A method and device for automatically detecting an optical module

Abstract

The present invention discloses an automated detection method and device for an optical module, the method comprising collecting raw fluorescence data of an optical module port; performing spectral analysis on the raw fluorescence data to separate fluorescent nano-marker signals and background signals; dividing the optical module port into multiple detection areas, respectively calculating the average intensity of the fluorescent nano-marker signal in each detection area; comparing the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold; if the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold, it is determined that there are residual stains on the optical module port; if the average intensity of the fluorescent nano-marker signal in all detection areas is lower than or equal to the corresponding preset first threshold, it is determined that the optical module port is cleaned. The technical solution of the present invention can solve the problem that the traditional image detection method cannot effectively identify pollutants with a color similar to that of the surface of the optical module port.

Description

Technical Field

[0001] The present invention relates to the technical field of optical module port detection, and in particular to an automatic detection method and device for an optical module. Background Art

[0002] The optical module is a key component in the fiber-optic communication system, which is used to realize the conversion and transmission of photoelectric signals. It usually contains a transmitter, a receiver and an optical interface, in which the optical interface (i.e., the optical module port) is directly connected to the optical fiber and is a key node for signal transmission. If there are contaminants in the optical module port, such as dust, grease or other particles, it may lead to consequences such as signal attenuation, increased reflection loss and communication interruption. Therefore, regular inspection and cleaning of the optical module port is an important measure to maintain the performance of the optical communication network, which can effectively extend the life of the equipment, reduce the occurrence of failures, and improve the overall stability and efficiency of the network.

[0003] In the automation equipment of the optical module of the exemplary technology, it is usually determined whether the port is clean by obtaining the image data of the optical module port. However, this method has significant limitations and it is difficult to effectively detect certain specific types of pollutants. For example, submicron dust particles, whose size is much smaller than the wavelength of visible light, are difficult to distinguish in ordinary optical imaging systems. In addition, the ceramic or metal materials commonly used in optical modules may have tiny scratches or pits, and these surface defects are similar to the performance of actual pollutants in the image, which increases the difficulty of identification. There are also some transparent or translucent pollutants, such as some organic residues or detergent residues, whose refractive index is close to that of the port surface material and produces almost no visible contrast under normal lighting conditions. What's more, some electrostatically adsorbed ultrafine fibers or film-like pollutants may cover the port surface, but due to their extremely thin characteristics, it is difficult to show a clear boundary in a two-dimensional image. Although these pollutants that are not easy to find are small, they may significantly affect the transmission of optical signals, resulting in increased insertion loss or worsened return loss. Therefore, the detection method that relies solely on image data has obvious shortcomings and it is difficult to fully guarantee the clean quality of the optical module port. Summary of the invention

[0004] The main purpose of the present invention is to propose an automated detection method for an optical module, aiming to solve the technical problem that traditional image detection methods cannot effectively identify pollutants with a color similar to that of the surface of the optical module port.

[0005] To achieve the above-mentioned purpose, the present invention proposes an automated detection method for optical modules, wherein the automated detection method for optical modules is applied to an automated detection device for optical modules, wherein the automated detection device for optical modules uses a cleaning agent containing a fluorescent nano-marker to decontaminate the optical module port, wherein the fluorescent nano-marker can bind to the stains in the optical module port;

[0006] The automated detection method of the optical module includes:

[0007] Performing multi-wavelength laser scanning on the optical module port, collecting the fluorescence signal of the optical module port, and obtaining raw fluorescence data;

[0008] Performing spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal;

[0009] Dividing the optical module port into a plurality of detection areas, and calculating the average intensity of the fluorescent nano-marker signal in each detection area;

[0010] comparing the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold;

[0011] If the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold, it is determined that there is residual stain on the optical module port;

[0012] If the average intensity of the fluorescent nano-marker signals in all the detection areas is lower than or equal to the corresponding preset first threshold, it is determined that the cleaning of the optical module port is completed.

[0013] Optionally, performing multi-wavelength laser scanning on the optical module port to collect the fluorescence signal of the optical module port to obtain raw fluorescence data includes:

[0014] selecting a plurality of excitation wavelengths, the plurality of excitation wavelengths being within a predetermined wavelength range;

[0015] The optical module ports are scanned sequentially using the multiple excitation wavelengths, each scan comprising:

[0016] Adjusting the laser power to a preset power level, controlling the laser beam to perform a raster scan on the surface of the optical module port, and collecting a fluorescent signal at each scanning point of the raster scan, wherein the fluorescent signal includes an intensity value and a corresponding wavelength value;

[0017] Perform signal denoising on all collected fluorescence signals;

[0018] The denoised fluorescence signal is sorted according to the scanning position and excitation wavelength to obtain the original fluorescence data.

[0019] Optionally, performing spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal includes:

[0020] Performing baseline correction on the raw fluorescence data;

[0021] The baseline-corrected data were decomposed using a multivariate curve resolution algorithm to obtain multiple characteristic spectra;

[0022] Calculating similarity scores between the plurality of characteristic spectra and each spectrum in a preset fluorescent nano-marker characteristic spectrum library;

[0023] Based on the similarity scores, matching and identifying the multiple characteristic spectra;

[0024] Based on the matching and recognition results, the fluorescent nanomarker signals are extracted and the remaining signals are regarded as background signals.

[0025] Optionally, the pre-processed data is decomposed using a multivariate curve resolution algorithm to obtain a plurality of characteristic spectra, including:

[0026] Constructing an initial estimation matrix, wherein the initial estimation matrix includes a spectral component matrix and a concentration component matrix;

[0027] The spectral component matrix and the concentration component matrix are alternately optimized using an alternating least squares method until a preset convergence condition is reached;

[0028] Apply non-negative matrix factorization algorithm to the optimized spectral component matrix for non-negative constraint processing;

[0029] Based on the spectral component matrix processed with non-negative constraints, multiple characteristic spectra are extracted.

[0030] Optionally, dividing the optical module port into a plurality of detection areas and calculating the average intensity of the fluorescent nano-marker signal in each detection area respectively includes:

[0031] Based on the geometric features of the optical module port, the optical module port is divided into a core area and an edge area;

[0032] Performing uniform grid division on the core area to obtain a plurality of core sub-areas of equal size;

[0033] Performing non-uniform grid division on the edge region to obtain a plurality of edge sub-regions of different sizes;

[0034] Taking the core sub-region and the edge sub-region as initial detection regions;

[0035] Calculating the mean intensity of the fluorescent nano-marker signal in each of the initial detection areas;

[0036] Based on the intensity mean and spatial adjacent relationship of the initial detection areas, adjacent initial detection areas are merged according to a preset merging rule to obtain a plurality of merged detection areas;

[0037] The weighted intensity average of the fluorescent nano-marker signal in each of the combined detection areas is calculated as the average intensity of the fluorescent nano-marker signal in the detection area.

[0038] Optionally, based on the intensity mean and spatial adjacent relationship of the initial detection areas, adjacent initial detection areas are merged according to a preset merging rule to obtain a plurality of merged detection areas, including:

[0039] Calculate the difference between the mean intensity of each initial detection area and its adjacent initial detection areas;

[0040] Determining the initial detection area pairs to be merged according to a preset second threshold;

[0041] Execute the region growing algorithm on the initial detection regions to be merged to form a preliminary merged region;

[0042] Calculate the shape complexity of the preliminary merged area, and if it exceeds the preset complexity threshold, simplify the boundary of the area;

[0043] Repeat the steps of calculating the intensity mean difference, determining the region pairs to be merged, executing the region growing algorithm and calculating the shape complexity until no further merging is possible or the preset maximum number of iterations is reached, and obtain the merged detection area.

[0044] Optionally, performing a region growing algorithm on the initial detection regions to be merged to form a preliminary merged region includes:

[0045] Selecting an initial detection area whose intensity mean is higher than a preset third threshold as a seed area;

[0046] Calculate the intensity mean similarity between the seed region and the adjacent initial detection region;

[0047] If the similarity of the mean intensity values ​​of the adjacent regions exceeds a preset fourth threshold, the adjacent region is added to the seed region;

[0048] Update the boundary and intensity mean of the merged region;

[0049] The steps of calculating similarity, determining merging conditions, and updating regions are repeated until no adjacent regions satisfying the merging conditions can be found, thereby forming a preliminary merging region.

[0050] The present invention also provides an automatic detection device for an optical module, comprising:

[0051] A transport component for transporting the optical module;

[0052] A first cleaning component, a second cleaning component, and a detection component are sequentially arranged in a transport direction of the optical module, and at least one of the first cleaning component and the second cleaning component is configured to use a cleaning agent containing a fluorescent nano-marker to decontaminate the optical module port;

[0053] The detection component includes a laser scanning unit, a signal acquisition unit, a spectrum analysis unit, a region division unit, an intensity calculation unit and a determination unit;

[0054] The laser scanning unit is used to perform multi-wavelength laser scanning on the optical module port;

[0055] The signal acquisition unit is used to collect the fluorescence signal of the optical module port to obtain raw fluorescence data;

[0056] The spectral analysis unit is used to perform spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal;

[0057] The area division unit is used to divide the optical module port into a plurality of detection areas;

[0058] The intensity calculation unit is used to calculate the average intensity of the fluorescent nano-marker signal in each detection area;

[0059] The determination unit is used to compare the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold value. The comparison process includes: if the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold value, it is determined that there are residual stains on the optical module port; if the average intensity of the fluorescent nano-marker signal in all detection areas is lower than or equal to the corresponding preset first threshold value, it is determined that the optical module port is cleaned.

[0060] Optionally, the first cleaning component includes a first driving member and a water absorbing member, the first driving member is in transmission connection with the water absorbing member to drive the water absorbing member to move in the optical module port, and the water absorbing member is provided with a cleaning agent containing a fluorescent nano-marker;

[0061] The second cleaning component includes a second driving member, a nozzle and a pump body component. The second driving member is connected to the nozzle for driving the nozzle to move into the optical module port. A fluid flow channel and an air port connected to the fluid flow channel are formed on the nozzle. The pump body component is connected to the fluid flow to transport high-pressure pure fluid to the optical module port through the fluid flow channel and the air port, or to form a negative pressure environment in the fluid flow channel.

[0062] The automatic detection method of the optical module proposed by the present invention is based on the use of a cleaning agent containing fluorescent nano-markers to perform a decontamination treatment on the optical module port. This special cleaning agent can combine the fluorescent nano-markers with the pollutants on the port surface.

[0063] During the inspection of the optical module port, the treated port is scanned by multi-wavelength laser scanning technology. If the port has been completely cleaned, no fluorescent signal will be detected because the fluorescent nano-markers and dirt have been completely removed during the cleaning process. However, if there is any residual contamination, the fluorescent nano-markers combined with it will produce detectable fluorescent signals under laser excitation. This method can effectively identify various types of imperceptible contaminants. For example, submicron dust particles are smaller than the wavelength of visible light, but they can produce obvious signals when combined with fluorescent nano-markers; tiny scratches or pits on the port surface can be distinguished from actual contaminants by analyzing the characteristics of the fluorescent signal; transparent or translucent contaminants will be detected through fluorescent signals as long as they remain even if the refractive index is close to the port surface; electrostatically adsorbed ultra-fine fibers or film-like contaminants, no matter how thin, can be accurately identified as long as they exist. By dividing the port into multiple detection areas, calculating the average intensity of the fluorescent nano-marker signal in each area, and comparing it with the preset threshold, the pollution distribution can be accurately located and quantified. This method not only overcomes the limitations of traditional detection methods, but also realizes accurate evaluation of the cleaning process, effectively reduces insertion loss and return loss, and significantly improves the performance and reliability of optical communication systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0065] Figure 1 It is a structural schematic diagram of an embodiment of an automatic detection device for optical modules of the present invention;

[0066] Figure 2 for Figure 1 A partial enlarged view of the middle A;

[0067] Figure 3 for Figure 1 A partial enlarged view of point B in the middle;

[0068] Figure 4 for Figure 1 A partial enlarged view of point C in the middle;

[0069] Figure 5 A simplified schematic diagram of an embodiment of a detection component;

[0070] Figure 6 The present invention is a flow chart of an automated detection method for optical modules.

[0071] Description of Figure Numbers:

[0072] 1. Transport component; 2. First cleaning component; 21. First driving member; 22. Water absorbing member; 23. Cleaning agent storage tank; 3. Second cleaning component; 31. Second driving member; 32. Nozzle; 321. Air port; 4. Detection component; 41. Laser scanning unit; 42. Signal acquisition unit; 43. Spectral analysis unit; 44. Area division unit; 45. Intensity calculation unit; 46. Judgment unit; 5. Loading and unloading mechanism.

[0073] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0076] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0077] The present invention provides an automatic detection device for an optical module.

[0078] In the embodiment of the present invention, Figures 1 to 5 As shown, the automated detection of the optical module includes:

[0079] The transport component 1 is used for transporting the optical module;

[0080] A first cleaning component 2, a second cleaning component 3 and a detection component 4 are sequentially arranged in the transport direction of the optical module, and at least one of the first cleaning component 2 and the second cleaning component 3 is configured to use a cleaning agent containing a fluorescent nano-marker to decontaminate the optical module port;

[0081] The detection component 4 includes a laser scanning unit 41, a signal acquisition unit 42, a spectrum analysis unit 43, a region division unit 44, an intensity calculation unit 45 and a determination unit 46;

[0082] The laser scanning unit 41 is used to perform multi-wavelength laser scanning on the optical module port;

[0083] The signal acquisition unit 42 is used to collect the fluorescence signal of the optical module port to obtain raw fluorescence data;

[0084] The spectrum analysis unit 43 is used to perform spectrum analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal;

[0085] The area division unit 44 is used to divide the optical module port into multiple detection areas;

[0086] The intensity calculation unit 45 is used to calculate the average intensity of the fluorescent nano-marker signal in each detection area;

[0087] The determination unit 46 is used to compare the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold value. The comparison process includes if the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold value, it is determined that there are residual stains on the optical module port; if the average intensity of the fluorescent nano-marker signal in all detection areas is lower than or equal to the corresponding preset first threshold value, it is determined that the optical module port is cleaned.

[0088] Specifically, in this embodiment, the transport component 1 may include a conveyor belt, a turntable (such as Figure 1The first cleaning component 2 and the second cleaning component 3 may include various devices capable of cleaning the optical module port, such as cotton swabs, brushes, spray devices, etc. The various units in the detection component 4 may be independent hardware devices or different functional modules integrated on the same housing. Among them, the laser scanning unit 41 may be a laser with adjustable wavelength in combination with a scanning mirror; the signal acquisition unit 42 may be a photomultiplier tube or a CCD camera; the spectral analysis unit 43, the area division unit 44, the intensity calculation unit 45 and the determination unit 46 may be processors or application-specific integrated circuits with corresponding functions.

[0089] In this embodiment, the "cleaning agent containing fluorescent nano-markers" is a specially formulated cleaning solution that contains nano-scale particles that can bind to various contaminants and emit specific fluorescence. Specifically, corresponding fluorescent nano-markers are selected for the three common types of contaminants in optical module ports:

[0090] 1. For inorganic dust, CdSe / ZnS core-shell structure quantum dots with surface modified with -COOH groups are used, with a particle size of 5nm and an emission wavelength of about 520nm.

[0091] 2. For organic residues, polystyrene nanospheres encapsulated with rhodamine B were selected, with a diameter of 50 nm and an emission wavelength of about 580 nm.

[0092] 3. For oil and fat pollutants, erbium-doped NaYF4 upconversion nanocrystals are used with a particle size of 20nm, which has characteristic emissions at 540nm and 660nm under 980nm excitation.

[0093] The concentration of these three fluorescent nano-markers in the cleaning agent is 0.01wt%. The pH value of the cleaning agent is adjusted to 7.0±0.2 by adding a small amount of acetic acid and sodium hydroxide to ensure that the ceramic or metal materials of the optical module port will not be corroded. This multi-component design enables the cleaning agent to simultaneously mark and detect multiple types of contaminants, providing a basis for subsequent multi-wavelength laser scanning and fluorescence signal analysis.

[0094] The basic solvent of the cleaning agent is anhydrous ethanol with a concentration of 95%. This choice can effectively dissolve a variety of pollutants without causing damage to the optical fiber port.

[0095] Optionally, the first cleaning component 2 includes a first driving member 21 and a water absorbing member 22, the first driving member 21 is in transmission connection with the water absorbing member 22, so as to drive the water absorbing member 22 to move in the optical module port, and a cleaning agent containing a fluorescent nano-marker is provided on the water absorbing member 22;

[0096] The second cleaning component 3 includes a second driving member 31, a nozzle 32 and a pump body component (not shown in the figure). The second driving member 31 is connected to the nozzle 32 in a transmission manner to drive the nozzle 32 to move into the optical module port. A fluid flow channel (not shown in the figure) and an air port 321 connected to the fluid flow channel are formed on the nozzle 32. The pump body component is connected to the fluid flow to transport high-pressure pure fluid to the optical module port through the fluid flow channel and the air port 321, or to form a negative pressure environment in the fluid flow channel.

[0097] Specifically, in the present invention, the first driving member 21 can be composed of a plurality of driving structures to realize the diversified movement of the water absorbing member 22 in the optical module port. Specifically, the first driving member 21 can include a rotational driving unit and a linear driving unit. The rotational driving unit can be a motor, a stepping motor or a servo motor, which is used to drive the water absorbing member 22 to perform rotational movement in the port. The linear driving unit can be an electric push rod, a cylinder, a linear motor or a screw nut mechanism, which is used to drive the water absorbing member 22 to perform linear reciprocating movement. These two units can be controlled independently or work together, so that the water absorbing member 22 can achieve compound movement in the optical module port, including rotation, linear reciprocating movement, and even spiral movement. In addition, the first driving member 21 can also be equipped with a sophisticated control system, such as a PLC or an embedded controller, for accurately adjusting the rotation speed, reciprocating frequency and movement amplitude to adapt to optical module ports of different sizes and degrees of contamination.

[0098] The rotation drive unit of the first drive member 21 can be connected to the shaft of the water absorbing member 22 by gear transmission, belt transmission or direct coupling to achieve the transmission of rotational motion. The linear drive unit can be connected to the supporting structure of the water absorbing member 22 by a connecting rod mechanism, a slider guide or a precision linear module to achieve the transmission of linear reciprocating motion. In addition, the entire drive system can be connected to the main frame by an adjustable bracket or a mechanical arm, allowing precise positioning in three-dimensional space to accommodate optical module ports of different heights and angles.

[0099] The absorbent 22 is a cleaning tool, which can be made of a variety of materials. It can be a traditional cotton swab or dust-free cloth, or it can be a more advanced microfiber material or nanofiber structure. The shape of the absorbent 22 can be customized according to the geometric characteristics of the optical module port, and can be cylindrical, conical, or a special contour shape. In order to improve the cleaning effect, the absorbent 22 can have a microscopic concave-convex structure or a special fiber arrangement to increase the contact area with the pollutants.

[0100] In the present application, a cleaning agent storage tank 23 is provided below the water absorbing member 22 , and the cleaning agent storage tank 23 is used to store cleaning agent. The water absorbing member 22 can be driven by the first driving member 21 to move into the cleaning agent storage tank 23 to contact with the cleaning agent.

[0101] The second drive member 31 may also include a plurality of drive units working in coordination. It may be composed of a high-precision stepper motor or servo motor, and cooperate with a precision reducer and encoder to achieve precise positioning and complex movement of the nozzle 32. The second drive member 31 may adopt a multi-axis design, such as a three-axis or five-axis motion system, allowing the nozzle 32 to move flexibly in a three-dimensional space. The main body of the nozzle 32 is made of corrosion-resistant and high-pressure resistant materials, such as stainless steel, ceramics or special engineering plastics. The number of air ports 321 on the nozzle 32 may be one or more. In the present application, the number of air ports 321 is at least two, and the nozzle 32 is arranged in a columnar shape so as to extend into the optical module port. At least one air port 321 is arranged on the peripheral wall of the nozzle 32, and at least one air port 321 is arranged on the end of the nozzle 32, so as to effectively spray high-pressure gas or liquid for the optical module port, or to recover the cleaned gas or liquid.

[0102] The operation process of the optical module automatic detection device of the present invention is as follows:

[0103] First, the loading and unloading mechanism 5 of the optical module automatic detection device accurately places the optical module to be detected on the transport component 1. The transport component 1 is started and smoothly transports the optical module to the working area of ​​the first cleaning component 2. Here, the water absorbing member 22 of the first cleaning component 2 is pre-soaked with a special cleaning agent containing fluorescent nano-markers, and begins to perform preliminary cleaning of the optical module port. Under the control of the first driving member 21, the water absorbing member 22 rotates and reciprocates in a straight line in the port to fully wipe the surface and interior of the port.

[0104] After cleaning is completed, the transport component 1 continues to work and transports the optical module to the location of the second cleaning component 3. At this time, the second drive member 31 controls the nozzle 32 to move precisely into the port of the optical module. The pump body component is then started, and high-pressure gas or high-pressure liquid (usually treated pure water) is delivered to the inside of the port through the nozzle 32 to deeply flush the port. This process can effectively remove residual contaminants and cleaning agents. After flushing, the pump body component switches to negative pressure mode, and the nozzle 32 forms a negative pressure environment to quickly recover the used cleaning liquid and residual dirt to ensure that the port surface is dry and clean.

[0105] Next, the transport component 1 transports the double-cleaned optical module to the working area of ​​the detection component 4. The laser scanning unit 41 of the detection component 4 begins to perform multi-wavelength laser scanning on the optical module port. At the same time, the signal acquisition unit 42 accurately captures the fluorescence signal emitted from the port surface. The collected fluorescence data is then transmitted to the spectrum analysis unit 43 for in-depth data processing and analysis.

[0106] The area division unit 44 virtually divides the port surface into multiple detection areas, and the intensity calculation unit 45 accurately calculates the intensity of the fluorescence signal in each area. Finally, the determination unit 46 compares the calculated fluorescence intensity of each area with the preset threshold value to determine the cleanliness status of the port. Based on the determination result, the system will decide whether to send the optical module back to the cleaning station for re-cleaning, or confirm that the cleaning is completed and transport it to the unloading area.

[0107] The entire process is highly automated, with each component working in coordination to ensure the efficiency and accuracy of optical module port cleaning and inspection.

[0108] The present invention also provides an automatic detection method for an optical module, which is applied to an automatic detection device for an optical module, and the specific structure of the automatic detection method for an optical module is as described in the above embodiment. Figure 6 The present invention is a flow chart of an automated detection method for optical modules.

[0109] The automatic detection method of the optical module includes:

[0110] See also Figure 6 , performing multi-wavelength laser scanning on the optical module port, collecting the fluorescence signal of the optical module port, and obtaining original fluorescence data;

[0111] In one embodiment of the present invention, the multi-wavelength laser scanning of the optical module port, collecting the fluorescence signal of the optical module port, and obtaining the original fluorescence data include: selecting multiple excitation wavelengths, wherein the multiple excitation wavelengths are within a predetermined wavelength range; sequentially scanning the optical module port with the multiple excitation wavelengths, wherein each scan includes: adjusting the laser power to a preset power level, controlling the laser beam to perform a raster scan on the surface of the optical module port, collecting the fluorescence signal at each scanning point of the raster scan, wherein the fluorescence signal includes an intensity value and a corresponding wavelength value; performing signal denoising processing on all collected fluorescence signals; and sorting the denoised fluorescence signals according to the scanning position and the excitation wavelength to obtain the original fluorescence data.

[0112] Specifically, it is necessary to first select multiple excitation wavelengths that are optimized for the three fluorescent nano-markers in the cleaning agent. Specifically, at least three excitation wavelengths of 405nm, 532nm and 980nm are selected. 405nm is used to excite CdSe / ZnS quantum dots, 532nm is used to excite nanospheres wrapped with rhodamine B, and 980nm is used to excite erbium-doped NaYF4 upconversion nanocrystals. In this way, comprehensive detection capabilities for inorganic dust, organic residues and grease pollutants are ensured.

[0113] Next, the three excitation wavelengths are used in turn to scan the optical module port. Take the optical module as an LC type fiber optic connector as an example, and its end face diameter is 2.5mm. For each wavelength scan, first adjust the laser power to the preset power level. Specifically, the 405nm laser is set to 5mW, the 532nm is set to 10mW, and the 980nm is set to 50mW. This power setting can ensure sufficient excitation intensity while avoiding damage to the fiber end face. Subsequently, the laser beam is controlled to perform a grid scan on the surface of the optical module port. The 2.5mm diameter end face is divided into 100×100 grids, each grid size is 25μm×25μm. The scan adopts a serpentine path, starting from the upper left corner, first scanning a row from left to right, then moving down one grid, and then scanning the next row from right to left, and so on until the entire end face is scanned.

[0114] At each scanning point of the grid scanning, the system collects fluorescence signals. The fluorescence signal includes the intensity value and the corresponding wavelength value. A high-sensitivity spectrometer is used to collect a data point every 0.5nm in the range of 400nm to 700nm (this range covers the emission wavelength range of the fluorescent nano-markers listed above and corresponds to the visible spectrum of the human eye), and a spectrum containing 600 data points is obtained. Each data point contains a wavelength value and a corresponding intensity value.

[0115] To make it easier to understand, the process is explained again to more accurately reflect the actual operation:

[0116] 1. First, use a 405nm laser. The laser beam scans the entire optical module port surface according to the preset grid pattern (100×100). At each scanning point, use a high-sensitivity spectrometer to collect fluorescence spectra in the range of 400-700nm, with one data point every 0.5nm, and obtain 600 data points (note that 600 data points are collected at each scanning point) until all 10,000 (100×100) scanning points are collected.

[0117] 2. Switch to 532nm laser, scan the entire port surface again, and collect the fluorescence spectrum of each point.

[0118] 3. Finally, switch to the 980nm laser and perform a third full scan to complete the last round of fluorescence spectrum acquisition.

[0119] Next, all collected fluorescence signals are denoised to improve data quality. Specifically, the wavelet transform denoising method is used. After that, the fluorescence signals are sorted according to the scanning position and excitation wavelength, and a four-dimensional array D

[100]

[100] [3]

[600] is constructed to store the raw fluorescence data. In this array, D[x][y][z][w] represents the position at (x, y) coordinates, when the zth excitation wavelength is used, at a wavelength of (400 w 0.5) nm. Specifically, the range of x and y are both 1 to 100, corresponding to a 100×100 scanning grid; the value of z is 0, 1, and 2, corresponding to the three excitation wavelengths of 405nm, 532nm, and 980nm respectively; w is the array index, ranging from 0 to 599, corresponding to the emission wavelength range of 400nm to 699.5nm. For example, D

[50]

[50] [1]

[360] represents the fluorescence intensity measured at 580nm at the center position (50, 50) of the end face when using a 532nm excitation wavelength; while D[1][1][0]

[240] represents the fluorescence intensity measured at 520nm at the upper left corner of the end face when using a 405nm excitation wavelength. This data structure makes it easy to extract the complete fluorescence spectrum at a specific position and a specific excitation wavelength in subsequent analysis. For example, by comparing the values ​​of D[x][y][0]

[240] and D[x][y][1]

[360] , it can be determined whether inorganic dust (CdSe / ZnS quantum dots) or organic residues (rhodamine B) are more likely to exist at the (x, y) position. In addition, to obtain the complete spectrum of a certain position (x, y) at all excitation wavelengths, the data of D[x][y][:][:] can be extracted; to analyze the entire end face fluorescence distribution under a specific excitation wavelength z, the data of D[:,:][z][w] can be extracted, where w is the index corresponding to the characteristic emission wavelength of the specific pollutant. This data sorting method not only provides a basis for the accurate identification and location of different types of pollutants, but also greatly improves the efficiency and accuracy of subsequent data analysis, providing reliable data support for the precise cleaning and maintenance of optical module ports.

[0120] Please continue reading Figure 6 , performing spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal;

[0121] In one embodiment of the present invention, the spectral analysis of the raw fluorescence data to separate the fluorescent nanomarker signal and the background signal includes: performing baseline correction on the raw fluorescence data; decomposing the baseline-corrected data using a multivariate curve resolution algorithm to obtain a plurality of characteristic spectra; calculating similarity scores between the plurality of characteristic spectra and each spectrum in a preset fluorescent nanomarker characteristic spectrum library; matching and identifying the plurality of characteristic spectra based on the similarity scores; and extracting the fluorescent nanomarker signal based on the matching and identification results, and treating the remaining signal as a background signal.

[0122] In one embodiment of the present invention, the preprocessed data is decomposed using a multivariate curve resolution algorithm to obtain a plurality of characteristic spectra, including: constructing an initial estimation matrix, the initial estimation matrix including a spectral component matrix and a concentration component matrix; using an alternating least squares method to alternately optimize the spectral component matrix and the concentration component matrix until a preset convergence condition is reached; applying a non-negative matrix decomposition algorithm to the optimized spectral component matrix for non-negative constraint processing; and extracting a plurality of characteristic spectra based on the spectral component matrix processed with non-negative constraints.

[0123] Specifically, first, baseline correction is performed on the raw fluorescence data. This step is performed for each spectrum D[x][y][z][:] in the four-dimensional array D

[100]

[100] [3]

[600] . The specific operation is to apply the adaptive iterative polynomial fitting method, using the third-order polynomial f(λ)= + λ+ + The fitting is performed with λ being the wavelength (400 nm to 699.5 nm, with an interval of 0.5 nm). Initially, Let be the average intensity of the spectrum, , , Set to 0. These coefficients are then iteratively optimized using the least squares method until the residual change is less than 0.1% or 10 iterations are reached. Finally, the fitted baseline f(λ) is subtracted from the original spectrum to obtain the corrected spectrum. For example, for D

[50]

[50] [1][:] (which represents the spectrum obtained using a 532nm excitation wavelength at the center of the end face), the intensity of the original data at 400nm may be 105, which may become 6.5 after baseline correction. This process is performed for all 300,000 spectra (100×100×3) separately, and finally a new four-dimensional array D'

[100]

[100] [3]

[600] is obtained after baseline correction.

[0124] Next, the baseline-corrected data were decomposed using a multivariate curve resolution algorithm to obtain multiple characteristic spectra. Specifically, the multivariate curve resolution-alternating least squares method (MCR-ALS) was used. First, the four-dimensional array D' was reorganized into a two-dimensional matrix X[30000×600], where 30000=100×100×3, representing the combination of all scanning points and excitation wavelengths, and 600 represents the number of spectral data points. The number of groups was set to 5, including 3 known fluorescent nano-markers (CdSe / ZnS quantum dots, nanospheres wrapped with rhodamine B, and erbium-doped NaYF4 upconversion nanocrystals) and 2 possible background signals (such as instrument autofluorescence and ambient light interference).

[0125] The pure variable method is used to select the initial estimated spectrum and construct the initial spectrum matrix S[600×5]. Specifically, first find the spectrum with the largest intensity in the X matrix as the first estimated spectrum, then find the spectrum least correlated with the first spectrum in the remaining data as the second estimated spectrum, and so on, until 5 initial estimated spectra are selected. These 5 spectra constitute the initial value of the S matrix, and each column represents an estimated component spectrum.

[0126] Then, by alternately optimizing the concentration matrix C[30000×5] and the spectrum matrix S[600×5], we make X[30000×600]≈C[30000×5]· [5×600], where X is a two-dimensional matrix formed by reorganizing the four-dimensional array D', the C matrix represents the relative concentration of each component in each sample, the S matrix represents the characteristic spectrum of each component, and T represents the transpose operation of the matrix. The optimization process is as follows:

[0127] 1. Fix S[600×5] and solve C[30000×5], the formula is:

[0128] C[30000×5]=X[30000×600]·S[600×5]· [5×5].

[0129] 2. Fix C[30000×5] and solve S[600×5], the formula is:

[0130] [30000×600];

[0131] 3. Apply non-negativity constraints (all elements ≥ 0) and monotonicity constraints (spectral shape remains monotonic) to C and S. The non-negativity constraint ensures that the concentration and spectral intensity are always positive, and the monotonicity constraint helps maintain the rationality of the spectral shape.

[0132] 4. Calculate the residual (The residual R represents the difference between the original data X and the reconstructed data C The difference between ), the formula is:

[0133] ;

[0134] 5. If the residual change is less than 0.1% or reaches 100 iterations, stop; otherwise return to step 1 (that is, fix S and solve C). This condition ensures that the algorithm stops when sufficient accuracy or the maximum number of iterations is reached.

[0135] Each column in the final S matrix is ​​a characteristic spectrum. For example, S[:, 0] may represent the characteristic spectrum of CdSe / ZnS quantum dots, with a significant peak at around 520nm; S[:, 1] may represent the characteristic spectrum of rhodamine B, with a peak at around 580nm; and so on.

[0136] Next, the similarity scores of these characteristic spectra and the spectra in the preset fluorescent nano-marker characteristic spectrum library are calculated. The characteristic spectrum library L[600×3] contains the standard spectra of three fluorescent nano-markers. The cosine similarity method is used to calculate the similarity scores of the characteristic spectra. And the spectrum in the library , the similarity calculation formula is:

[0137] ;

[0138] in, and Represents vectors and The norm of .

[0139] Calculate the similarity of each characteristic spectrum with all spectra in the library to form a similarity matrix SM[5×3]. Based on the calculated similarity scores, match and identify multiple characteristic spectra. Set the similarity threshold to 0.95. For each row in the SM matrix, if the maximum value exceeds the threshold, the characteristic spectrum is matched to the corresponding fluorescent nanomarker; if all values ​​are below the threshold, the characteristic spectrum is marked as unknown. For example, if SM[1,:]=[0.98, 0.45, 0.32], the first characteristic spectrum is identified as the signal of CdSe / ZnS quantum dots.

[0140] Finally, according to the matching and recognition results, the fluorescent nanomarker signals are extracted, and the remaining signals are regarded as background signals. For each identified fluorescent nanomarker signal, its corresponding spectrum (corresponding column in the S matrix) and concentration distribution (corresponding column in the C matrix) are saved as valid signals. For example, if the first, third, and fourth characteristic spectra are identified as fluorescent nanomarker signals, S[:, 0], S[:, 2], S[:, 3] and C[:, 0], C[:, 2], C[:, 3] are extracted as valid signals, and the rest are background signals.

[0141] This method can not only accurately identify and locate different types of contaminants, but also effectively remove background interference, providing reliable data support for the precise cleaning and maintenance of optical module ports. Through this comprehensive and detailed data analysis process, the contamination status of the optical module port can be accurately judged, providing precise guidance for subsequent cleaning work.

[0142] Please continue reading Figure 6 , dividing the optical module port into a plurality of detection areas, and calculating the average intensity of the fluorescent nano-marker signal in each detection area respectively;

[0143] In one embodiment of the present invention, the optical module port is divided into multiple detection areas, and the average intensity of the fluorescent nano-marker signal in each detection area is calculated respectively, including: based on the geometric characteristics of the optical module port, the optical module port is divided into a core area and an edge area; the core area is uniformly gridded to obtain multiple core sub-areas of equal size; the edge area is non-uniformly gridded to obtain multiple edge sub-areas of different sizes; the core sub-areas and the edge sub-areas are used as initial detection areas; the intensity mean of the fluorescent nano-marker signal in each of the initial detection areas is calculated; based on the intensity mean and spatial adjacent relationship of the initial detection areas, the adjacent initial detection areas are merged according to a preset merging rule to obtain multiple merged detection areas; the weighted intensity mean of the fluorescent nano-marker signal in each merged detection area is calculated as the average intensity of the fluorescent nano-marker signal in the detection area.

[0144] In one embodiment of the present invention, based on the intensity mean and spatial adjacent relationship of the initial detection areas, adjacent initial detection areas are merged according to a preset merging rule to obtain multiple merged detection areas, including: calculating the intensity mean difference between each initial detection area and its adjacent initial detection area; determining the initial detection area pairs to be merged according to a preset second threshold; executing a region growing algorithm on the initial detection area pairs to be merged to form a preliminary merged area; calculating the shape complexity of the preliminary merged area, and if it exceeds a preset complexity threshold, simplifying the boundary of the area; repeating the steps of calculating the intensity mean difference, determining the area pairs to be merged, executing the region growing algorithm and calculating the shape complexity until further merging is impossible or the preset maximum number of iterations is reached to obtain a merged detection area.

[0145] In one embodiment of the present invention, the initial detection areas to be merged are subjected to a region growing algorithm to form a preliminary merged area, including: selecting an initial detection area whose intensity mean is higher than a preset third threshold as a seed area; calculating the similarity between the seed area and the intensity mean of an adjacent initial detection area; if the intensity mean similarity of an adjacent area exceeds a preset fourth threshold, adding the adjacent area to the seed area; updating the boundary and intensity mean of the merged area; and repeatedly executing the steps of calculating the similarity, judging the merging condition and updating the area until no adjacent area that meets the merging condition can be found, thereby forming a preliminary merged area.

[0146] It should be noted that the process of dividing the optical module port into the core area and the edge area needs to be flexibly adjusted according to the characteristics of different types of optical module ports. Taking LC and SC fiber optic connectors as an example, for LC fiber optic connectors, the end face is usually circular with a diameter of 2.5mm. In this case, the central 60% of the area can be defined as the core area, that is, the circular area with a radius of 0.968mm, and the remaining 40% of the area is defined as the edge area. This division takes into account the characteristics that the optical signal is mainly transmitted through the central area of ​​the port, while the edge area is more likely to accumulate pollutants.

[0147] For SC type fiber optic connectors, the end face is usually square with a side length of about 2.5mm. Considering the geometric characteristics of the SC connector, different division strategies can be adopted. One method is to define the square area of ​​1.5mm×1.5mm in the center (accounting for 36% of the total area) as the core area, and the remaining edge part as the edge area. Another more sophisticated method is to use rounded square division, that is, a rounded square with a radius of 0.9mm with the center of the end face as the center of the circle as the core area (accounting for about 50% of the total area), and the remaining part is the edge area. This rounded square division method takes into account both the square characteristics of the SC connector and the circular characteristics of the optical transmission path.

[0148] In this way, both the circular LC connector and the square SC connector can be reasonably divided into regions, laying the foundation for subsequent fine grid division and pollution detection. The flexibility and adaptability of this division strategy can also be extended to other types of optical module ports, such as MTP / MPO connectors, etc., just by making corresponding adjustments according to their specific geometric shapes and sizes.

[0149] After that, the core area is uniformly meshed. This process aims to divide the core area into multiple sub-areas of equal size for more accurate and systematic contamination detection. The specific implementation method needs to be adjusted appropriately according to the shape characteristics of the optical module port. For LC-type fiber optic connectors, the core area is circular. The circular core area can be inscribed in a square, and then the square is uniformly meshed. First, the minimum side length of the square that can completely cover the circular core area is calculated, that is, the diameter is 1.936mm (2×0.968mm). Next, the square is divided into a 12×12 grid to obtain 144 small squares. The side length of each small square is about 0.161mm (1.936mm / 12). Then, only those small squares whose center points fall within the circle are retained as valid core sub-areas. Through simple geometric calculations, it can be known that the center points of about 112 small squares are located within the circle, which constitute the core sub-areas.

[0150] For SC type fiber optic connectors, if the square core area is divided, the Cartesian coordinate system is used for uniform grid division. Assuming that the core area is a 1.5mm×1.5mm square, it can be divided into a 12×12 grid, resulting in 144 square core sub-areas of 0.125mm×0.125mm in size.

[0151] This uniform grid division method ensures uniformity and accuracy in the detection of the core area, and each sub-area can receive equal attention. Secondly, it is also convenient for subsequent data processing and analysis, and it is easy to compare the pollution conditions in different locations. In addition, the grid density can be flexibly adjusted as needed to balance the detection accuracy and computational efficiency.

[0152] Afterwards, the edge area is divided into non-uniform grids to obtain multiple edge sub-areas of different sizes. The implementation of this method needs to take into account the particularity of the edge area, because the edge area is more likely to accumulate pollutants, and the pollution distribution is usually uneven. Taking the LC type fiber optic connector as an example, after completing the division of the core area, the remaining annular area is the edge area. The specific operation of non-uniform grid division of this annular area is as follows: First, the edge area is divided into three concentric rings in the radial direction, and the width ratio from the inside to the outside is set to 3:2:1, so that the ring close to the core area is wider, and the ring close to the port edge is narrower. Then, each ring is divided into 16 sectors in the circumferential direction. In this way, the entire edge area is divided into 48 sub-areas of different sizes, among which the sub-area area of ​​the innermost ring is the largest and the sub-area area of ​​the outermost ring is the smallest.

[0153] For example, assuming that the end face diameter of the LC fiber optic connector is 2.5mm and the radius of the core area is 0.968mm, the width of the edge area is 0.282mm (i.e. 1.25mm-0.968mm). According to the ratio of 3:2:1, the widths of the three rings are 0.141mm, 0.094mm and 0.047mm respectively. Therefore, the sub-area area of ​​the innermost ring is about 0.0345mm², the sub-area area of ​​the middle ring is about 0.0230mm², and the sub-area area of ​​the outermost ring is about 0.0115mm². The advantage of this non-uniform division method is that it can pay more attention to the areas close to the core while ensuring the overall detection accuracy of the edge area, because these areas have a greater impact on the transmission of optical signals. At the same time, by reducing the area of ​​the outer sub-area, the detection resolution of the outermost area of ​​the port is improved, which is conducive to capturing contaminants that may penetrate from the outside.

[0154] Afterwards, the core sub-region and edge sub-region obtained in the previous step are uniformly regarded as the initial detection region, and then a separate signal intensity analysis is performed on each region. In the specific implementation, it is necessary to use the previously collected original fluorescence data, which is stored in a four-dimensional array D

[100]

[100] [3]

[600] , where the first two dimensions represent the scanning position, the third dimension represents the excitation wavelength, and the last dimension represents the emission spectrum.

[0155] Taking the LC fiber optic connector as an example, assuming that the core area is divided into 112 sub-areas and the edge area is divided into 48 sub-areas, there are a total of 160 initial detection areas. For each initial detection area, it is necessary to determine its corresponding set of scanning points. For example, for a sub-area located in the center of the core area, it may contain 25 scanning points (5×5 grid). Next, the fluorescence signal data of these 25 scanning points are processed. Considering that three excitation wavelengths (405nm, 532nm and 980nm) are used, the mean signal intensity at each excitation wavelength needs to be calculated separately.

[0156] The specific calculation process is as follows: First, for each scanning point, extract its fluorescence spectrum (600 data points) at a specific excitation wavelength. Then, find the intensity value corresponding to the emission wavelength of the fluorescent nanomarker in this spectrum. For example, for CdSe / ZnS quantum dots, focus on the intensity value around 520nm; for Rhodamine B, focus on the intensity value around 580nm. Next, calculate the average intensity of all scanning points in the sub-area at this wavelength. Finally, combine the average values ​​obtained under the three excitation wavelengths to form the characteristic intensity value of the initial detection area.

[0157] This method provides an effective way to compress high-resolution scan data into easy-to-process regional statistics, greatly reducing the computational workload of subsequent processing. Secondly, by considering the signals at different excitation wavelengths separately, it is possible to fully capture the characteristics of different types of fluorescent nanomarkers and improve the accuracy of pollutant identification. Furthermore, this region-based analysis method retains the spatial information of pollution distribution, which is conducive to subsequent pollution pattern recognition and cleaning strategy formulation. Finally, by calculating the intensity mean for each sub-region separately, it provides flexibility for subsequent regional merging and pollution judgment, allowing the judgment criteria to be adjusted according to actual needs.

[0158] Afterwards, the intensity mean difference between each initial detection area and its adjacent initial detection area is calculated, and the initial detection area pair that needs to be merged is determined based on the preset second threshold. This process first requires defining the adjacent relationship of each initial detection area, then calculating the intensity difference between adjacent areas, and finally determining whether merging is required based on the set threshold.

[0159] For ease of understanding, taking the LC fiber optic connector as an example, assume that after the previous steps, the intensity means of 160 initial detection areas (including core areas and edge areas) have been obtained. First, it is necessary to establish an adjacent relationship matrix to record the adjacent areas of each area. For a sub-area in the core area, it usually has 4 adjacent areas (upper, lower, left, and right), while for the sub-area in the edge area, the number of adjacent areas may be different and requires special processing. Next, for each initial detection area, calculate the difference between its intensity means and all adjacent areas. For example, for a core sub-area with an intensity mean of 100 (arbitrary units), the intensity means of its four adjacent areas are 98, 102, 95, and 105, respectively. Then, the calculated intensity differences are 2, 2, 5, and 5, respectively. This process needs to be repeated for all 160 initial detection areas, and finally a data set containing the intensity differences of all adjacent area pairs is obtained.

[0160] Then, a preset second threshold is set to determine whether the adjacent regions need to be merged. The selection of this second threshold needs to take into account the actual application scenario and the desired degree of merging. For example, the threshold can be set to 10% of the intensity mean. In the above example, if the threshold is set to 10, the adjacent regions whose intensity difference with the core sub-region is less than 10 (i.e., the two regions with intensities of 98 and 102) will be marked as potential merge objects.

[0161] This method provides a data-driven adaptive region merging strategy that can dynamically adjust the size and shape of the detection area according to the actual pollution distribution. Secondly, by comparing the intensity differences between adjacent regions, the boundaries and distribution patterns of pollution can be effectively identified, thereby improving the accuracy of pollution detection. Furthermore, while retaining important details, this method can effectively reduce data redundancy and improve the efficiency of subsequent processing. Finally, by adjusting the second threshold, the degree of merging can be flexibly controlled to meet different application requirements and pollution detection accuracy requirements.

[0162] After that, the initial detection area with an intensity mean higher than the preset third threshold is selected as the seed area, and then the area range is gradually expanded through iterative calculation and merging, and finally a preliminary merged area is formed. In specific implementation, it is first necessary to set a suitable third threshold. For example, the 75% quantile of the intensity mean of all initial detection areas can be selected as the threshold. Taking the LC fiber optic connector as an example, assuming that there are 160 initial detection areas, the area with the highest intensity mean is 200 (arbitrary unit), and the third threshold is set to 150. Then, all initial detection areas with an intensity mean higher than 150 will be regarded as potential seed areas.

[0163] Next, the region with the highest mean intensity is selected from these potential seed regions as the starting seed region, and then the similarity between the mean intensity of the seed region and its adjacent initial detection region is calculated. The similarity can be calculated in a variety of ways, such as the relative difference formula: similarity = 1-|seed region intensity-adjacent region intensity| / max(seed region intensity, adjacent region intensity). For example, if the mean intensity of the seed region is 200 and the mean intensity of the adjacent region is 180, then their similarity is 1-|200-180| / 200=0.9.

[0164] Then, a preset fourth threshold is set to determine whether to merge the adjacent area into the seed area. The selection of this threshold needs to balance the degree of merging and the need to retain details. The fourth threshold can be set to 0.85. If the calculated similarity exceeds 0.85, the adjacent area is added to the seed area. After each merge, the boundary and intensity mean of the merged area need to be updated. Boundary updating involves recording the new area range, and the update of the intensity mean can adopt the area-weighted average method, that is, the new intensity mean = (original seed area area × original seed area intensity + newly added area area × newly added area intensity) / (original seed area area + newly added area area).

[0165] This process will be repeated continuously, and each time the similarity between the updated seed area and its new adjacent area is calculated to determine whether the merging conditions are met, and the corresponding updates are performed until no adjacent area that meets the merging conditions can be found. The final area is a preliminary merged area. It is worth noting that this process may form multiple unconnected preliminary merged areas, each of which represents a relatively independent contaminated area.

[0166] The method of the present application can adaptively process pollution areas of different shapes and sizes without being restricted by a preset grid; secondly, by growing from high-intensity areas, it can effectively capture the core areas of pollution; thirdly, the similarity-based merging strategy can effectively reduce data fragmentation while retaining the details of pollution distribution; finally, the iterative nature of this method enables it to handle complex pollution distribution patterns and adapt to various practical situations.

[0167] After that, it is necessary to calculate the shape complexity of the preliminary merged area, and then decide whether to simplify the boundaries based on the preset threshold, and finally obtain the final merged detection area through iterative optimization. The shape complexity is calculated using the ratio of the square of the perimeter to the area, that is, complexity = perimeter² / (4π×area). This indicator is the smallest when it is a circle, equal to 1, and the more irregular the shape, the larger the value. Taking a contaminated area on an LC fiber optic connector as an example, assuming that an irregular shape is formed after the preliminary merger, with a perimeter of 10mm and an area of ​​5mm², then its complexity is 10² / (4π×5)≈1.59. Set a preset complexity threshold, such as 1.5. If the calculated complexity exceeds this threshold, the boundary of the area needs to be simplified.

[0168] The Douglas-Peucker algorithm is used for boundary simplification. This algorithm simplifies the polygon boundary by gradually removing vertices that contribute less to the overall shape. In specific operations, the boundary of the preliminary merged area is first represented as a series of vertices, and then a distance threshold is set, such as 0.05mm. The algorithm draws a straight line from the starting point and the end point of the boundary, calculates the distance from all intermediate points to this straight line, and directly replaces the original boundary with this straight line if the distance of all points is less than the threshold; otherwise, the point with the largest distance is selected to divide the boundary into two segments, and the above process is repeated for each segment until the distance threshold is met or the preset number of iterations is reached.

[0169] After the boundary simplification is completed, the intensity mean of the region needs to be recalculated. Assuming that the region before simplification contains 100 initial detection points, and 80 are retained after simplification, the new intensity mean will be the average of these 80 points. This process ensures that the main features of the region are retained while simplifying the shape.

[0170] Next, the steps of calculating the mean difference of intensity, determining the region pairs to be merged, executing the region growing algorithm, and calculating the shape complexity are repeated. Each iteration may generate new merged regions, which will become the input for the next round of iteration. For example, after the first round of iteration, five merged regions may be formed, and the second round of iteration may further merge these five regions into three larger regions. This process will continue until no region pairs that meet the merging conditions can be found, or the preset maximum number of iterations, such as 10, is reached.

[0171] The final merged detection area retains the main features of the pollution distribution and has a relatively simple shape, which is convenient for subsequent analysis and processing. This method avoids overly complex regional shapes by controlling the shape complexity, improving the efficiency of subsequent processing; secondly, the boundary simplification process reduces the amount of data while retaining the main features, which is conducive to storage and transmission; thirdly, the iterative optimization process can adaptively handle various complex pollution distribution patterns and is suitable for pollution of different types and degrees; finally, this method provides a more intuitive and easy-to-understand representation of the pollution area while ensuring detection accuracy.

[0172] In the subsequent steps, it is necessary to clarify the initial detection areas contained in each merged detection area, because in the previous merging process, multiple initial detection areas may have been combined into a larger area. For example, on the end face of an LC fiber optic connector, a merged detection area may contain the original 5 core sub-areas and 3 edge sub-areas. Next, it is necessary to trace back to the original data of these initial detection areas, that is, the fluorescence signal data stored in the four-dimensional array D

[100]

[100] [3]

[600] .

[0173] For each initial detection area, its corresponding fluorescence signal data needs to be extracted. Assuming that an initial detection area corresponds to 25 scanning points in the array (a 5×5 grid), all spectral data of these 25 points under three excitation wavelengths need to be processed. Specifically, for each scanning point, it is necessary to find the intensity value corresponding to the emission wavelength of the fluorescent nanomarker in its spectrum. For example, for CdSe / ZnS quantum dots, pay attention to the intensity value around 520nm; for Rhodamine B, pay attention to the intensity value around 580nm; for erbium-doped NaYF4 upconversion nanocrystals, it is necessary to pay attention to the intensity values ​​at 540nm and 660nm at the same time.

[0174] Then, the average intensity values ​​of all scanning points in each initial detection area at these specific wavelengths are calculated. This process actually performs a dimensionality reduction process on the original data, compressing a large amount of spectral data into a few representative values. The advantage of this is that it greatly reduces the complexity of subsequent calculations while retaining the most critical information.

[0175] The weight of each initial detection area in the merged large area can be determined based on its area or the number of scan points it contains. For example, if a merged detection area contains 3 core sub-areas and 2 edge sub-areas, assuming that the area of ​​the core sub-area is twice that of the edge sub-area, then the core sub-area can be given a weight of 2 and the edge sub-area can be given a weight of 1. This process takes into account the importance of different areas, so that the final result can better reflect the actual pollution status of the entire merged area. The core area usually has a greater impact on the transmission of optical signals, so it is reasonable to give it a higher weight.

[0176] In addition, this weighted average method can be flexibly adjusted to adapt to different types of optical module ports or specific application requirements. For example, for some ports with special shapes, more complex weight distribution schemes can be designed based on the importance of the optical signal transmission path. For example, in a rectangular port, a higher weight may be given to the area close to the center, while the weight of the edge area may be gradually reduced.

[0177] Finally, the calculated weighted intensity mean is used as the average intensity of the fluorescent nano-marker signal in the combined detection area. This value will be used for subsequent threshold comparison and pollution determination. In this way, not only can a comprehensive index reflecting the pollution degree of the entire area be obtained, but also the spatial information of the pollution distribution can be retained to a certain extent. This is of great significance for formulating accurate cleaning strategies and evaluating cleaning effects.

[0178] Please continue reading Figure 6 , comparing the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold;

[0179] If the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold, it is determined that there is residual stain on the optical module port;

[0180] If the average intensity of the fluorescent nano-marker signals in all the detection areas is lower than or equal to the corresponding preset first threshold, it is determined that the cleaning of the optical module port is completed.

[0181] This process requires setting a suitable threshold for each detection area, which may vary depending on the location and importance of the area. For example, for LC-type fiber optic connectors, a lower threshold, such as 50 (assumed units), can be set for the core area, while a slightly higher threshold, such as 70, can be set for the edge area. This differentiated threshold setting takes into account the degree of influence of different areas on the transmission of optical signals. In actual operation, the system compares the average intensity of the fluorescent nano-marker signal of each detection area with its corresponding preset first threshold one by one. If the signal intensity of any detection area is found to exceed its corresponding threshold, the system will immediately determine that there is residual stain on the optical module port. For example, if the signal intensity of a detection area in the core area is 55, which exceeds the preset threshold of 50, the system will determine that the cleaning is not completed even if the signal intensity of all other areas is below the threshold. This strict judgment standard ensures the comprehensive cleaning of the optical module port. Only when the signal intensity of all detection areas is lower than or equal to their respective preset thresholds, the system will determine that the optical module port is cleaned. The advantage of this method is that it can accurately locate the location of residual stains and provide guidance for subsequent precise cleaning. At the same time, it also provides an objective and quantifiable cleaning standard, effectively avoiding the errors and inconsistencies that may be caused by human judgment. In addition, by adjusting the thresholds of different areas, this method can also flexibly adapt to different types of optical modules and different levels of cleaning requirements, providing great flexibility and adaptability for the entire automated detection system.

[0182] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. An automated detection method for an optical module, characterized in that: The automated detection method of the optical module is applied to an automated detection device for the optical module, and the automated detection device for the optical module uses a cleaning agent containing a fluorescent nano-marker to decontaminate the optical module port, and the fluorescent nano-marker can bind to the stains in the optical module port; The automated detection method of the optical module includes: Performing multi-wavelength laser scanning on the optical module port, collecting the fluorescence signal of the optical module port, and obtaining original fluorescence data; performing multi-wavelength laser scanning on the optical module port, collecting the fluorescence signal of the optical module port, and obtaining original fluorescence data, including: selecting multiple excitation wavelengths, the multiple excitation wavelengths are within a predetermined wavelength range; sequentially scanning the optical module port with the multiple excitation wavelengths, each scan including: adjusting the laser power to a preset power level, controlling the laser beam to perform a grid-type scan on the surface of the optical module port, collecting a fluorescence signal at each scanning point of the grid-type scan, the fluorescence signal including an intensity value and a corresponding wavelength value; performing signal denoising on all collected fluorescence signals; and arranging the denoised fluorescence signals according to the scanning position and the excitation wavelength to obtain original fluorescence data; Performing spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal; performing spectral analysis on the raw fluorescence data to separate the fluorescent nano-marker signal and the background signal, including: performing baseline correction on the raw fluorescence data; decomposing the baseline-corrected data using a multivariate curve resolution algorithm to obtain a plurality of characteristic spectra; calculating similarity scores between the plurality of characteristic spectra and each spectrum in a preset fluorescent nano-marker characteristic spectrum library; matching and identifying the plurality of characteristic spectra based on the similarity scores; extracting the fluorescent nano-marker signal according to the matching and identification results, and treating the remaining signal as a background signal; The optical module port is divided into multiple detection areas, and the average intensity of the fluorescent nano-marker signal in each detection area is calculated respectively; the said dividing the optical module port into multiple detection areas, and the average intensity of the fluorescent nano-marker signal in each detection area is calculated respectively, including: based on the geometric characteristics of the optical module port, the optical module port is divided into a core area and an edge area; the core area is uniformly gridded to obtain a plurality of core sub-areas of equal size; the edge area is non-uniformly gridded to obtain a plurality of edge sub-areas of unequal size; the core sub-areas and the edge sub-areas are used as initial detection areas; the intensity mean of the fluorescent nano-marker signal in each of the initial detection areas is calculated; based on the intensity mean and spatial adjacent relationship of the initial detection areas, the adjacent initial detection areas are merged according to a preset merging rule to obtain a plurality of merged detection areas; the weighted intensity mean of the fluorescent nano-marker signal in each of the merged detection areas is calculated as the average intensity of the fluorescent nano-marker signal in the detection area; comparing the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold; If the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold, it is determined that there is residual stain on the optical module port; If the average intensity of the fluorescent nano-marker signals in all the detection areas is lower than or equal to the corresponding preset first threshold, it is determined that the cleaning of the optical module port is completed.

2. The automated detection method for optical modules according to claim 1, characterized in that: The pre-processed data is decomposed using a multivariate curve resolution algorithm to obtain multiple characteristic spectra, including: Constructing an initial estimation matrix, wherein the initial estimation matrix includes a spectral component matrix and a concentration component matrix; The spectral component matrix and the concentration component matrix are alternately optimized using an alternating least squares method until a preset convergence condition is reached; Apply non-negative matrix factorization algorithm to the optimized spectral component matrix for non-negative constraint processing; Based on the spectral component matrix processed with non-negative constraints, multiple characteristic spectra are extracted.

3. The automated detection method for optical modules according to claim 1, characterized in that: Based on the intensity mean and spatial adjacent relationship of the initial detection area, the adjacent initial detection areas are merged according to a preset merging rule to obtain a plurality of merged detection areas, including: Calculate the difference between the mean intensity of each initial detection area and its adjacent initial detection areas; Determining the initial detection area pair to be merged according to a preset second threshold; Execute the region growing algorithm on the initial detection regions to be merged to form a preliminary merged region; Calculate the shape complexity of the preliminary merged area, and if it exceeds the preset complexity threshold, simplify the boundary of the area; Repeat the steps of calculating the intensity mean difference, determining the region pairs to be merged, executing the region growing algorithm and calculating the shape complexity until no further merging is possible or the preset maximum number of iterations is reached, and obtain the merged detection area.

4. The automated detection method for optical modules according to claim 3, characterized in that: The performing of a region growing algorithm on the initial detection regions to be merged to form a preliminary merged region comprises: Selecting an initial detection area whose intensity mean is higher than a preset third threshold as a seed area; Calculate the intensity mean similarity between the seed region and the adjacent initial detection region; If the similarity of the mean intensity of the adjacent regions exceeds a preset fourth threshold, the adjacent region is added to the seed region; Update the boundary and intensity mean of the merged region; The steps of calculating similarity, determining merging conditions, and updating regions are repeated until no adjacent regions satisfying the merging conditions can be found, thereby forming a preliminary merging region.

5. An automatic detection device for an optical module, characterized in that: The automatic detection device of the optical module adopts the automatic detection method of the optical module according to any one of claims 1 to 4, and the automatic detection device of the optical module comprises: A transport component (1), used for transporting the optical module; A first cleaning component (2), a second cleaning component (3) and a detection component (4) are sequentially arranged in the transport direction of the optical module, and at least one of the first cleaning component (2) and the second cleaning component (3) is configured to use a cleaning agent containing a fluorescent nano-marker to decontaminate the optical module port; The detection component (4) comprises a laser scanning unit (41), a signal collection unit (42), a spectrum analysis unit (43), a region division unit (44), an intensity calculation unit (45) and a determination unit (46); The laser scanning unit (41) is used to perform multi-wavelength laser scanning on the optical module port; The signal collection unit (42) is used to collect the fluorescence signal of the optical module port to obtain raw fluorescence data; The spectrum analysis unit (43) is used to perform spectrum analysis on the original fluorescence data to separate the fluorescent nano-marker signal and the background signal; The area division unit (44) is used to divide the optical module port into a plurality of detection areas; The intensity calculation unit (45) is used to calculate the average intensity of the fluorescent nano-marker signal in each detection area; The determination unit (46) is used to compare the average intensity of the fluorescent nano-marker signal in each detection area with the corresponding preset first threshold value, and the comparison process includes: if the average intensity of the fluorescent nano-marker signal in any detection area exceeds the corresponding preset first threshold value, it is determined that there is residual stain on the optical module port; if the average intensity of the fluorescent nano-marker signal in all detection areas is lower than or equal to the corresponding preset first threshold value, it is determined that the optical module port is cleaned.

6. The automatic detection device for optical modules according to claim 5, characterized in that: The first cleaning component (2) comprises a first driving member (21) and a water absorbing member (22), wherein the first driving member (21) is in driving connection with the water absorbing member (22) so as to drive the water absorbing member (22) to move within the optical module port, and a cleaning agent containing a fluorescent nano-marker is provided on the water absorbing member (22); The second cleaning component (3) comprises a second driving member (31), a nozzle (32) and a pump body component. The second driving member (31) is connected to the nozzle (32) in a transmission manner so as to drive the nozzle (32) to move into the optical module port. A fluid flow channel and an air port (321) connected to the fluid flow channel are formed on the nozzle (32). The pump body component is connected to the fluid flow so as to transport high-pressure pure fluid to the optical module port through the fluid flow channel and the air port (321), or to form a negative pressure environment in the fluid flow channel.

Images