiers give buyers confidence in
                    sensor performance without requiring in-house testing infrastructure.
Living Standard: While keeping the version 1.x backward-compatibility principle, versions 1.1 and 2.0
                    plan to incorporate new technologies such as event-based sensors, NeRF, and neural representations.
Ecosystem Enablement: By standardizing the foundation, WIA-SEMI-013 accelerates innovation
                    at higher levels - algorithms, applications, and complete solutions.

        

Review Questions

Explain the four-phase architecture of WIA-SEMI-013. Why is this progressive approach beneficial compared to a monolithic standard?
What are the three certification tiers (Bronze, Silver, Gold) and what performance levels do they represent? Why is tiered certification useful?
Describe the WIA depth map binary format header structure. What information does it encode and why is each field necessary?
Compare the TypeScript SDK approach versus the REST API approach for accessing 3D sensors. When would you choose each?
A manufacturer wants to make their proprietary ToF sensor WIA-SEMI-013 compliant. Outline the steps and what they need to implement at each phase.
What is the purpose of the temperature compensation table in the calibration format? Why can't sensors just use fixed calibration values?
Explain how WIA-SEMI-013 supports multi-sensor fusion. What specific formats and APIs enable combining data from different sensors?
Why does the standard include both binary and JSON/CSV formats for the same data? What are the use cases for each?

弘益人間 (Hongik Ingan)
Broadly Benefiting Humanity