Discharge ultraviolet panoramic imaging system and method capable of expanding wave band

Abstract

The invention provides a discharge ultraviolet panoramic imaging system and method capable of expanding wave bands, and belongs to the technical field of insulation state monitoring of high-voltage equipment, the system comprises an upper computer, a wavelength tunable filtering device, a multi-view-field panoramic imaging module, a self-adaptive fusion module and a panoramic splicing module, the control module is used for controlling transmission wavelength adjustment, servo turntable movement and data processing; the wavelength tunable filtering device comprises a liquid crystal tunable filter and is used for executing transmission wavelength adjustment; the multi-view-field panoramic imaging module is used for synchronously collecting ultraviolet and visible light images of a target wavelength through rotary scanning; the self-adaptive fusion module is used for dynamically fusing the ultraviolet image and the visible light image; and the panoramic stitching module is used for performing feature matching and panoramic stitching on the dynamically fused image.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-voltage equipment insulation status monitoring, and in particular relates to an ultraviolet panoramic imaging system and method with an extendable wavelength band. Background Art

[0002] Existing UV imaging technology has the following defects in power equipment discharge detection:

[0003] ① Traditional UV imaging systems only work in the solar-blind UV band (240-280nm) and cannot adapt to the spectral characteristics of different discharge types (such as the visible light component of corona discharge at 300-400nm).

[0004] ②The field of view of a single lens is usually less than 60°, and the shooting angle needs to be adjusted frequently, making it difficult to cover panoramic monitoring of large equipment.

[0005] ③ At night or in complex lighting environments, the UV signal signal-to-noise ratio (SNR) is lower than 30dB and is easily interfered by background noise.

[0006] For example, but not limited to, prior art document 1 (CN119860842A) discloses a combined filter multispectral imaging solution in the field of multispectral imaging technology. By setting a light-transmitting hole in the center of the multispectral filter array and embedding a liquid crystal tunable filter, the wide-range spectroscopic capability of the filter array and the narrow-band tuning characteristics of the liquid crystal tunable filter are combined to achieve wide spectrum coverage and high-precision spectrum adjustment. However, it is not optimized for the ultraviolet band and cannot adapt to the discharge type detection requirements of power equipment. It relies on a static combination structure, resulting in insufficient adaptability to dynamic scenes and limited imaging quality under complex backgrounds.

[0007] Prior art document 2 (CN113865707A) discloses a fusion solution of LCTF and ridge regression spectral models in the field of hyperspectral imaging technology. It improves the light flux through the LCTF structure of two-stage Lyot glass cascade, and constructs a system transfer matrix combined with a ridge regression algorithm to invert high-resolution spectral images from low-spectral resolution observation data. However, it is not optimized for ultraviolet band imaging. Due to structural limitations, the spectral transmittance is insufficient in the ultraviolet region, and it relies on the ridge regression algorithm with high computational complexity to invert hyperspectral data, which makes it difficult to meet real-time requirements and cannot achieve seamless coverage of large-scale scenes. Summary of the Invention

[0008] To address the shortcomings of existing UV imaging technology, this paper proposes a scalable UV panoramic imaging system and method for partial discharge detection in power equipment. By leveraging a wavelength-tunable filter system and multi-field panoramic imaging technology, this system and method achieves all-weather, multi-band coverage for partial discharge detection in power equipment. This system and method are suitable for UV monitoring of power equipment such as GIS equipment, transformer bushings, and insulators in locations like substations and transmission lines.

[0009] The present invention adopts the following technical solutions.

[0010] A first aspect of the present invention provides a discharge ultraviolet panoramic imaging system with an extendable wavelength band, the system comprising:

[0011] Host computer, wavelength tunable filter device, multi-field panoramic imaging module, adaptive fusion module and panoramic stitching module, wherein:

[0012] The host computer is used to control the transmission wavelength adjustment, servo turntable movement and data processing;

[0013] A wavelength tunable filter device, comprising a liquid crystal tunable filter for performing transmission wavelength adjustment;

[0014] Multi-field panoramic imaging module, used to synchronously capture ultraviolet and visible light images of target wavelengths through rotational scanning;

[0015] Adaptive fusion module for dynamically fusing UV images with visible light images;

[0016] The panoramic stitching module is used to perform feature matching and panoramic stitching on the dynamically fused images.

[0017] Optionally, the liquid crystal tunable filter covers a wavelength band of 240-400 nm, and the transmission wavelength is switched in a preset step size.

[0018] Optionally,

[0019] The multi-field panoramic imaging module includes:

[0020] The lens array consists of a set number of wide-angle lenses;

[0021] Servo turntable, used to drive the lens array to rotate and scan at full angles at a set speed;

[0022] A spectrometer for splitting incident light into an ultraviolet light path and a visible light path;

[0023] An ultraviolet imaging unit, disposed in the ultraviolet light path, for generating an ultraviolet image;

[0024] A visible light imaging unit, disposed in the visible light path, for generating a visible light image;

[0025] FPGA synchronization controller, used to synchronously control the exposure timing of the ultraviolet and visible light image acquisition units.

[0026] A second aspect of the present invention provides a method for panoramic ultraviolet imaging of discharge with an extendable wavelength band, based on the method for panoramic ultraviolet imaging of discharge with an extendable wavelength band described in the first aspect of the present invention, comprising the following steps:

[0027] Select a target wavelength according to the discharge type, and control the liquid crystal tunable filter to switch to the target wavelength in a set step size;

[0028] After locking the target wavelength, the servo turntable is triggered to drive the lens array to rotate, synchronously capturing multi-field visible light images and ultraviolet images;

[0029] The multi-field fusion image is generated by dynamically fusing the ultraviolet image and the visible light image of the multi-field through the adaptive fusion algorithm;

[0030] Based on the fusion images of multiple fields of view, the SIFT algorithm is used to perform panoramic stitching of the multi-field images.

[0031] Optionally, the dynamically fusing the ultraviolet image and the visible light image by using an adaptive fusion algorithm includes:

[0032] Dynamically normalize the radiation intensity of ultraviolet images;

[0033] Calculate the spectral similarity between the UV image and the visible light image;

[0034] Fusion weights are dynamically assigned according to the radiation intensity and spectral similarity to generate a fused image.

[0035] Optionally, the dynamically normalizing the ultraviolet image radiation intensity includes:

[0036] Setting a radiation intensity threshold of the ultraviolet image pixel, including setting it to the maximum value when the radiation intensity is greater than a first preset ratio of the maximum value, and setting it to 0 when the radiation intensity is less than a second preset ratio of the minimum value;

[0037] Selecting a normalization mode according to lighting conditions, including selecting global dynamic normalization or local dynamic normalization according to the rate of change of lighting;

[0038] The processed radiation intensity is converted to a value range from 0 to 1 through a linear mapping.

[0039] Optionally, calculating the spectral similarity between the ultraviolet image and the visible light image includes:

[0040] Extracting the radiation intensity data of ultraviolet image pixels and visible light image pixels;

[0041] The spectral similarity value of two images is calculated by normalized cross-correlation.

[0042] Optionally, dynamically allocating fusion weights according to radiation intensity and spectrum similarity includes:

[0043] The UV image weight is determined by a preset linear combination of UV image pixel radiation intensity and spectral similarity;

[0044] The visible light image weight is 1 minus the ultraviolet image weight;

[0045] The weighted UV image and the visible light image are added together to generate a fused image.

[0046] Optionally, performing panoramic stitching on the multi-field images by using the SIFT algorithm includes:

[0047] Extract SIFT feature points from multi-field fusion images;

[0048] Based on the SIFT feature points, a spatial consistency check is performed to eliminate mismatched points;

[0049] Panoramic image stitching is performed based on the filtered feature points.

[0050] Optionally, performing a spatial consistency check to eliminate mismatched points further includes optimizing stitching parameters by a bundle adjustment method:

[0051] Based on the feature matching points retained after spatial consistency verification, the coordinates of each feature point in three-dimensional space are inverted and calculated;

[0052] Constructing an objective function to minimize the reprojection error between the three-dimensional space coordinates and their two-dimensional projection coordinates in adjacent lens images;

[0053] The rotation matrix and translation matrix of each lens are adjusted through iterative optimization to achieve a stitching error less than the preset pixel size.

[0054] Compared with the prior art, the beneficial effects of the present invention include at least:

[0055] 1. The wavelength-tunable filter device provided by the present invention breaks through the detection limitations of the traditional solar-blind ultraviolet band by dynamically adjusting the ultraviolet transmission band, achieving broad spectrum coverage and accurate identification of the characteristic spectra of various discharge types of power equipment;

[0056] 2. The multi-field panoramic imaging module provided by the present invention uses a servo turntable to drive the wide-angle lens array and synchronize the exposure of multiple lenses, achieving seamless and efficient scanning of the entire panorama of large-scale power equipment, significantly improving the dynamic monitoring adaptability of complex scenes;

[0057] 3. The adaptive fusion module provided by the present invention effectively suppresses background noise interference in complex lighting environments through dynamic normalization of ultraviolet radiation intensity and multi-spectral weight distribution mechanism, significantly improving the image signal-to-noise ratio of discharge characteristics;

[0058] 4. The panoramic stitching module provided by the present invention realizes high-precision, low-distortion panoramic stitching of multi-field images through feature point gradient screening and spatial consistency optimization, solving the mismatching problem of traditional algorithms in weak texture areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a schematic diagram of system structure connection provided according to an embodiment of the present invention;

[0060] Figure 2 1 is a schematic diagram of the positional relationship between a visible light camera and a spectrometer according to an embodiment of the present invention;

[0061] Figure 3 is a flow chart of an adaptive fusion algorithm provided in accordance with an embodiment of the present invention;

[0062] Figure 4 This is a flowchart of an improved SIFT panorama stitching process provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0064] The present invention provides a discharge ultraviolet panoramic imaging system with an extendable wavelength band in embodiment 1, such as Figure 1 As shown, the system includes:

[0065] Host computer, wavelength tunable filter device, multi-field panoramic imaging module, adaptive fusion module and panoramic stitching module, among which,

[0066] The host computer is used to control wavelength adjustment, servo turntable movement and data processing;

[0067] A wavelength tunable filter device, comprising a liquid crystal tunable filter (LCTF), covering the 240-400 nm band, and switching the transmission wavelength in a set step size, preferably but not limited to 5 nm;

[0068] Multi-field panoramic imaging module, including a servo turntable, lens array, spectrometer, ultraviolet imaging unit, visible light imaging unit and FPGA synchronization controller;

[0069] The lens array is composed of a set number of wide-angle lenses, which are mounted on a servo turntable. The lens array is preferably composed of at least four groups of wide-angle lenses with a field of view of 120°.

[0070] The servo turntable drives the lens to rotate at a set speed, and a single scan covers 360°. The set speed is preferably 15° / s;

[0071] The light splitting device splits the incident light into ultraviolet light and visible light;

[0072] An adaptive fusion module configured to fuse multi-band UV and visible images through dynamic weight allocation and employ an improved non-local means filter to suppress noise;

[0073] The panoramic stitching module is configured to achieve seamless stitching of multi-field images through grayscale gradient screening and bundle adjustment.

[0074] Preferably, the functions of the host computer include:

[0075] Controlling the band switching of the wavelength tunable filter device;

[0076] Issue servo turntable motion instructions;

[0077] Receive and process multi-field image data;

[0078] Generate a panoramic inspection report.

[0079] Preferably, the wavelength tunable filter device comprises an LCTF, which adjusts the transmission wavelength by voltage control, and the wavelength covers 240-400 nm, and is switched in a set step size to adapt to the spectral characteristics of different discharge types. The set wavelength is preferably 5 nm;

[0080] Further preferably, the thickness of the liquid crystal layer of the LCTF is 20 microns, the voltage adjustment range is 0-10V, and the transmission wavelength expands linearly with increasing voltage.

[0081] Specifically, LCTF is used. LCTF is based on the birefringence effect of liquid crystals. By applying voltage, the arrangement of liquid crystal molecules is changed, thereby adjusting the transmission wavelength. The relationship between the transmission wavelength λ and the voltage V can be expressed as:

[0082]

[0083] Where Δn(V) is the voltage-dependent birefringence, d is the liquid crystal layer thickness, and m is the diffraction order. Wavelength adjustment in 5nm steps is achieved through step voltage control. Experiments have shown that the half-width of the discharge spectrum of power equipment is typically in the range of 10-20nm. A 5nm step size can effectively distinguish the characteristic peaks of different discharge types while avoiding switching delays and data redundancy caused by too small a step size.

[0084] Based on the target wavelength, the host computer sends instructions to the LCTF controller; the LCTF then automatically switches to the target wavelength, completing the spectral screening. Finally, the spectral data is fed back to the image processing unit in real time. By dynamically adjusting the filter wavelength to adapt to the spectral characteristics of different discharge types, the system breaks through the limitations of traditional fixed wavelengths and significantly improves detection sensitivity and adaptability.

[0085] Preferably, the multi-field panoramic imaging module includes a servo turntable, a lens array, a spectrometer, an ultraviolet imaging unit, a visible light imaging unit and an FPGA synchronization controller;

[0086] The lens array is composed of at least four groups of wide-angle lenses with a field of view of 120°, the wide-angle lenses have a field of view of 120° and a focal length of 8 mm, and the lens array is arranged in a cross shape and mounted on the servo turntable;

[0087] The servo turntable drives the lens to rotate at a speed of 15° / s, and a single scan covers a 360° range, achieving a 360° panoramic scan to meet the seamless monitoring needs of a large range of power equipment;

[0088] like Figure 2 As shown, the spectroscopic device is used to split the incident light, reflect the ultraviolet light to the ultraviolet imaging unit, and transmit the visible light to the visible light imaging unit; illustratively, the spectroscopic device includes a dichroic mirror.

[0089] Further preferably, the ultraviolet imaging unit is used to collect multi-band ultraviolet images, and the ultraviolet image collection adopts the parallel light path ultraviolet imaging method. The discharged ultraviolet light is suppressed by the filter to suppress the interference of sunlight, and is focused to the ICCD photocathode window through the objective lens, and is converted into a visible light image through the image intensifier.

[0090] Further preferably, the visible light imaging unit includes a high-sensitivity CMOS camera, and the spectral response range covers the 400-700nm visible light band.

[0091] The FPGA synchronization controller is used to synchronously control the exposure time of the four groups of lenses, with a synchronization error of less than 1 millisecond, ensuring the time consistency of multi-field images.

[0092] Specifically, multi-lens exposure synchronization is achieved through FPGA hardware. Assuming the number of lenses is N and the exposure time is T, the synchronization error must satisfy:

[0093]

[0094] In this system, T = 20ms, N = 4, so Δt < 0.5ms, and the actual design error is < 1ms, ensuring the temporal consistency of multi-field images.

[0095] Further preferably, the visible light imaging unit further includes a synchronization control device for achieving exposure time synchronization by sharing the same FPGA controller with the visible light camera and the ultraviolet camera, thereby ensuring the temporal consistency of the multispectral image.

[0096] Preferably, the adaptive fusion module is used to fuse the multi-band ultraviolet image and the visible light image, and is configured to dynamically assign fusion weights according to ultraviolet radiation intensity and spectral similarity, and adopt an improved non-local mean filter to suppress background noise;

[0097] Further preferably, the improved non-local means filtering enhances the texture detail preservation capability of the discharge area by fusing the ultraviolet radiation intensity to correct the filtering weights.

[0098] Preferably, the panoramic stitching module is configured to achieve seamless stitching of multi-field-of-view images through grayscale gradient screening and spatial consistency check, with a stitching error of less than 0.1 pixel.

[0099] It should be noted that the host computer sends the target wavelength instruction to the wavelength tunable filter device, triggering the LCTF to switch to the specified band; after the wavelength is locked, the host computer synchronously controls the servo turntable to start rotation and the FPGA exposure synchronization of the multi-field panoramic imaging module to collect ultraviolet images and visible light images of the current band; the imaging data is transmitted to the adaptive fusion module in real time to complete the weight distribution and noise reduction processing of the ultraviolet and visible light images; the fused image is input into the panoramic stitching module for feature matching and panoramic stitching, and finally the discharge positioning result is fed back to the host computer to generate a detection report, forming a full closed-loop control link of "wavelength adjustment → image acquisition → fusion → stitching → output".

[0100] In Example 2, the present invention provides a method for discharge ultraviolet panoramic imaging with an extendable wavelength band. Based on the discharge ultraviolet panoramic imaging system with an extendable wavelength band described in Example 1, the method includes the following steps:

[0101] Step 1: Select a target wavelength according to the type of discharge to be detected, and switch to the target wavelength in 5 nm steps through a liquid crystal tunable filter.

[0102] Preferably, the step 1 comprises:

[0103] Step 1.1, select the target wavelength according to the type of discharge to be detected, including:

[0104] The target wavelength for corona discharge is 300-400nm, and the preferred characteristic wavelength is 365nm;

[0105] The target wavelength band for partial discharge is 240-280nm, and the preferred characteristic wavelength is 280nm;

[0106] The characteristic wavelength corresponding to the discharge type is used as the target wavelength;

[0107] Step 1.2: Switch to the target wavelength in 5 nm steps through LCTF.

[0108] The voltage adjustment range of the LCTF is 0-10V, and the transmission wavelength expands linearly with increasing voltage;

[0109] Each target wavelength stays for 5 seconds, and the wavelength deviation is checked in real time and automatically corrected.

[0110] Step 2: After locking the target wavelength, trigger the servo turntable to drive the lens array to rotate and synchronously capture multi-field visible light images and ultraviolet images.

[0111] Preferably, in step 2, the rotation speed of the servo turntable is 15° / s, the single scanning time is 24 seconds, and the multi-field image is associated with the current transmission wavelength label.

[0112] Preferably, in step 2, synchronously acquiring the visible light image and the ultraviolet image of the current band includes:

[0113] The incident light is divided into ultraviolet light and visible light by the spectrometer. The FPGA controller triggers the ultraviolet camera and the visible light camera to expose simultaneously. The exposure time is unified at 20ms, and the synchronization error is <1ms.

[0114] Specifically, the trigger relationship between step 2 and step 1 is: after the LCTF locks the target wavelength, the synchronous acquisition of the servo turntable and the lens array is automatically started.

[0115] Step 3: Apply the adaptive fusion algorithm to perform signal-to-noise ratio optimization fusion on the multi-field ultraviolet image and the visible light image to generate a multi-field fused image.

[0116] Preferably, if Figure 3 As shown, step 3 includes:

[0117] The adaptive fusion algorithm optimizes the signal-to-noise ratio by dynamically assigning weights to the UV image and the visible light image. It extracts the radiation intensity and spectral distribution characteristics of the discharge area and dynamically assigns fusion weights to enhance the signal-to-noise ratio (>45dB) in low-light environments. It also uses non-local mean filtering (NL-Means) to process the background area to suppress noise interference. The details are as follows:

[0118] The visible light image provides background texture information, and the ultraviolet image radiation intensity is I UV0 , directly collected by UV sensor, firstly preprocess the data to eliminate noise or transient interference, and set the upper and lower thresholds I UV0 >0.95I max Time Force I UV0 For I max , I UV0 <0.05I max Time Force I UV0 is 0). Then normalize it by linear mapping method:

[0119]

[0120] Among them I min is the minimum measurable value of the sensor, I max is the maximum saturation value of the sensor.

[0121] To improve adaptability, the system supports dynamic normalization, including:

[0122] Global dynamic normalization: Based on the I of all pixels in a single scan UV The maximum and minimum values ​​are mapped. This method is suitable for scenes with large changes in lighting conditions (such as day and night).

[0123] Local normalization: Calculate I separately for sub-regions in the image (such as the discharge core area) UV Maximum and minimum values ​​are used to avoid global extreme values ​​from masking local features.

[0124] The visible light image is I Vis , spectral similarity S spectral Calculated by normalized cross correlation (NCC):

[0125]

[0126] in It is the average value of the radiation intensity of all pixels in the ultraviolet image and the visible light image.

[0127] Fusion image I Fused It can be expressed as:

[0128] I Fused =w UV I UV +w Vis I Vis

[0129] Among them, the weight w UV and w Vis By the ultraviolet radiation intensity I UV and spectral similarity S spectralJoint decision. The specific formula is:

[0130]

[0131] w Vis =1-w UV

[0132] Where α and β are adjustment parameters that control the contribution ratio of radiation intensity and spectral similarity, respectively, and are optimized and determined in experimental data to maximize the signal-to-noise ratio; γ is a smoothing term that prevents the denominator from being zero.

[0133] The present invention uses non-local mean filtering (NL-Means) to smooth the background area. The traditional NL-Means weight calculation is:

[0134]

[0135] Where Patch(p) and Patch(q) are the neighborhood blocks centered on pixels p and q, respectively, and h is the filter control parameter (attenuation coefficient), which determines the decay rate of the weight as the distance increases.

[0136] The present invention introduces the ultraviolet radiation intensity weight, which is modified to:

[0137]

[0138] Where, I UV (p) and I UV (q) are the neighborhood blocks centered on pixels p and q, respectively, I UVmax The maximum radiation intensity of the region. While preserving texture details, it significantly suppresses noise in low radiation areas.

[0139] Step 4: Based on the fused images of multiple fields of view, high-precision panoramic stitching of the multi-field of view images is performed using an improved SIFT (Scale-Invariant Feature Transform) algorithm.

[0140] Preferably, if Figure 4 As shown, in step 4, the panoramic stitching technology of the improved SIFT algorithm includes:

[0141] Based on traditional SIFT feature matching, the grayscale gradient constraint of the discharge area and the spatial consistency verification mechanism are introduced to build a false matching elimination model.

[0142] First, SIFT feature points are extracted from the multi-field image; then the grayscale gradient variance of the discharge area around the feature points is calculated:

[0143]

[0144] Among them, G i is the gradient amplitude, the threshold is set to 0.1, and the mismatch in the low variance area (non-discharge area) is eliminated; σ 2 is the gray gradient variance, is the arithmetic mean of the gradient amplitudes of all pixels in the region, and N is the total number of pixels in the region;

[0145] The RANSAC algorithm is used to check spatial consistency. Four sets of matching points are randomly selected to calculate the homography matrix H. The inliers are selected iteratively, and the model with the most inliers is finally selected. Finally, the bundle adjustment method is used to optimize the stitching parameters and construct an objective function to minimize the reprojection error:

[0146]

[0147] Among them, R i , t i is the rotation and translation matrix of the i-th lens, X j is the 3D point coordinate, π(·) is the projection function, and xij is the 3D scene point X j 2D observation coordinates in camera view ij;

[0148] Achieve high-precision, low-distortion panoramic stitching of ultraviolet images with a stitching error of less than 0.1 pixel.

[0149] In order to more clearly introduce the outstanding essential features of the present invention and the significant progress it brings to the prior art, an application example of implementing the present invention is introduced below.

[0150] The following is a detailed description of an embodiment of the present invention, which specifically includes:

[0151] This method was implemented in a 500kV substation GIS equipment partial discharge monitoring. The equipment is 3.2 meters high and 8.5 meters in circumference.

[0152] Hardware layout and parameter settings:

[0153] Wavelength tunable filter system:

[0154] Using LCTF, the wavelength adjustment range is 240-400nm, the step size is 5nm, the liquid crystal layer thickness is d = 20μm, and the diffraction order is m = 1. The voltage adjustment range is 0-10V, corresponding to the transmission wavelength:

[0155]

[0156] Experimental measurements show that Δn(V) varies linearly in the range of 0.1-0.3.

[0157] Multi-field panoramic imaging system:

[0158] Four wide-angle lenses (120° field of view, 8mm focal length) are arranged in a cross pattern on a servo turntable, with a lens spacing of 50cm. The turntable rotates at 15° / s, with a single scan time of 24 seconds (360° / 15° / s). FPGA-controlled exposure synchronization error is less than 0.8ms (T = 20ms, N = 4).

[0159] Adaptive fusion algorithm parameters: α = 0.6, β = 0.4, γ = 0.01; non-local means filter window size 7 × 7.

[0160] Improved SIFT algorithm: The grayscale gradient variance threshold was set to 0.1, the number of RANSAC iterations was 1000, and the reprojection error after bundle adjustment optimization was <0.08 pixels.

[0161] Implementation steps:

[0162] Dynamic wavelength adjustment:

[0163] The host computer sends instructions to the LCTF controller to switch to the target wavelength (such as 280nm, 320nm, and 365nm) in sequence, staying in each band for 5 seconds.

[0164] Ultraviolet image data were collected for corona discharge (characteristic peak 365nm) and partial discharge (characteristic peak 280nm).

[0165] Synchronous acquisition of multi-field images:

[0166] The servo turntable rotates at a constant speed of 15° / s, and the FPGA triggers the synchronous exposure of four lens groups (exposure time 20ms), acquiring a 360° panoramic raw image (resolution 2048×2048) in a single scan. Experimental measurements show that the overlap area of ​​adjacent lens images is ≥30%, ensuring the feasibility of stitching.

[0167] Adaptive fusion and signal-to-noise ratio optimization:

[0168] Preprocess the 365nm band ultraviolet image: the radiation intensity threshold is set to I max =1000 (12-bit sensor), when I UV >950 is forced to 1000, I UV Forced to 0 when <50.

[0169] Local normalization is used, and the IUV of the discharge core region (ROI) is mapped to 0-1000, and the background region is mapped to 0-200.

[0170] The spectral similarity Sspectral (NCC value 0.85) is calculated, the fusion weight wUV = 0.72, and the signal-to-noise ratio of the generated fused image is improved to 48dB (the original UV image SNR = 32dB).

[0171] Improved SIFT panorama stitching:

[0172] The SIFT feature points of the four groups of lenses were extracted (an average of 5,000 per image), and the feature points with grayscale gradient variance < 0.1 (accounting for about 15%) were removed.

[0173] The optimal homography matrix is ​​calculated using the RANSAC algorithm, with an internal point ratio of 92%. The stitching error after bundle adjustment optimization is 0.07 pixels.

[0174] Effect verification:

[0175] Multi-band coverage: Successfully detected the characteristic peaks of 280nm (partial discharge) and 365nm (corona discharge), reducing the missed detection rate by 80% compared with traditional solar-blind UV systems (240-280nm).

[0176] Panoramic coverage efficiency: A single scan completes 360° monitoring in 24 seconds. Traditional single-lens systems require manual angle adjustment, which takes more than 5 minutes.

[0177] Improved signal-to-noise ratio: The SNR of the fused image is >45dB, and the standard deviation of background noise is reduced from 15.2 to 3.8. The visible light background information effectively suppresses the nighttime ambient noise.

[0178] The discharge ultraviolet panoramic imaging system and method with scalable wavelengths provided by the present invention have the following technical effects in application examples:

[0179] 1. Successfully detected the characteristic peaks of 280nm (partial discharge) and 365nm (corona discharge), reducing the missed detection rate by 80% compared with the traditional solar-blind UV system (240-280nm);

[0180] 2. A single scan takes 24 seconds to cover an 8.5-meter circumference GIS device, with a signal-to-noise ratio of 48dB and a stitching error of 0.07 pixels. The visible light background information effectively suppresses nighttime ambient noise.

[0181] 3. An improved SIFT algorithm is proposed, and the stitching error is <0.1 pixel through grayscale gradient variance screening and bundle adjustment method.

[0182] To address the problems of existing ultraviolet imaging technology, such as fixed band (only solar-blind ultraviolet), low panoramic coverage efficiency, poor signal-to-noise ratio, and insufficient real-time performance, the present invention provides a discharge ultraviolet panoramic imaging method and system with an expandable band, which has the following technical effects:

[0183] To address the issue of fixed wavelength bands and compatibility with discharge types, the present invention provides a wavelength-tunable filter device, including a liquid crystal tunable filter. This device achieves dynamic adjustment in 5nm steps within the UV wavelength range of 240-400nm, covering the characteristic wavelength of corona discharge at 365nm and the characteristic wavelength of partial discharge at 280nm, thus overcoming the limitations of conventional fixed solar-blind wavelength bands. In Example 3, the characteristic peaks at 280nm and 365nm were successfully detected, reducing the missed detection rate by 80% compared to conventional solar-blind UV systems.

[0184] To address the problems of low panoramic coverage efficiency and insufficient adaptability to dynamic scenes: the present invention provides a multi-field panoramic imaging module driven by a servo turntable, which includes multiple groups of wide-angle lens arrays with a field of view of 120 degrees and an FPGA synchronization controller. It achieves 360-degree seamless panoramic coverage in 24 seconds with a single scan, meeting the real-time monitoring needs of large-scale power equipment. The scanning efficiency is significantly improved compared to the traditional single-lens manual adjustment system.

[0185] To address the issues of low signal-to-noise ratio and complex background interference, this invention provides an adaptive fusion module, including a dynamic weight allocation algorithm based on UV radiation intensity and spectral similarity, and an improved non-local means filter. This module achieves efficient fusion of UV and visible light images and suppresses background noise. In Example 3, the signal-to-noise ratio of the fused image increased to 48 decibels, and the standard deviation of background noise decreased by 75%, effectively improving imaging quality in complex lighting environments.

[0186] To address the problems of high computational complexity, poor real-time performance, and insufficient panoramic stitching accuracy: the present invention provides FPGA hardware synchronization control to ensure that the multi-lens exposure synchronization error is less than 1 millisecond, and combines it with an improved SIFT stitching algorithm, including grayscale gradient variance screening and bundle adjustment optimization, to achieve high-precision, low-distortion panoramic stitching of multi-field-of-view images.

[0187] The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A discharge ultraviolet panoramic imaging system with an expandable wavelength band, characterized in that: The system includes: Host computer, wavelength tunable filter device, multi-field panoramic imaging module, adaptive fusion module and panoramic stitching module, wherein: The host computer is used to control the transmission wavelength adjustment, servo turntable movement and data processing; A wavelength tunable filter device, comprising a liquid crystal tunable filter for performing transmission wavelength adjustment; Multi-field panoramic imaging module, used to synchronously capture ultraviolet and visible light images of target wavelengths through rotational scanning; Adaptive fusion module for dynamically fusing UV images with visible light images; The panoramic stitching module is used to perform feature matching and panoramic stitching on the dynamically fused images.

2. The discharge ultraviolet panoramic imaging system with an expandable wavelength band according to claim 1, characterized in that: The liquid crystal tunable filter covers the 240-400nm band, and the transmission wavelength is switched in a preset step size.

3. The discharge ultraviolet panoramic imaging system with an extendable wavelength band according to claim 1, characterized in that: The multi-field panoramic imaging module includes: The lens array consists of a set number of wide-angle lenses; Servo turntable, used to drive the lens array to rotate and scan at full angles at a set speed; A spectrometer for splitting incident light into an ultraviolet light path and a visible light path; An ultraviolet imaging unit, disposed in the ultraviolet light path, for generating an ultraviolet image; A visible light imaging unit, disposed in the visible light path, for generating a visible light image; FPGA synchronization controller, used to synchronously control the exposure timing of the ultraviolet and visible light image acquisition units.

4. A discharge ultraviolet panoramic imaging method with an extendable wavelength band, based on the discharge ultraviolet panoramic imaging system with an extendable wavelength band according to any one of claims 1 to 3, characterized in that: The steps include: Select a target wavelength according to the discharge type, and control the liquid crystal tunable filter to switch to the target wavelength in a set step size; After locking the target wavelength, the servo turntable is triggered to drive the lens array to rotate, synchronously capturing multi-field visible light images and ultraviolet images; The multi-field fusion image is generated by dynamically fusing the ultraviolet image and the visible light image of the multi-field through the adaptive fusion algorithm; Based on the fusion images of multiple fields of view, the SIFT algorithm is used to perform panoramic stitching of the multi-field images.

5. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 4, characterized in that: The dynamically fusing the ultraviolet image and the visible light image by using the adaptive fusion algorithm includes: Dynamically normalize the radiation intensity of ultraviolet images; Calculate the spectral similarity between the UV image and the visible light image; Fusion weights are dynamically assigned according to the radiation intensity and spectral similarity to generate a fused image.

6. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 5, characterized in that: The dynamic normalization processing of the ultraviolet image radiation intensity includes: Setting a radiation intensity threshold of the ultraviolet image pixel, including setting it to the maximum value when the radiation intensity is greater than a first preset ratio of the maximum value, and setting it to 0 when the radiation intensity is less than a second preset ratio of the minimum value; Selecting a normalization mode according to lighting conditions, including selecting global dynamic normalization or local dynamic normalization according to the rate of change of lighting; The processed radiation intensity is converted to a value range from 0 to 1 through a linear mapping.

7. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 5, characterized in that: The calculating of the spectral similarity between the ultraviolet image and the visible light image comprises: Extracting the radiation intensity data of ultraviolet image pixels and visible light image pixels; The spectral similarity value of two images is calculated by normalized cross-correlation.

8. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 5, characterized in that: The dynamic allocation of fusion weights according to radiation intensity and spectrum similarity includes: The UV image weight is determined by a preset linear combination of UV image pixel radiation intensity and spectral similarity; The visible light image weight is 1 minus the ultraviolet image weight; The weighted UV image and the visible light image are added together to generate a fused image.

9. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 4, characterized in that: The panoramic stitching of multi-field images by using the SIFT algorithm includes: Extract SIFT feature points from multi-field fusion images; Based on the SIFT feature points, a spatial consistency check is performed to eliminate mismatched points; Panoramic image stitching is performed based on the filtered feature points.

10. The method for discharge ultraviolet panoramic imaging with an extendable wavelength band according to claim 9, characterized in that: The performing of the spatial consistency check to eliminate mismatched points also includes optimizing the stitching parameters by using the bundle adjustment method: Based on the feature matching points retained after spatial consistency verification, the coordinates of each feature point in three-dimensional space are inverted and calculated; Constructing an objective function to minimize the reprojection error between the three-dimensional space coordinates and their two-dimensional projection coordinates in adjacent lens images; The rotation matrix and translation matrix of each lens are adjusted through iterative optimization to achieve a stitching error less than the preset pixel size.

Images