Infrared measurement system with switchable filter for vision measurement

Abstract

The invention discloses an infrared measurement system with switchable filters for vision measurement, and relates to the technical field of computer vision measurement. The multispectral collaborative sensing hardware architecture comprises a multi-mode imaging unit and a switchable filter module, the global shutter industrial camera is installed in the shell, the shell is located at the lens end and surrounds and fixes an infrared light supplementing lamp, the steering engine is fixed to the top of the shell and is connected with a driving support, and the support is located in front of the lens end and carries an optical filter. The optical filter is integrated with an infrared optical filter and a visible light channel; the intelligent control and decision-making unit realizes microsecond-level synchronization of filter switching, light source modulation and camera exposure; a multi-modal data processing algorithm is combined with visible light high-frequency texture and infrared low-frequency heat distribution characteristics to realize background interference suppression. A mechanical switchable optical filter is adopted, an infrared optical filter and a visible light channel are integrated, a single industrial camera can synchronously obtain visible light and infrared information, and the industrial camera has the advantages of being low in cost, high in efficiency and high in anti-interference capacity.

Description

Technical Field

[0001] The invention relates to the technical field of computer vision measurement, in particular to an infrared measurement system with switchable filters for vision measurement. Background Art

[0002] In visual measurement systems, multispectral information acquisition is a crucial foundation for environmental perception and target recognition. Spectral adaptability in dynamic environments directly impacts the accuracy and reliability of the measurement system. Traditional infrared measurement systems often use fixed filters, which can lead to significant signal-to-noise ratio drops and feature information loss in scenarios with drastic lighting changes or complex spectral interference. This is particularly true in areas requiring 24 / 7 operation, such as industrial inspection and security monitoring. Existing infrared measurement systems struggle to balance high sensitivity with strong anti-interference capabilities.

[0003] Studies have shown that cross-interference between the near-infrared band in the solar spectrum and artificial light sources can cause traditional single-filter systems to have a feature misjudgment rate of more than 17.6%. When dealing with rapid changes in light intensity such as the transition from dawn to dusk, the dynamic response time of fixed-band filters exceeds 300ms, seriously restricting real-time measurement performance.

[0004] Furthermore, current advanced solutions generally rely on multispectral sensor arrays or computational imaging algorithms to improve environmental adaptability. However, the former leads to a surge in hardware costs, approximately 2.8 times that of single-sensor solutions, while the latter introduces up to 73% additional computational load. While tunable filters based on electrically controlled liquid crystals can be electronically switched, they suffer from a transmission efficiency loss of over 35% and a response latency of 8-12ms, making them difficult to meet the demands of high-speed motion scenarios.

[0005] In view of this, the present invention innovatively designs a mechanical switchable filter driven by a servo, integrating an 850nm~1000nm wide-bandpass infrared filter and a visible light channel, and realizes 200ms-level switching of the filter through high-precision angle control, so that a single ordinary industrial camera can simultaneously obtain visible light texture and infrared thermal feature information. Combined with the background suppression algorithm based on frequency domain fusion, the target signal-to-noise ratio is improved to 42dB in complex industrial scenarios, reducing the misjudgment rate by 19.7% compared with the traditional single-band infrared solution. The designed multi-spectral collaborative perception hardware architecture effectively solves the difficult problem of balancing cost, speed and accuracy in multi-spectral vision systems, provides a cost-effective solution for dynamic target recognition, and has broad application prospects in industrial automation detection, intelligent security, autonomous driving and robot navigation. Summary of the Invention

[0006] To address the problems of high hardware cost, heavy computational load, low light transmission efficiency, and response delay in traditional infrared measurement, the present invention provides an infrared measurement system with a switchable filter for visual measurement. The system uses a mechanically switchable filter and integrates an infrared filter with a visible light channel, enabling a single industrial camera to simultaneously acquire visible light and infrared information. This system has the characteristics of low cost, high efficiency, and strong anti-interference capability.

[0007] To achieve the above-mentioned object, the present invention adopts the following technical solutions: an infrared measurement system with switchable filters for visual measurement, comprising a multi-spectral collaborative perception hardware architecture, an intelligent control and decision-making unit, and a multimodal data processing algorithm;

[0008] The multispectral collaborative sensing hardware architecture includes a multimodal imaging unit and a switchable filter module. The multimodal imaging unit is equipped with a global shutter industrial camera and is installed inside a housing. The housing is located at the lens end and is surrounded by multiple infrared fill lights. The servo of the switchable filter module is fixed to the top of the housing. The output end of the servo is connected to the drive bracket. The bracket is located in front of the lens end of the global shutter industrial camera and is provided with a hole and equipped with a filter. The filter integrates an 850nm~1000nm wide-bandpass infrared filter and a visible light channel. The servo can achieve 200ms-level mechanical switching of the filter, realizing the synchronous capture of visible light texture and infrared thermal characteristics.

[0009] The intelligent control and decision-making unit uses an embedded chip as the control core and is based on an FPGA+ARM heterogeneous computing architecture to achieve microsecond-level synchronization of filter switching, light source modulation, and camera exposure, ensuring the spatiotemporal consistency of multispectral data.

[0010] The multimodal data processing algorithm includes a frequency domain fusion algorithm and a background suppression algorithm, which combines the high-frequency texture of visible light and the low-frequency thermal distribution characteristics of infrared light to achieve a background interference suppression rate of >85%.

[0011] Furthermore, a wireless charging interface is integrated on the top of the shell, and a battery, a control unit and a battery management module are also installed inside the shell. The wireless charging interface is connected to the battery to form a wireless charging module. The battery management module is provided with a voltage divider module to adaptively supply the battery's electrical energy to the power-consuming components. The control unit is equipped with an intelligent control and decision-making unit for overall control of the system.

[0012] Furthermore, the light transmittance of the filter is above 92%, and the wavelength of the infrared fill light is 940nm.

[0013] Furthermore, the intelligent control and decision-making unit adopts the Xilinx Zynq UltraScale+ MPSoC chip that integrates FPGA and ARM. The FPGA side realizes the angle control of the servo, PWM modulation of the infrared fill light and camera trigger signal generation, with a timing synchronization accuracy of less than 5μs. The ARM side runs the Linux real-time kernel, deploys an adaptive exposure algorithm and an abnormality diagnosis module, dynamically adjusts the fill light intensity and exposure time, and monitors the hardware status in real time.

[0014] Furthermore, the multimodal data processing algorithm consists of a multi-source data registration module, a frequency domain feature decoupling module, a dynamic weight fusion module, a background interference suppression module and a multi-scale reconstruction optimization module. The multi-source data registration module realizes sub-pixel alignment of visible light and infrared images. The frequency domain feature decoupling module separates high-frequency texture and low-frequency thermal features through improved wavelet packet transform. The dynamic weight fusion module dynamically adjusts the fusion weight according to the ambient light intensity and local statistical characteristics of the image. The background interference suppression module combines frequency domain notching and time domain Gaussian mixture modeling to suppress complex interference. The multi-scale reconstruction optimization module improves the signal-to-noise ratio of the target area through inverse wavelet transform and image enhancement optimization.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. This invention achieves a breakthrough in multispectral sensing capabilities. By using dual-modal data fusion, a single industrial camera can simultaneously acquire visible light and infrared information, reducing hardware costs by 60% compared to traditional multi-camera solutions. It has dynamic spectral adaptability, supports different operating modes, and can achieve 99.9% filter switching reliability in an environment of -30°C to 70°C.

[0017] 2. This invention is robust in extreme environments. Through the mechanical filter + digital noise reduction algorithm, it has strong anti-interference ability and maintains extremely high detection accuracy under strong light interference, which is 23% higher than the detection accuracy of traditional solutions. It can achieve all-weather autonomous operation and adaptive operation in the illumination range of 5lux to 150,000lux, without the need for manual parameter adjustment.

[0018] 3. The present invention has improved both efficiency and accuracy. The single-piece detection takes 280ms and supports industrial-grade high-speed detection of more than 12,000 pieces per hour. It can detect surface cracks of 0.1mm and thermal anomalies of 0.5℃ temperature difference, with a missed detection rate of <0.8%. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the overall framework of the infrared measurement system of the present invention;

[0020] Figure 2 This is a schematic diagram of the overall structure of the multi-spectral collaborative sensing hardware architecture in the present invention;

[0021] Figure 3 It is a rear-view structural diagram of the multi-spectral collaborative sensing hardware architecture in the present invention;

[0022] Figure 4 This is a cross-sectional view of the internal structure of the multi-spectral collaborative sensing hardware architecture of the present invention;

[0023] Figure 5 Schematic diagram of the electrical structure of the infrared measurement system of the present invention;

[0024] Figure 6 It is a schematic diagram of the algorithm structure of the infrared measurement system of the present invention;

[0025] Figure 7 It is an algorithm flow chart of the infrared measurement system of the present invention.

[0026] In the figure: 1. Servo; 2. Multimodal imaging unit; 3. Bracket; 4. Housing; 5. Wireless charging port; 6. Filter; 7. Battery; 8. Control unit; 9. Battery management module; 10. Infrared fill light. DETAILED DESCRIPTION

[0027] The technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] like Figures 1 to 7 As shown, an infrared measurement system with switchable filters for visual measurement, the overall framework of which is combined with Figure 1 As shown in the figure, it is divided into three parts: multi-spectral collaborative perception hardware architecture, intelligent control and decision-making unit, and multimodal data processing algorithm.

[0029] The multispectral collaborative perception hardware architecture includes a multimodal imaging unit 2 and a switchable filter module. The switchable filter module uses a servo 1 to drive a bracket 3 equipped with a filter 6. The filter 6 integrates an 850nm~1000nm wide-bandpass infrared filter and a visible light channel. The high-precision angle control of the servo 1 realizes the 200ms-level mechanical switching of the filter 6, and the light transmittance efficiency of the filter 6 is above 92%. The multimodal imaging unit 2 is equipped with a global shutter industrial camera and cooperates with a programmable infrared fill light 10 to realize the synchronous capture of visible light texture and infrared thermal characteristics.

[0030] Specific, combined Figures 2 to 4As shown, the global shutter industrial camera of the multimodal imaging unit 2 is installed in the middle position inside the housing 4. The housing 4 is located at the lens end and is surrounded by multiple infrared fill lights 10. The wavelength of the infrared fill lights 10 is 940nm. The servo 1 of the switchable filter module is fixed on the top of the housing 4. The output end of the servo 1 is connected to the drive bracket 3. The bracket 3 is located in front of the lens end of the global shutter industrial camera and is provided with a hole and equipped with a filter 6. The filter 6 integrates an 850nm~1000nm wide-bandpass infrared filter and a visible light channel. The servo 1 controls the deflection of the bracket 3 to achieve 200ms-level mechanical switching of the filter 6. In addition, a wireless charging interface 5 is integrated on the top of the shell 4, and a battery 7, a control unit 8 and a battery management module 9 are also installed inside the shell 4. The wireless charging interface 5 is connected to the battery 7 to form a wireless charging module for charging tasks. The battery management module 9 is provided with a voltage divider module to adaptively supply the electrical energy of the battery 7 to the power-consuming components. The control unit 8 is equipped with an intelligent control and decision-making unit for the overall control of the system. The control combination of the electrical structure is shown in Figure 5.

[0031] The multispectral collaborative sensing hardware architecture is designed based on three core principles: ensuring spectral-spatial consistency, optimizing anti-interference, and balancing cost and performance. It aims to meet the demands for high precision, robustness, and cost-effectiveness in industrial inspection. The architecture utilizes a mechanical filter wheel structure with optical axis offset controlled to within 0.01°. Combined with a global shutter CMOS camera, this ensures subpixel alignment of multimodal data. A wide-bandpass infrared filter covers characteristic absorption peaks while avoiding interfering wavelengths. Combined with a ring-shaped programmable infrared fill light (10), this reduces reflective interference by 40% and significantly improves anti-interference capabilities. Time-sharing multiplexing of a single camera replaces a multi-camera system, reducing hardware costs by 60% and simplifying the calibration process. The servo drive solution achieves a light transmission efficiency exceeding 92%, is polarization-independent, and achieves 200ms switching performance at one-third the cost, meeting production line requirements and offering high cost-performance.

[0032] The workflow involves controlling the servo 1 via a field-programmable gate array (FPGA) to mechanically switch the filter 6 in 200ms. Switching to infrared mode also synchronizes the activation of the infrared fill light 10, providing a dual-wavelength light source (850nm-1000nm / 940nm). A global shutter CMOS camera (120fps) is triggered synchronously with the filter switching and light source modulation hardware, achieving a timing alignment accuracy of <5μs. In visible light mode, texture information is captured at the 0.05mm / pixel level, while in infrared mode, thermal distribution characteristics are acquired through a 0.3mm surface layer, enabling multispectral collaborative sensing. The collected image data is processed and analyzed, combining the characteristics of different spectral channels to enable comprehensive inspection and analysis, such as electronic component defect detection and metal part inspection.

[0033] In summary, the multi-spectral collaborative sensing hardware architecture of the present invention achieves an organic combination of efficient spectral switching, precise imaging, strong anti-interference ability and high cost performance through mechanical-optical-electronic collaborative design. It is highly innovative and has high practical value, and can effectively meet the diverse needs of industrial detection.

[0034] The intelligent control and decision-making unit mainly uses an embedded chip as a high-speed control core to achieve microsecond-level synchronization of filter switching, light source modulation and camera exposure, ensuring the spatiotemporal consistency of multispectral data.

[0035] The intelligent control and decision-making unit is implemented using a Xilinx Zynq UltraScale+ MPSoC chip based on an FPGA+ARM heterogeneous computing architecture. This chip integrates high-performance FPGA logic resources and a multi-core ARM processor, creating a multi-level control architecture that meets the high-precision, real-time, and reliability requirements of multispectral collaborative sensing systems in complex industrial environments. At the hardware acceleration layer, a dedicated state machine written in Verilog on the FPGA implements angle control of servo 1 (with 0.1° resolution), PWM modulation of infrared fill light 10 (with a 10kHz frequency), and camera trigger signal generation for multimodal imaging unit 2. Timing synchronization accuracy is less than 5μs, ensuring deterministic real-time response for critical control tasks. At the decision optimization layer, an ARM Cortex-A53 processor running the Linux real-time kernel deploys an adaptive exposure algorithm (dynamically adjusting fill light intensity and exposure time based on a PID controller) and an anomaly diagnosis module (real-time monitoring of filter wear, light source attenuation, and other conditions). Dynamic adjustment of fill light intensity and exposure time, along with real-time hardware status monitoring, ensure long-term stable system operation. At the communication interface layer, an EtherCAT slave controller is integrated to support <1ms-level command interaction with PLCs, robotic arms and other equipment, accurately matching the production line rhythm and meeting the real-time requirements of large-scale industrial production scenarios.

[0036] The workflow includes environmental perception triggering, mode decision-making, and hardware synchronization execution. Light intensity sensors and temperature and humidity modules collect environmental data, generating control instructions based on preset strategies or AI models. Hardware synchronization ensures the integrity and consistency of multispectral image data. Design features include nanosecond-level response guarantees, adaptive fault tolerance, energy-efficient design, and industrial-grade reliability. The system has passed IEC 61000-4-4 and MIL-STD-810G vibration testing, ensuring 24 / 7 operation in harsh industrial environments.

[0037] In summary, the intelligent control and decision-making unit of the present invention achieves an organic combination of high real-time performance, high precision, high reliability and low power consumption through the deep collaboration of hardware logic and software algorithms. It has significant innovation and practicality, and can effectively meet the needs of multi-spectral collaborative perception systems in complex industrial environments.

[0038] The multimodal data processing algorithm is mainly a frequency domain fusion algorithm and a background suppression algorithm, which combines the high-frequency texture of visible light and the low-frequency thermal distribution characteristics of infrared light to achieve a background interference suppression rate of >85% in complex scenes.

[0039] The multimodal data processing algorithm consists of a multi-source data registration module, a frequency domain feature decoupling module, a dynamic weight fusion module, a background interference suppression module and a multi-scale reconstruction optimization module. Figure 6 As shown, each module works closely together to form a complete process. First, the multi-source data registration module achieves subpixel alignment of visible light and infrared images, laying the foundation for subsequent processing. The frequency domain feature decoupling module uses a modified wavelet packet transform to separate high-frequency texture and low-frequency thermal features. The dynamic weight fusion module dynamically adjusts fusion weights based on ambient light intensity and local image statistics to optimize the fusion effect. The background interference suppression module combines frequency domain notching with time domain Gaussian mixture modeling to accurately suppress complex interference. The multiscale reconstruction optimization module uses inverse wavelet transform and image enhancement optimization to improve the signal-to-noise ratio of the target area. This algorithm is based on the two core concepts of physical interference feature separation and multimodal information complementarity. It eliminates periodic noise through frequency domain notch filtering and leverages the complementary characteristics of visible light and infrared images to overcome the limitations of single-modality perception. The dynamic weight mechanism adjusts the weight coefficients in real time based on ambient light sensor data and local image statistics, achieving stable output under all working conditions, from bright light to dark fields.

[0040] Combine Figure 7 As shown in the figure, the algorithm workflow involves the following: First, the registered visible and infrared images are input and radiometrically corrected to optimize image contrast and brightness. Next, the corrected images undergo a four-layer wavelet packet decomposition to extract high-frequency subbands (containing texture and detail information) and low-frequency subbands (containing thermal distribution features). At the highest decomposition layer, frequency-domain notch filtering is used to precisely remove periodic background noise without damaging target features. Subsequently, feature fusion is performed: weighted fusion is applied to the low-frequency components, with the infrared image as the primary focus to preserve thermal distribution features, while a pulse-coupled neural network (PCNN) is used to optimize the high-frequency components to enhance texture details in the visible image. The fused coefficients at each layer are then reconstructed using an inverse wavelet transform to create a preliminary fused image. Foreground extraction is performed on this preliminary fused image using a temporal Gaussian mixture model to isolate the target image after background suppression. Finally, the target image undergoes CLAHE enhancement and edge-preserving filtering to improve the visual quality and signal-to-noise ratio of the target region, resulting in the final detection result.

[0041] In summary, the multimodal data processing algorithm of the present invention has significant characteristics and advantages. It innovatively adopts a cascade architecture of frequency domain feature optimization and time domain dynamic modeling to effectively separate background interference and target features. In metal surface detection, the background suppression rate is increased to 89%, which is 31% higher than the traditional spatial domain method, significantly enhancing the accuracy of target detection. In addition, the light intensity sensor provides real-time feedback on the ambient light intensity, dynamically adjusting the fusion weight of visible light and infrared images, and the image quality is improved even when the light intensity changes drastically (ΔL>10 4 lux / s), the peak signal-to-noise ratio (PSNR) fluctuation of the fused image remains less than 0.5dB, ensuring the algorithm's stability and robustness under complex lighting conditions. Furthermore, integrating computationally intensive operations such as wavelet transforms and PCNN optimization into the FPGA logic unit significantly improves the algorithm's operational efficiency, enabling real-time processing at 30 frames per second at 1080p resolution with end-to-end latency of less than 33 milliseconds, meeting the stringent real-time requirements of industrial scenarios.

[0042] The specific implementation steps of the system of the present invention include:

[0043] Step 1: Install the system of the present invention on the target to be measured or on the measuring platform, ensuring that the relative position between the system and the target meets the measurement requirements;

[0044] Step 2: Start the measurement system through the main switch, check the connection status of each system component, and confirm that the system is operating normally;

[0045] Step 3: Based on the current ambient light conditions and measurement requirements, adjust the switchable filter module through software, select the appropriate wide-bandpass infrared filter or visible light channel, and adjust the infrared fill light intensity to ensure that the camera can clearly identify the target features and obtain high-quality image data.

[0046] The present invention can be applied to material defect detection scenarios on industrial production lines. In the quality inspection of heat sinks for electronic products, the system is deployed above a high-speed conveyor belt. The bandpass filter wheel (850nm~1000nm) driven by a servo works in conjunction with a 2-megapixel industrial camera to achieve active spectral switching 5 times per second. In visible light mode (the filter is switched to the transparent channel), the system captures surface scratches, oxidation and other texture defects on the heat sink with a resolution of 0.05mm / pixel. When switching to the 1000nm infrared band, the array infrared fill light is simultaneously activated to penetrate the surface oil and detect hidden defects such as internal structural cracks, with a thermal feature sensitivity of 0.3°C. Combined with a background suppression algorithm based on multi-scale frequency domain fusion, a defect detection rate of 97.6% can be achieved in a 120dB dynamic lighting environment, an improvement of 23.8% compared to traditional single-band detection solutions, while keeping the false positive rate below 1.2%. According to actual tests, the system can still maintain a stable recognition rate of 94.3% under conditions of strong sunlight interference (>80,000 lux) and low illumination in a dark room (<5 lux). The single-piece inspection takes only 280ms, which is 15 times more efficient than manual visual inspection.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other configurations without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations coming within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0048] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. An infrared measurement system with switchable filters for visual measurement, characterized by: Including multi-spectral collaborative perception hardware architecture, intelligent control and decision-making unit, and multimodal data processing algorithm; The multi-spectral collaborative sensing hardware architecture includes a multi-modal imaging unit (2) and a switchable filter module, wherein the multi-modal imaging unit (2) is equipped with a global shutter industrial camera and is installed inside a housing (4), wherein the housing (4) is located at the lens end and is surrounded by a plurality of infrared fill lights (10), and the servo (1) of the switchable filter module is fixed on the top of the housing (4), wherein the output end of the servo (1) is connected to a driving bracket (3), wherein the bracket (3) is located in front of the lens end of the global shutter industrial camera and is provided with a hole and equipped with a filter (6), wherein the filter (6) integrates an 850nm~1000nm wide-bandpass infrared filter and a visible light channel, and the servo (1) can realize 200ms-level mechanical switching of the filter (6), thereby realizing synchronous capture of visible light texture and infrared thermal features; The intelligent control and decision-making unit uses an embedded chip as the control core and is based on an FPGA+ARM heterogeneous computing architecture to achieve microsecond-level synchronization of filter switching, light source modulation, and camera exposure, ensuring the spatiotemporal consistency of multispectral data. The multimodal data processing algorithm includes a frequency domain fusion algorithm and a background suppression algorithm, which combines the high-frequency texture of visible light and the low-frequency thermal distribution characteristics of infrared light to achieve a background interference suppression rate of >85%.

2. The infrared measurement system with switchable filters for visual measurement according to claim 1, characterized in that: A wireless charging interface (5) is integrated and installed on the top of the housing (4), and a battery (7), a control unit (8) and a battery management module (9) are also mounted inside the housing (4). The wireless charging interface (5) is connected to the battery (7) to form a wireless charging module. The battery management module (9) is provided with a voltage divider module to adaptively supply the electric energy of the battery (7) to the power-consuming components. The control unit (8) is equipped with an intelligent control and decision-making unit for overall control of the system.

3. The infrared measurement system with switchable filters for visual measurement according to claim 1 or 2, characterized in that: The light transmittance efficiency of the filter (6) is above 92%, and the wavelength of the infrared fill light (10) is 940 nm.

4. The infrared measurement system with switchable filters for visual measurement according to claim 1, characterized in that: The intelligent control and decision-making unit adopts a Xilinx Zynq UltraScale+MPSoC chip integrating FPGA and ARM. The FPGA side realizes the angle control of the servo (1), the PWM modulation of the infrared fill light (10) and the generation of the camera trigger signal. The timing synchronization accuracy is less than 5μs. The ARM side runs the Linux real-time kernel, deploys the adaptive exposure algorithm and the abnormality diagnosis module, dynamically adjusts the fill light intensity and exposure time, and monitors the hardware status in real time.

5. The infrared measurement system with switchable filters for visual measurement according to claim 1, characterized in that: The multimodal data processing algorithm consists of a multi-source data registration module, a frequency domain feature decoupling module, a dynamic weight fusion module, a background interference suppression module and a multi-scale reconstruction optimization module. The multi-source data registration module realizes sub-pixel alignment of visible light and infrared images. The frequency domain feature decoupling module separates high-frequency texture and low-frequency thermal features through improved wavelet packet transform. The dynamic weight fusion module dynamically adjusts the fusion weight according to the ambient light intensity and local statistical characteristics of the image. The background interference suppression module combines frequency domain notching and time domain Gaussian mixture modeling to suppress complex interference. The multi-scale reconstruction optimization module improves the signal-to-noise ratio of the target area through inverse wavelet transform and image enhancement optimization.

Images