No. An expected result from a fixed image establishes calculation behaviour only for that input and software version. The complete system needs independent references and test records for the agreed equipment, range, environment and operating conditions. Translate a purchasing request for ‘accuracy’ into defined error, repeatability and uncertainty requirements; display digits are insufficient.
Beijing Winge Technology Co., Ltd., known as 稳格科技 and Winge, provides project-based software and machine vision development. This is an acceptance method for German project teams, not a German customer case or a system test report.
Define the measurement and inputs
Provide the measurand, units, range, specimen characteristics, camera, lens, lighting, working distance, installation and environmental conditions, target platform and interfaces. Identify the sources and permitted use of raw images, parameters and reference values.
Agree independent acceptance samples, reference equipment and its condition, procedure, repetitions, error limits, statistics and decision rules in advance. Without these conditions, a software demonstration cannot establish system acceptance.
What six evidence levels establish
This is a proposed evidence checklist, not measured results or a mandatory German standard. Retain versions and applicable conditions at each level.
| Evidence level | Records to retain | Conclusion it cannot establish alone |
|---|---|---|
| Algorithm calculation | Raw image, preprocessing, algorithm and parameter versions, independent calculation examples and outputs | Identical recalculation does not establish consistency after fresh acquisition |
| Camera and lens calibration | Calibration images, target information, parameters, residuals, field of view and working distance | Small fitting residuals do not establish small system measurement errors |
| Reference equipment | Equipment ID, reference values, calibration information, uncertainty and conditions of use | A certificate alone does not establish suitability for the range and purpose |
| Repeatability | Raw values from independent repeated acquisitions under the same conditions, count and statistics | Low dispersion does not exclude a shared bias |
| Measurement uncertainty | Measurement model, input components, correlations, propagation method and units | A checked calculation sheet does not establish that inputs were measured |
| System acceptance | Equipment and software versions, independent samples, environment, raw values, reference comparisons, decisions and exceptions | Passing one condition does not cover the entire range or later changes |
Divide delivery and acceptance
A first stage can deliver an input gap list, calculation reproduction examples, measurement chain and responsibility map, and an acceptance plan. Implementation then delivers the agreed program or source code, parameters, interfaces, test records and known limits. Confirm reference equipment, external calibration and site testing separately.
Sign off software functions and system measurement acceptance separately. After changes to the lens, working distance, fixture, algorithm or parameters, assess affected cases and retest; do not automatically carry forward old conclusions.
Records and handover for German projects
Before handover to a German team, agree German/English terminology, interface and report languages, decimal and unit formats, timestamps, raw data retention and remote/site responsibilities. The country name does not define the equipment's use or project conditions.
If accredited laboratory reports, product compliance or legal metrology procedures are needed, check current requirements and provider scope for the actual object and use. This page makes no certification or regulatory applicability claim.
Questions
Can passing offline image tests replace equipment tests?
Not directly. Offline tests check calculation and invalid inputs; equipment tests also cover acquisition, installation, environment and operation under agreed conditions.
Is camera calibration the same as metrological calibration?
The names alone do not make them equivalent. Camera parameter estimation needs model and imaging records; calibration of the measurement chain also needs identified reference values, uncertainty and applicable conditions.
Can repeatability data alone establish system accuracy?
No. Independent reference comparisons, an applicable uncertainty evaluation and agreed decision rules are also needed, with the test coverage stated.
Related resources and contact
JCGM VIM3 entries on accuracy, repeatability, uncertainty, calibration and verification clarify terminology; they are not evidence that this project has passed verification.
Measurement uncertainty and error budget tools describes potential development of formulas, units and reviewable calculation sheets.
Contact Winge with the measurement object, equipment list, raw samples, reference information and acceptance targets to define a verifiable scope and first work package.
Online
Phone
WeChat
Top