When a laser contour has gaps or spikes, compare the original observation with the expected geometry before choosing how to process it. Missing contours, multiple bright stripes from reflections and real geometric steps are different problems. Smoothing every abrupt change can remove edges that must be measured.
Preserve evidence before processing
Record raw images or device-output contours, exposure, valid-point flags, sampling position and device status. If a finished sensor does not expose images, retain the data and quality fields available through its interface rather than claiming unavailable raw pixels. Use the manufacturer's format definition to distinguish invalid values from valid numbers. Use the device's format to decide whether a zero is valid; do not treat every zero as missing data.
Check the source of each symptom
| Symptom | Initial checks | Processing limits |
|---|---|---|
| Local missing data | Illumination and viewing occlusion, working range and validity | Do not fill unknown geometry and present it as directly measured |
| Excessively bright or wide stripe | Exposure, surface reflection and saturation | Recapture after adjustments and retain versions |
| Multiple candidate stripes | Multiple reflections, background and stripe-selection rules | Preserve selection reasons and uncertain states |
| Isolated points or spikes | Original observations, neighborhoods and real edges | Deviation from a smooth curve alone is not a deletion criterion |
| Sustained dimensional jumps | Geometry, pose, sampling positions and processing settings | Review real changes separately from processing faults |
State the purpose of interpolation and filtering
Points added for display are not direct observations. If interpolation is permitted for measurement, define allowed gap sizes, regions and validity rules beforehand. Retain original points, interpolation flags and parameters. Compare key dimensions before and after processing and against references to check whether filtering hides grooves, holes or boundary changes.
Carry invalid states into downstream processing
Report invalid or pending-review status when valid coverage is insufficient or stripe selection is ambiguous. Successful geometric fitting does not prove sufficient original observations. When the current observation is invalid, report it as invalid and never label a previous measurement as a valid current result. Record causes, result states and recovery separately for equipment and software.
See the manufacturer's technical documentation for extraction, occlusion and metric-calibration principles. This article describes diagnosis and processing methods and reports no specified-sensor integration, accuracy or test acceptance.
Related services: 3D vision and point-cloud processing, project consultation.