3D body data can both identify a person and reveal body shape, which places it in the strictest category under China's Personal Information Protection Law. Five concrete actions cover capture, retention, export and deletion.
3D body data looks like "a pile of points". In fact it has two properties at once: it can identify a specific person, and it reflects body shape. Together they place it in the strictest tier of personal information protection. Institutions buying equipment usually ask about accuracy and rarely ask who owns the data afterwards, how long it is kept, and who can see it.
Compliance is not a document for the legal team. It is five concrete actions. Do them and daily operations stay clear of the obvious traps.
1. Separate three data categories — they are handled differently
Category
Sensitivity and handling
Raw point cloud and 3D model
Most sensitive. Allows reconstruction of body surface, is personal information and in most cases sensitive personal information. In principle stays local: no external transfer, no public hosting.
Structured report and indicators
Dimensions, angles and composition once the model is removed. Still personal information, but usable for statistics and follow-up provided de-identification and authorisation scope are clear.
Identity and contact details
Name, ID number, phone. Stored separately from the two categories above and linked by number. Never put identity into a model filename.
The commonest failure is mixing the three: a report named "Zhang San – 20261101", a model folder containing a phone number. If one item leaks, all three leak. Use a subject number throughout, and keep the identity mapping table separate and separately authorised.
2. Before capture: what to disclose, how consent is given
Disclosure must answer five questions. Miss one and it is not clear disclosure.
What is captured. 3D body surface data, height and weight — and state explicitly whether colour photographs are captured.
What it is for. Assessment, report, follow-up, research. The purpose must be specific; "to improve our service" is not a purpose.
How long it is kept. Give a period, or the rule that determines it.
Who may see it. Which internal roles, and whether any third party (a custom-product manufacturer, for example).
How to delete it. The channel for requesting deletion and the response time.
Consent must be obtained separately, not buried in a long service agreement. Sensitive personal information requires separate consent; minors' data requires guardian consent. Paper or electronic both work — what matters is that it can be retained and traced to a specific capture.
3. During capture: minimum necessary
Do not collect what the purpose does not need. A posture assessment does not need the full medical history; foot matching does not need an ID number.
Local by default. Raw data stays on the institution's own server; cloud storage is enabled only when needed, with named access rights.
Physical separation at capture. No screen facing the waiting area, no bystanders watching. This is both a privacy requirement and a factor in cooperation.
Re-tests overwrite the original record. Do not keep a second capture from a failed attempt, or the same person will have two inconsistent raw records from one day.
4. Export and sharing: de-identify, authorise, log
De-identify. Remove name, ID number and contact details; name model files by number; keep identity out of the report header. Where a human needs to recognise the subject, use the number plus the mapping table.
Authorise. Any external provision, such as sending data to a custom-product manufacturer, must be stated at disclosure, with purpose limited and re-disclosure prohibited.
Log. Who exported which records, and when. Without a log, the scope of an incident cannot be established.
Transfer. Do not send raw models through chat apps or public links. Use an approved channel, encrypted where necessary.
5. Minors: separate consent and presence
Guardian consent. Personal information of children under fourteen is sensitive personal information and requires guardian consent, with dedicated processing rules.
Presence. A guardian or teacher should be present during capture, and no unrelated person should photograph the area.
Who is told the result. Screening results go to the guardian, not announced in class or in a group chat. Exception lists follow the referral route only.
Research use separately. Using minors' 3D data for teaching or research requires separate consent and de-identification.
6. Retention and deletion: when the period ends, actually delete
"Retained indefinitely" does not survive scrutiny. Tie retention to purpose: data captured for one assessment can be archived or deleted once the assessment and follow-up are complete; data used for long-term health management is kept for the agreed period with periodic review of necessity.
Deletion requests. The channel named in the disclosure must actually work. Handle requests within the stated time and keep a record.
Media destruction. Clear local data when replacing a drive, scrapping a device or returning leased equipment. Deleting files without handling the media is not deletion.
Backups too. Backups are the usual omission. A deletion that does not cover backups is only apparent.
7. Three things to land inside the institution
Name an owner. One person responsible for data security, whom people contact when something goes wrong. IT or operations both work; a committee does not.
Keep a data inventory. Which category, stored where, visible to whom, kept how long. One page, but it must match reality.
Train once and keep the record. Front desk, technicians and sales staff all need to know what may be said, what may not be photographed and what may not be sent. The training record is important evidence later.
8. Campus screening: the four-party problem
Campus work is the most delicate case: the subjects are minors, the organiser is the school, the operator is a third party, and results go back to parents. Four parties, and any unclear link causes trouble.
Who discloses to whom. The school organises and discloses; the institution captures and reports. The content is confirmed jointly — the institution cannot push all responsibility onto the school.
Consent form. Signed by the guardian, listing capture items and purposes. A notice in a class group chat is not consent.
On-site management. Enclosed capture area, no bystanders, no photography. Teachers are present but do not read other students' reports.
Feedback. One report per person, given to the guardian. Exception lists follow referral only — no public ranking, no group comparison.
Retention. Keep or hand over within the agreed period after screening, and never transfer to a third party unrelated to the screening.
Writing these five into the cooperation agreement with the school is worth far more than explaining afterwards. Once screening volume rises, individual disputes follow; rules exist to draw the boundary in advance.
9. Two common misconceptions
"Local storage makes it safe." Local storage lowers the chance of external transfer. It does not address internal over-reach, residue on scrapped equipment, or backup leakage.
"Once de-identified it can be used freely." De-identification lowers risk; it does not change the nature of the data. A new purpose still has to fall inside the disclosed scope, and anything beyond it requires fresh consent. This is the point most often missed when data is reused across scenarios.
Next step
If you need a disclosure and consent template for a specific scenario, contact the account manager. Cooperation runs as direct purchase or through a dealer arrangement; details are discussed directly.
Note: this article describes operational practice under China's Personal Information Protection Law and is not legal advice. Confirm your own documentation with qualified counsel. This article contains no pricing.
Measurement conditions and what the result covers
The accuracy figures and posture data discussed here all rest on the same measurement conditions: the person wears close-fitting clothing or shorts, is barefoot, stands naturally, stays as still as possible during the scan (about 3–5 seconds), and has hair tied back or pinned up so it does not cover the head and shoulder outline.
When these conditions are met, the data is used for preliminary screening, risk indication and follow-up comparison of posture and spinal status; when they are not met (loose clothing, movement, hair covering the outline), the result is an estimate for reference only. The device does not provide a medical diagnosis — any abnormal finding needs review at a medical institution.
Accuracy is millimetre-level (vendor nominal ±1 mm). A living body is not a calibration artefact: millimetre-level variation across three consecutive scans of the same person is normal, so read the trend from the range of several measurements rather than from a single number.
Where the data comes from
Specifications here come from the device datasheet and user manual — nothing is estimated or inferred.
Case data comes from client project records; organization names can be verified by phone. We do not use client names without authorization.
Qualification records can be looked up by certificate number in the National Medical Products Administration database.
FAQ
Is 3D body data sensitive personal information?
In most settings, yes. It can identify a specific person and it reflects body form, which ties it directly to the person. Treating it under sensitive-personal-information rules is the safer course: specific notice, separate consent, a defined retention period. Holding to the stricter standard keeps later reuse of the data unconstrained.
Is storing the data locally enough?
No. Local storage lowers the chance of external transfer but does not address three things: internal access beyond authority, data left on media when the device is retired, and backup leakage. The complete set is local retention, access control, export logging and media destruction. Miss one and the chain has a gap.
In a school screening, can a student's report go to the form teacher?
To the guardian, not to the form teacher. At class level what is needed is an aggregated list of flagged students for organising referrals, not individual reports. Publishing individual reports to a class or a group chat goes beyond the scope that was notified.
A custom-product manufacturer wants the model data. Can we share it?
Yes, provided the purpose was notified and consent obtained at capture, with use limited and onward provision prohibited. De-identify on delivery, using an ID rather than a name. Where structured dimensions are enough, do not hand over the full model. The less you pass on, the less liability you carry.