Service scope and implementation plan
direction numberD43
Demand scenario8 items
main inputSWIR/hyperspectral L5; NIR/SWIR L5; LWIR+NIR/SWIR L4/L5; LWIR/MWIR L2/L3
Typical routeT12、T13、T16、T03、T05、T04、T19、T02
The project scope is determined based on samples, equipment, working conditions and written indicators; general modules, special models and multi-sensor integration are configured according to the evaluation results.
Application scenarios and development content
IR337Bruising, spoilage and abnormal water content of fruits and vegetables SWIR Sorting
- Minimum input and technical route
- SWIR/Hyperspectral L5;T12+T13
- target output
- Defect categories, areas and sorting results
- Implementation
- Need to slice/physical and chemical truth value
- critical acceptance
- Each type of Recall, missed points/10,000 pieces, batch migration
IR338Spectral Classification of Meat Fat, Moisture and Foreign Matter
- Minimum input and technical route
- NIR/SWIR L5；T13
- target output
- Ingredient categories/regression values and foreign matter
- Implementation
- Requires laboratory truth value
- critical acceptance
- RMSE, each type of Recall, temperature batch stratification
IR339Auxiliary assessment of fish freshness and temperature distribution
- Minimum input and technical route
- LWIR+NIR/SWIR L4/L5；T12+T16
- target output
- Surface temperature, spectral characteristics and freshness proxies
- Implementation
- Need to test the true value
- critical acceptance
- External verification, cold chain condition stratification
IR340Baking, frying and cooking surface temperature uniformity
- Minimum input and technical route
- LWIR/MWIR L2/L3；T03+T05
- target output
- Food surface temperatures, hot zones and process curves
- Implementation
- Partially reusable
- critical acceptance
- Compare with core temperature/quality; surface temperature is not equal to degree of ripening
IR341Frozen product warming, thawing and refreezing thermal modes
- Minimum input and technical route
- LWIR L2/L3；T05
- target output
- Surface temperature profiles and anomaly candidates
- Implementation
- Product thermal model required
- critical acceptance
- Compare with core temperature/ice crystal/quality test
IR342Food packaging heat sealing notch and sealing temperature
- Minimum input and technical route
- LWIR L2/L3；T03+T04
- target output
- Sealing Continuity, Temperature and OK/NG
- Implementation
- Production line data required
- critical acceptance
- Compare to leak/peel test
IR343Container filling level and hot filling temperature
- Minimum input and technical route
- LWIR L2/L3；T04+T19
- target output
- Vessel, level thermal boundaries and temperatures
- Implementation
- Container material verification required
- critical acceptance
- Compare with weighing/level/contact temperature
IR344Food production line CIP cleaning and residual liquid heat trace review
- Minimum input and technical route
- LWIR L2/L3；T02+T05
- target output
- Residual liquid/hot and cold zone candidates and process stages
- Implementation
- Hygiene verification required
- critical acceptance
- Compare with sampling/conductivity/visual inspection; does not replace sanitary release
Implementation and delivery methods
- Project input confirmationCheck camera, spectral band, bit depth, SDK, environmental conditions, sample and acceptance true value.
- Sample verificationVerify key processing links, applicable conditions and abnormal working conditions, and determine the development scope and acceptance methods.
- Algorithm development and integrationImplement data parsing, algorithms, quality gating, interfaces and target platform deployment.
- Test acceptance and handoverRetest according to the agreed test set and target platform, and hand over the deliverables, test results and applicable conditions.
- Requirements and technical solutions, input and output and interface field definitions
- Algorithm library, model or SDK/API; delivery form is confirmed in the project plan
- Sample program, parameter configuration, error status and calling instructions
- Test set version, indicator definition, stratification results and failed sample records
- Target platform adaptation, resource usage and running delay records
- Version, dependencies, license, reproduction steps and applicable conditions description
Acceptance design
Each scenario establishes independent test cases according to "key acceptance". When sharing models across scenarios, results are reported separately for each scenario, distance, load, environment, and target size, supplemented by overall statistics.
When high-risk alarms or controls are involved, the algorithm output is used for review or linkage judgment; safety functions, regulatory responsibilities, and automatic disposal require separate design and acceptance.
Frequently Asked Questions for Enterprise Customers
Can I purchase directly based on one of the scenes?
Can be evaluated first as a separate work package. The formal scope should describe the device, data, target objects, environment, interfaces, delivery form, and acceptance thresholds.
Can it be started when there are no abnormal samples?
The data interface, normal baseline, acquisition plan and simulation/controlled test design can be completed first; the abnormal detection rate and fault classification effect can be evaluated after obtaining sufficient abnormal true values.
Can the algorithm directly trigger alarms or equipment actions?
Duration, deduplication, quality gating, manual review, communication failure and safety interlocking need to be designed based on business risks. High-risk actions should be activated after completion of independent safety mechanisms and on-site acceptance.