Application Practice of CHNSpec FS-13 Hyperspectral Camera in Leather Defect Detection

In the leather production and quality control process, subtle defects such as glue leakage and scratches directly affect product grading and market value. Traditional manual visual inspection is easily affected by subjective judgment and fatigue, leading to problems such as low efficiency, inconsistent standards, and frequent missed inspections. Conventional optical testing equipment mostly relies on spatial morphological information and has limited ability to identify optical differences caused by microscopic changes in materials, making it difficult to meet the needs of refined quality inspection.

Hyperspectral imaging technology can simultaneously obtain the spatial image and continuous spectral information of the target, with each pixel corresponding to a complete high-resolution spectral curve. Since there are differences in composition and surface structure between leather defect areas and normal areas, the reflection spectra and colorimetric parameters of the two form quantifiable differences in specific bands, providing data support for objective and stable defect identification.

I. Experimental Scheme and Equipment Configuration

In this case, the CHNSpec FS-13 hyperspectral camera was used to carry out leather defect detection verification. The equipment and parameter settings were tailored to the characteristics of leather samples:

• Spectral Range: 400–1000nm
• Spectral Resolution: 2.5nm
• Working Mode: External push-broom scanning
• Key Parameters: Exposure time 200μs, motor movement speed 30 mm/s
• Sample: Leather specimens containing glue leakage defects
• Detection Goal: Extract and distinguish the spectral and colorimetric characteristics of defect areas and normal areas, and complete defect localization and visual presentation.

II. Detection Process and Data Processing

1.Data Acquisition: Scanning the entire leather surface in push-broom mode, simultaneously collecting full-band spectral data and colorimetric parameters such as L, a, b, X, Y, Z for each pixel. Reflectance curves are generated in real-time, forming an integrated "spatial + spectral" dataset.

2.Data Preprocessing and Analysis: Performing calibration and noise reduction on the raw data, focusing on comparing the morphology of reflectance curves between defect areas and normal areas, quantifying colorimetric parameter differences, extracting optical features that can be used to distinguish defects, and establishing a stable identification basis.

III. Application Effects and Measured Performance

1.Clear Spectral Feature Differences: Within the 400–1000 nm band, the reflectance curves of the glue leakage area and the normal area show quantifiable waveform differences in peak values, slopes, and characteristic wavelength positions, providing an objective basis for defect determination.

2.Good Discrimination of Colorimetric Parameters: Taking D65/10° standard observation conditions as an example, there are significant differences in L, a, b, and other values between the glue leakage area and the normal area, enabling rapid defect discrimination through numerical thresholds.

3.Precise and Traceable Defect Localization: Combining spatial images with spectral features, the distribution range and boundaries of defects can be accurately locked. Visual detection results and quantified data are outputted, making the detection process reproducible and the results traceable, which facilitates quality control and process optimization.