CHNSpec FigSpec FS-27 Hyperspectral Camera Assists Breakthroughs in Underground Engineering Research, Achieving Non-Destructive and Precise Monitoring of Lining Structural Compressive Strength

Recently, a research team from Tongji University has made important progress in the detection of underground concrete lining structures—relying on the FigSpec FS-27 hyperspectral camera developed by Hangzhou CHNSpec Technology, the team successfully built a non-destructive detection technology integrating hyperspectral imaging (HSI) and deep neural networks (DNN). This technology enables automated, high-precision monitoring and defect visualization of the compressive strength of underground tunnel linings. The related findings have been accepted for publication by the international top journal Journal of Rock Mechanics and Geotechnical Engineering (SCIE Q1 Top journal).

Pain Points in Underground Engineering Detection Highlight the Need for Innovation in Traditional Methods

With the acceleration of urbanization, underground infrastructures such as metro tunnels and cross-river passages continue to expand. As the core protective structure in underground engineering, the compressive strength of concrete linings directly affects structural integrity and operational safety. However, the underground environment is enclosed and complex, and traditional detection methods face many limitations: destructive methods such as core sampling and pull-out testing damage the structure and are time-consuming and labor-intensive; conventional non-destructive technologies such as ultrasound and rebound hammers lack sufficient accuracy and cannot achieve spatial visualization of strength distribution, making it difficult to precisely identify local defects such as cracks, spalling, and seepage.

Operational data from cross-river tunnels such as Shanghai Metro Line 12 and Qingdao Metro Line 8 show that long-term groundwater erosion and geological stress can easily lead to lining strength degradation. Without timely and accurate monitoring, structural risks may occur. Therefore, developing non-contact, automated, high-precision detection technology has become an urgent need for underground engineering maintenance.

CHNSpec Hyperspectral Camera Provides Core Empowerment, Achieving Three Technical Innovations

The CHNSpec FigSpec FS-27 hyperspectral camera selected by the research team became the key hardware support behind this technological breakthrough. The camera is equipped with an InGaAs sensor capable of capturing 1024 consecutive spectral bands in the 900–1700 nm near-infrared spectral range, achieving a spectral resolution of 6.5 nm. It can precisely capture subtle spectral feature differences on the surface of concrete, providing abundant physicochemical information for strength assessment. Combined with halogen light sources and a standard reflectance calibration panel, the device can still stably obtain 1280×1280 high-definition spectral images in dim and complex tunnel environments, meeting the stringent requirements of on-site detection.

Based on the high-dimensional spectral data obtained by the camera, the team innovatively developed two DNN regression models. Among them, the segmented-spectrum deep neural regressor (SegS_DNR) performed the best: through spatial–spectral feature fusion using a 5×5 pixel window, the model achieved a testing-set coefficient of determination (Rp²) of 0.925 and a residual predictive deviation (RPD) of 5.28, far surpassing traditional partial least-squares regression (PLSR) and random forest (RFR) models. Meanwhile, it generated 2D compressive-strength distribution heatmaps that visually present strength differences across different lining regions, achieving the dual goals of “quantitative detection + visual presentation.”

In the on-site verification conducted in the cross-Huangpu River tunnel of Shanghai Metro Line 12, the technology not only accurately measured the compressive strength of normal sections (average about 47.6 MPa), but also clearly identified strength degradation features in areas with cracks, spalling, and seepage—defect region strength averages dropped to 33–37 MPa, highly consistent with actual engineering detection results, with recognition accuracy comparable to senior engineers.

Empowering Intelligent Maintenance and Leading a New Paradigm in Underground Engineering Detection

In this technological breakthrough, the high stability and high spectral resolution advantages of the CHNSpec FigSpec FS-27 hyperspectral camera were fully validated. Its collaborative application with deep learning algorithms has completely transformed the traditional detection model for underground lining structures. The technology requires no contact with the structure’s surface, can quickly complete large-area scanning detection, and the generated strength distribution maps can directly support maintenance decision-making, significantly improving the scientific nature and efficiency of underground engineering lifecycle management.

At present, the technology has completed pilot applications in underground engineering projects under complex geological conditions such as cross-river and subsea tunnels. In the future, it may be integrated with robotic inspection, IoT (Internet of Things), and BIM (Building Information Modeling) technologies to build an intelligent monitoring system for underground infrastructure. According to CHNSpec's relevant spokesperson, the company will continue optimizing the engineering adaptability of hyperspectral imaging equipment and provide more targeted technical solutions for non-destructive testing needs in transportation, civil engineering, geology, and other fields.

The implementation of this research achievement not only provides a new technical pathway for underground engineering safety assurance, but also demonstrates the core competitiveness of domestically produced hyperspectral equipment in high-end research and engineering applications, injecting new momentum into the safe and sustainable operation of China’s underground infrastructure.

FS-27 is a shortwave near-infrared imaging hyperspectral camera in the FigSpec® FS2X series. It adopts transmission grating spectral separation technology and features high resolution and high stability, making it suitable for professional spectral imaging analysis across multiple fields.

Core Product Features:
Excellent spectral performance: covering the 900–1700 nm shortwave near-infrared range, with wavelength resolution better than 6 nm, and supporting 1024 spectral channels. Reliable imaging quality: image resolution reaches 1280×1280, dynamic range is 12 bits, SNR is 600:1, and stray light level is only 0.5%. Convenient and efficient operation: built-in scanning imaging and auxiliary view camera can monitor the shooting area in real time; supports multiple regional ROI functions; imaging speed is 10 seconds per scan.