Depth Sensor Per-Pixel Calibration Non-Linear Error Model

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional depth sensor calibration techniques often result in limited accuracy, leading to incorrect distance measurements, which can be detrimental to applications such as 2D and 3D scene reconstruction, robot navigation, and manipulation, due to hardware capabilities and traditional calibration methods.

Innovation Solution

A method and system for correcting input depth images and calibrating depth sensors using a non-linear error model, where per-pixel correction values are determined and applied to compensate for systemic biases, utilizing a parameter matrix or lookup table for error correction, and patch-based calibration data to improve accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional per-sensor calibration is used during manufacturing, then the calibration process is simple and fast, but the measurement precision of depth readings is limited

Engineering Contradiction:
Improvedepth reading accuracyVSAvoidcalibration system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The calibration process is segmented into two distinct phases: (1) conventional per-sensor calibration during manufacturing for basic setup, and (2) a subsequent per-pixel calibration phase using captured depth images and reference measurements. This segmentation allows the system to achieve high measurement precision through the second phase while maintaining the simplicity and speed of the first phase during production.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies preliminary action by performing conventional calibration during manufacturing before deployment, establishing a baseline calibration state. Then, upon deployment, the system performs additional per-pixel calibration using captured depth images and reference distance measurements to achieve the desired high precision without requiring complex manufacturing processes.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If conventional calibration techniques are applied, then the calibration process is quick and easy, but depth sensors report incorrect distances compared to precision instruments

Engineering Contradiction:
Improvedistance measurement accuracyVSAvoidcalibration time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary conventional calibration during manufacturing to establish basic operational parameters quickly. Then, upon deployment, it performs a more thorough per-pixel calibration using captured depth images and reference measurements from precision instruments, achieving accurate distance measurements while accepting the additional calibration time only when needed for high-precision applications.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The calibration system performs self-calibration by capturing its own depth images, comparing measurements against reference distances, and automatically generating per-pixel correction values without requiring external intervention or complex manual adjustment procedures.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If per-pixel correction is applied to correct depth readings, then measurement precision improves, but the complexity of the calibration system increases

Engineering Contradiction:
Improvedepth value accuracyVSAvoidcalibration data structure complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The calibration data is segmented into per-pixel correction values stored in a structured format (parameter matrix or lookup table), separating the correction data from the conventional calibration parameters. This segmentation allows the system to maintain simple conventional calibration while adding sophisticated per-pixel corrections without overwhelming system complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary calibration layer that sits between the conventional per-sensor calibration and the final depth readings. This intermediary per-pixel correction layer processes the depth values using correction values derived from captured images and reference measurements, achieving high precision without directly modifying the core sensor hardware or conventional calibration processes.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Ease of manufacture

If conventional calibration is used, then manufacturing is simple and fast, but the depth sensor produces artifacts such as vertical bands in depth images

Engineering Contradiction:
Improvemanufacturing simplicityVSAvoiddepth image quality
Core Design Contradiction:
Ease of manufactureVSReliability

Solution Approach 1:

The calibration approach is segmented into conventional per-sensor calibration during manufacturing (maintaining simplicity) and per-pixel calibration after deployment (improving quality). This segmentation allows manufacturers to produce sensors quickly and simply while enabling end-users to achieve high-quality depth images without artifacts through the additional calibration step.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Conventional calibration is performed as a preliminary step during manufacturing to enable basic operation and keep the manufacturing process simple. The per-pixel calibration that eliminates artifacts is performed subsequently during deployment, allowing the system to achieve high reliability and clean depth images without complicating the manufacturing process.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3134866B1Depth sensor calibration and per-pixel correction
Publication Date: 2020.02.19 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3134866B1 patent drawingFigure 1
  • EP3134866B1 patent drawingFigure 2
  • EP3134866B1 patent drawingFigure 3

AI summary

Various technologies described herein pertain to correction of an input depth image captured by a depth sensor. The input depth image can include pixels, and the pixels can have respective depth values in the input depth image. Moreover, per-pixel correction values for the pixels can be determined utilizing depth calibration data for a non-linear error model calibrated for the depth sensor. The per-pixel correction values can be determined based on portions of the depth calibration data respectively corresponding to the pixels and the depth values. The per-pixel correction values can be applied to the depth values to generate a corrected depth image. Further, the corrected depth image can be output.