Optical Lens Inspection AI

Add a real-time inspection layer to every lens with vision AI for optical lens inspection. Built for the operations where the part is transparent, curved, and coated, a defect on the surface and a speck of dust sitting on it look the same in one light and different in another, the scratch-dig grade decides whether the lens ships or gets scrapped after the coating run, and a single particle sealed inside a module is a return. Whether you're polishing and coating precision optics, surfacing ophthalmic lenses, or assembling camera and sensor lens modules, Roboflow inspects the surface, the coating, the edge, and the assembly on every lens, with a graded record behind every serial.

Surface and Bulk:

• Detect scratches, digs, pits, and sleeks on convex, concave, and flat surfaces, and grade them against ISO 10110-7 and MIL-PRF-13830B scratch-dig limits, so the record carries the grade and the location
• Detect bubbles, inclusions, striae, and stones in the bulk material, so the flaw inside the glass or polymer is caught before it is coated
• Separate cleanable contamination from true surface defects, so a lens that needs a wipe is not scrapped and a lens with a real scratch is not shipped

Coating and Edge:

• Detect anti-reflective and functional coating pinholes, stains, spatter, crazing, and delamination, the defects that show in reflected light at one angle and disappear at another
• Detect edge chips, bevel and chamfer defects, and blocking or edging marks, so the edge is inspected with the surfaces instead of after them
• Trend defects by coating run, chamber, and polishing lot, so a chamber that starts leaving spatter surfaces before the next run

Assembly and Systems Integration:

• Detect cement bubbles and voids in doublets and triplets, particles sealed between elements, decentered or tilted elements visible in the module, and contamination inside camera and sensor lens modules
• Inspect at every value-added step, from polished blank to coated element to assembled module, so the defect is caught before the next step is spent on it
• Integrate with MES, QMS, and inspection stations, with per-lens records tied to serial, lot, station, grade, and imagery

Bring intelligence to every lens today. Stop scratches, coating stains, and sealed particles from becoming scrap after coating, module returns, or the lot held while the sample is re-graded.

More About Optical Lens Inspection

What is optical lens inspection with Vision AI?

Optical lens inspection with vision AI uses computer vision models to inspect lenses and lens assemblies at every stage of production, from polished blank to coated element to assembled module: scratches, digs, pits, and sleeks graded against scratch-dig limits, bubbles, inclusions, and striae in the bulk, coating pinholes, stains, spatter, and delamination, edge chips and bevel defects, dust and contamination separated from true defects, and cement voids, sealed particles, and element misalignment in doublets and modules. Models trained on your actual lenses, coatings, and lighting run on every lens, with per-lens records that support ISO 10110-7 and MIL-PRF-13830B grading, ANSI Z80.1, ISO 13485 for ophthalmic devices, and ISO 9001.

Can Vision AI tell a scratch from dust on a coated lens?

Telling a defect from contamination on a curved, coated, transparent surface is one of the highest-stakes calls in optics: a scratch and a fiber can look the same in bright field, a coating stain shows in reflected light at one angle and vanishes at another, and the grade decides whether a lens that already carries a polishing run and a coating run ships or gets scrapped. Deep-learning models trained on your actual lenses, coatings, and inspection lighting learn what a scratch, a dig, a coating pinhole, and a speck of dust each look like on your product, and multi-angle imaging with dark-field, bright-field, and reflected illumination gives the model the views an inspector would tilt the lens to get. Cleanable contamination is classified as such, so the lens gets a wipe and a re-inspection instead of a scrap ticket, and the true defect comes back sized and located against the specification, with the frame attached, so the grade is a record and not a judgment call.

Does this replace our interferometer and MTF testing?

No. Interferometry, MTF, and centration testing keep their role as the measurement of optical performance, surface figure, and wavefront, and the specification limits, the sampling plans, and the ISO 10110 drawing notes stay as they are. Vision AI adds what performance testing does not produce: cosmetic and surface imperfection grading on every lens instead of a sample, coating and edge inspection, contamination separated from defects, module and cement inspection after assembly, and per-lens imagery that documents the grade. Metrology results and vision inspections land in the same per-lens record, and the lens with a coating stain that would have passed MTF and failed the customer gets caught before it ships.

Can it integrate with our MES, QMS, and inspection stations?

Yes. Roboflow Inference exposes a standard API and supports common industrial protocols, so inspection results flow into your existing systems: MES and ERP platforms like SAP and Oracle, QMS platforms like MasterControl and ETQ, and SCADA and HMI platforms like Ignition and AVEVA, through REST, MQTT, OPC UA, and direct database writes. Station-level integration holds or diverts a lens the moment a check fails, and every event carries the lens serial, lot, station, defect class, size, location, grade, imagery, and disposition, with a full record behind every lens for customer certificates and audits.

‍