Shielding cover defect detection system based on multispectral imaging

Abstract

The invention discloses a shielding cover defect detection system based on multi-spectral imaging, relates to the technical field of shielding cover defect detection, and effectively improves the accuracy of shielding cover defect detection; the invention sets the same data acquisition range for each multi-spectral imager so that the data acquisition ranges of each multi-spectral imager intersect with each other, thereby acquiring normal multi-spectral raw data and real-time multi-spectral raw data of a target shielding cover, mutually maps the normal multi-spectral raw data with an intersecting acquisition range to acquire an ambient light spectrum, then spectrally filters the normal multi-spectral raw data and the real-time multi-spectral raw data through the ambient light spectrum, establishes a normal state image model and a real-time state image model according to the spectrally filtered normal multi-spectral raw data and the real-time multi-spectral raw data, overlaps and maps the normal state image model and the real-time state image model, thereby judging whether the target shielding cover currently has a defect.

Description

Technical Field

[0001] The invention relates to the technical field of shielding cover defect detection, and in particular to a shielding cover defect detection system based on multi-spectral imaging. Background Art

[0002] Mobile phone shielding covers are commonly used accessories on mobile phone motherboards to shield chip interference and dissipate heat. They are formed by metal stamping, which may produce defective products during the manufacturing process. Through the shielding cover AOI defect detection equipment, the detection and packaging are automatically completed, and the detection and packaging are automated, which greatly improves the product quality and output and saves a lot of manpower.

[0003] The existing mobile phone shielding cover automatic detection technology has the following defects:

[0004] Defect diversity: There are various types of defects on mobile phone shielding covers, including bright cracks, dark cracks, black film crushing, black film scratches, etc. The characteristics and morphology of each defect are quite different. It is difficult for existing technologies to conduct targeted detection for different defect situations, resulting in low detection accuracy;

[0005] Material surface reflection and lighting conditions: Since the material of the mobile phone shielding cover may be reflective and uneven lighting may occur during the production process, this will affect the visibility of defects, resulting in low quality of collected data and affecting the accuracy of subsequent mobile phone shielding cover defect detection.

[0006] Therefore, how to improve the quality of status data of mobile phone shielding covers while improving the accuracy of defect detection is a difficulty in the prior art. For this purpose, a shielding cover defect detection system based on multispectral imaging is provided. Summary of the invention

[0007] In order to solve the above technical problems, an object of the present invention is to provide a shielding cover defect detection system based on multi-spectral imaging.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A shielding cover defect detection system based on multispectral imaging includes a cloud computing platform, wherein the cloud computing platform is communicatively connected with a shielding cover data acquisition module, a spectral filtering module and a defect detection module;

[0010] The shielding cover data acquisition module is used to set a plurality of multispectral imagers, set the same data acquisition range for each multispectral imager, and arrange the multispectral imagers around the target shielding cover at a fixed spatial distance, so that the data acquisition ranges of each multispectral imager intersect with each other, thereby collecting normal multispectral raw data and real-time multispectral raw data of the target shielding cover, and sending the normal multispectral raw data and real-time multispectral raw data to the spectral filtering module;

[0011] The spectral filtering module is used to extract spectral signal spectra of multiple different bands from normal multi-spectral raw data at different acquisition angles, map spectral signal spectra of the same band in normal multi-spectral raw data with cross-relationship in acquisition ranges to each other, and obtain ambient light spectra at different acquisition angles, and then perform spectral filtering on the corresponding normal multi-spectral raw data and real-time multi-spectral raw data through the ambient light spectra at different acquisition angles;

[0012] The defect detection module is used to establish a normal state image model and a real-time state image model of the target shielding cover for the normal multi-spectral raw data and the real-time multi-spectral raw data after spectral filtering, respectively, overlap and map the normal state image model and the real-time state image model, and then determine whether the target shielding cover currently has defects.

[0013] Furthermore, the setting process of the multispectral imager includes:

[0014] The shielding cover data acquisition module is provided with n multi-spectral imagers, and each multi-spectral imager is marked with a number s1, s2, s3, ..., s n , where n is a natural number greater than 0;

[0015] The same data collection range and fixed spatial distance are set for each multispectral imager, and then the multispectral imagers are distributed around the target shielding cover, and the spatial distance between the multispectral imagers at adjacent spatial positions is equal to the fixed spatial distance, and the data collection ranges of the multispectral imagers at adjacent spatial positions are crossed with each other;

[0016] The acquisition interface of each multi-spectral imager is perpendicular to the surface of the target shielding cover, and the vertical distances between each multi-spectral imager and the target shielding cover are equal.

[0017] Furthermore, the acquisition process of the normal multi-spectral raw data and the real-time multi-spectral raw data includes:

[0018] A data collection cycle is set. Whenever a data collection cycle starts, the shielding cover data collection module generates a data collection instruction and sends it to each multi-spectral imager at the same time. Then, each multi-spectral imager simultaneously sends spectral signals of several wavelengths to the target shielding cover, thereby collecting real-time multi-spectral raw data within the corresponding data collection range.

[0019] When the data collection cycle is over, each multispectral imager uploads the real-time multispectral raw data it has collected to the shielding cover data collection module, and then the shielding cover data collection module marks the corresponding multispectral imager number and collection time for each real-time multispectral raw data;

[0020] At the same time, the shielding cover data acquisition module collects normal multispectral raw data of the target shielding cover according to the process of collecting real-time multispectral raw data, and marks the corresponding multispectral imager number;

[0021] The normal multi-spectral raw data and the real-time multi-spectral raw data are respectively composed of a plurality of spectral signal spectra.

[0022] Furthermore, the process of generating the ambient light spectrum includes:

[0023] Establish several two-dimensional coordinate systems, and extract the spectrum signal spectrum at each wavelength from each normal multi-spectral raw data, map the spectrum signal spectrum of the same wavelength in the normal multi-spectral raw data with adjacent numbers onto the same two-dimensional coordinate system, and take the second half of the first normal spectrum signal spectrum on the two-dimensional coordinate system and the first half of the third spectrum signal spectrum on the two-dimensional coordinate system in the order of numbers according to the data acquisition overlap range between the multi-spectral imagers with adjacent numbers, and splice the spectrum signal spectrum of the above two parts to obtain the reference spectrum signal spectrum;

[0024] A plurality of coordinate intervals of equal length are set on the two-dimensional coordinate system, and then the reference spectral signal spectrum and the spectral signal spectrum with intermediate numbers are divided into reference spectral segments and spectral signal spectrum segments according to the two-dimensional plane range corresponding to the coordinate interval;

[0025] Map and compare the spectrum segments of the spectrum signals of each wavelength with the reference spectrum segments in each coordinate interval. The completely overlapping parts of the spectrum segments of the spectrum signals and the reference spectrum segments are automatically ignored. For the parts of the spectrum segments of the spectrum signals and the reference spectrum segments that cannot be completely overlapped, the interval parts composed of the curves that do not overlap between the two are retained. The various interval parts are arranged in the order of the corresponding coordinates, and then the difference spectrum intervals of the corresponding coordinate intervals are obtained.

[0026] The planes of the target shielding cover are numbered γ1, ..., γ6 in sequence, and then the spectral signals of the same wavelength in the normal multi-spectral raw data generated on the same plane of the target shielding cover are overlapped with each other in the difference spectrum intervals corresponding to the same coordinate interval, so as to obtain the ambient light spectrum of the spectral signals of the corresponding wavelengths under each plane of the target shielding cover, and mark the plane number of the target shielding cover.

[0027] Furthermore, the process of performing spectral filtering on the normal multi-spectral raw data and the real-time multi-spectral raw data includes:

[0028] According to the ambient light spectrum of the spectral signals of each wavelength under each plane of the target shielding cover, the spectrum is overlapped with the corresponding real-time spectral signal spectrum in the real-time multi-spectral raw data, and then the ambient light spectrum in the corresponding real-time spectral signal spectrum is eliminated;

[0029] The above process of removing the ambient light spectrum from the real-time spectrum signal spectrum is repeated, and then the normal multi-spectral raw data and the real-time multi-spectral raw data are spectrally filtered.

[0030] Furthermore, the process of establishing the real-time state image model and the normal state image model includes:

[0031] According to normal multi-spectral raw data and real-time multi-spectral raw data after filtering each spectrum, a partial normal state image model and a partial real-time state image model of the target shielding cover in the corresponding data acquisition range are established;

[0032] Arrange the normal state image models and the real-time state image models according to the serial number sequence of the corresponding multi-spectral imagers, and set the shielding cover frame model according to the actual size of the target shielding cover;

[0033] Inputting each part of the normal state image model and the part of the real-time state image model into the shielding cover frame model respectively, dividing a number of three-dimensional space pixels of equal size from the part of the normal state image model and the part of the real-time state image model, and graying each surface of the three-dimensional space pixel, thereby obtaining the pixel value of each surface of the three-dimensional space pixel, and marking it on each surface of the three-dimensional space pixel;

[0034] The pixel values ​​of each surface of the three-dimensional space pixels in the same position in the shielding cover frame model but coming from different parts of the normal state image model are blurred, and the three-dimensional space pixels after the blurring are sequentially spliced ​​to obtain the normal state image model. At the same time, the process of establishing the normal state image model is adopted to establish the real-time state image model according to the partial real-time state image model.

[0035] Furthermore, the calculation formula of the fuzzy processing is:

[0036] Where Num represents the number of normal state image models associated with the three-dimensional space pixels at any position in the shield frame model, η j represents the pixel value of the jth face pixel after blurring, β i，j Represents the pixel value of the jth surface of the three-dimensional space pixel corresponding to the i-th partial normal state image model, where j=1,...,6.

[0037] Furthermore, the process of determining whether the target shielding cover has defects includes:

[0038] Overlapping and mapping the real-time state image model and the normal state image model, and obtaining the difference rate of each three-dimensional space pixel in the real-time state image model and the normal state image model;

[0039] When cracks or dirt appear on the metal surface or content of the target shielding cover, the absorption and refraction of the spectrum at the corresponding position will change, and then a difference threshold will be set. The difference rate of each three-dimensional spatial pixel will be compared with the difference threshold, and whether the target shielding cover has defects will be determined based on the comparison results.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. By setting the same data acquisition range for each multispectral imager, the data acquisition ranges of each multispectral imager are crossed with each other, and then the normal multispectral raw data and real-time multispectral raw data of the target shielding cover are collected, and the normal multispectral raw data with a cross-relationship in the acquisition range are mapped to each other to obtain the ambient light spectrum, and then the normal multispectral raw data and the real-time multispectral raw data are spectrally filtered through the ambient light spectrum, thereby effectively improving the status data quality of the target shielding cover and providing a data basis for the subsequent defect detection process.

[0042] 2. The present invention establishes a normal state image model and a real-time state image model based on normal multi-spectral raw data and real-time multi-spectral raw data filtered by spectral filtering, and overlaps and maps the normal state image model and the real-time state image model to determine whether the target shielding cover currently has defects, thereby effectively improving the accuracy of defect detection. The present invention collects the state data of the target shielding cover through spectral signals of multiple sections with different wavelengths, thereby effectively improving the versatility of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention.

[0044] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0045] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0046] like Figure 1 As shown, a shielding cover defect detection system based on multi-spectral imaging includes a cloud computing platform, and the cloud computing platform is communicatively connected with a shielding cover data acquisition module, a spectral filtering module and a defect detection module;

[0047] The shielding cover data acquisition module is used to set a plurality of multispectral imagers, set the same data acquisition range for each multispectral imager, arrange the multispectral imagers around the target shielding cover at a fixed spatial distance, so that the data acquisition ranges of each multispectral imager intersect with each other, set a data acquisition cycle, and then collect normal multispectral raw data of the target shielding cover and real-time multispectral raw data under each data acquisition cycle, and send the normal multispectral raw data and the real-time multispectral raw data to the spectral filtering module;

[0048] The spectral filtering module is used to extract spectral signal spectra of multiple different bands from normal multi-spectral raw data at different acquisition angles, map spectral signal spectra of the same band in normal multi-spectral raw data with cross-relationship in acquisition ranges to each other, and obtain ambient light spectra at different acquisition angles, and then perform spectral filtering on the corresponding normal multi-spectral raw data and real-time multi-spectral raw data through the ambient light spectra at different acquisition angles;

[0049] The defect detection module is used to establish a normal state image model and a real-time state image model of the target shielding cover for the normal multi-spectral raw data and the real-time multi-spectral raw data after spectral filtering, respectively, overlap and map the normal state image model and the real-time state image model, and then determine whether the target shielding cover currently has defects.

[0050] Further, the working principle of the present invention is described below by way of examples:

[0051] The shielding cover data acquisition module is provided with n multi-spectral imagers, and each multi-spectral imager is marked with a number s1, s2, s3, ..., s n , where n is a natural number greater than 0;

[0052] The same data collection range and fixed spatial distance are set for each multispectral imager, and then the multispectral imagers are distributed around the target shielding cover, and the spatial distance between the multispectral imagers at adjacent spatial positions is equal to the fixed spatial distance, and the data collection ranges of the multispectral imagers at adjacent spatial positions are crossed with each other;

[0053] For example, if the data collection ranges of the multispectral imagers numbered s1, s2, and s3 on the target shield are all 10 square centimeters, then 50% of the data collection ranges of the multispectral imagers numbered s1 and s3 overlap with the data collection range of the multispectral imager numbered s2;

[0054] It should be noted that the acquisition interface of each multi-spectral imager is perpendicular to the surface of the target shielding cover, and the vertical distance between each multi-spectral imager and the target shielding cover is equal;

[0055] A data collection cycle is set. Whenever a data collection cycle starts, the shielding cover data collection module generates a data collection instruction and sends it to each multi-spectral imager at the same time. Then, each multi-spectral imager simultaneously sends spectral signals of several wavelengths to the target shielding cover, thereby collecting real-time multi-spectral raw data within the corresponding data collection range.

[0056] When the data collection cycle is over, each multispectral imager uploads the real-time multispectral raw data it has collected to the shielding cover data collection module, and then the shielding cover data collection module marks the corresponding multispectral imager number and collection time for each real-time multispectral raw data;

[0057] At the same time, the shielding cover data acquisition module collects normal multispectral raw data of the target shielding cover according to the process of collecting real-time multispectral raw data, and marks the corresponding multispectral imager number;

[0058] It should be noted that the normal multi-spectral raw data and the real-time multi-spectral raw data are respectively composed of a plurality of spectral signal spectra, and the types and quantities of the spectral signal spectra contained in each normal multi-spectral raw data and the real-time multi-spectral raw data are the same.

[0059] Further, the shielding cover data acquisition module sends the normal multi-spectral raw data and the real-time multi-spectral raw data to the spectral filtering module;

[0060] Since the multispectral imagers are located directly above the target shielding cover during the generation of normal multispectral raw data and real-time multispectral raw data, each multispectral imager will generate a corresponding shadow part on the target shielding cover, so that the spectrum of the spectral signal contained in the corresponding generated normal multispectral raw data and real-time multispectral raw data will be affected by the ambient light;

[0061] The spectral filtering module establishes several two-dimensional coordinate systems and extracts the spectrum of spectral signals at various wavelengths from each normal multi-spectral raw data;

[0062] The spectrum signal spectra of the same wavelength in the normal multi-spectral raw data with adjacent numbers are mapped onto the same two-dimensional coordinate system, and according to the data acquisition overlap range between the multi-spectral imagers with adjacent numbers, the second half of the first normal spectrum signal spectrum on the two-dimensional coordinate system is taken in the order of numbers, and the first half of the third spectrum signal spectrum on the two-dimensional coordinate system is taken, and the spectrum signal spectra of the above two parts are spliced, and the spliced ​​spectrum signal spectrum is recorded as the reference spectrum signal spectrum;

[0063] It should be noted that, since there is space between each multispectral imager and it is perpendicular to the top of the target shielding cover, the shadows generated between the multispectral imagers with adjacent numbers affect each other. For example, among three multispectral imagers with adjacent numbers, the first and third multispectral imagers have the same degree of shadow influence on the middle multispectral imager, and the shadow influence on the data acquisition range is different relative to the middle multispectral imager.

[0064] A plurality of coordinate intervals of equal length are set on the two-dimensional coordinate system, and then the reference spectral signal spectrum and the spectral signal spectrum with intermediate numbers are divided into reference spectral segments and spectral signal spectrum segments according to the two-dimensional plane range corresponding to the coordinate interval;

[0065] Map and compare the spectrum segments of the spectrum signals of each wavelength with the reference spectrum segments in each coordinate interval. The completely overlapping parts of the spectrum segments of the spectrum signals and the reference spectrum segments are automatically ignored. For the parts of the spectrum segments of the spectrum signals and the reference spectrum segments that cannot be completely overlapped, the interval parts composed of the curves that do not overlap between the two are retained. The various interval parts are arranged in the order of the corresponding coordinates, and then the difference spectrum intervals of the corresponding coordinate intervals are obtained.

[0066] Since spectral signals of different wavelengths are affected by ambient light to different degrees, and the ambient light influence on normal multi-spectral raw data and real-time multi-spectral raw data generated by the multi-spectral imager on the same plane of the target shield is similar;

[0067] The planes of the target shielding cover are numbered γ1, ..., γ6 in sequence, and then the spectral signals of the same wavelength in the normal multi-spectral raw data generated on the same plane of the target shielding cover are overlapped with each other in the difference spectrum intervals corresponding to the same coordinate interval, so as to obtain the ambient light spectrum of the spectral signals of the corresponding wavelengths under each plane of the target shielding cover, and mark the plane number of the target shielding cover.

[0068] Furthermore, according to the ambient light spectrum of the spectral signals of each wavelength under each plane of the target shielding cover, the spectrum is overlapped with the corresponding real-time spectral signal spectrum in the real-time multi-spectral raw data, and then the ambient light spectrum in the corresponding real-time spectral signal spectrum is eliminated;

[0069] The above process of removing the ambient light spectrum from the real-time spectral signal spectrum is repeated, and then the normal multi-spectral raw data and the real-time multi-spectral raw data are spectrally filtered, and the normal multi-spectral raw data and the real-time multi-spectral raw data after spectrally filtering are sent to the defect detection module.

[0070] Furthermore, since the constituent materials of the target shielding cover have different reflection, absorption and transmission characteristics for light of the same wavelength, and any substance has different reflection, absorption and transmission characteristics for light of different wavelengths, the material composition structure of the target shielding cover can be obtained through the frequency spectrum arrangement order and amplitude status of the spectral signal of each wavelength in each multi-spectral raw data in the multi-spectral raw data;

[0071] Then, according to the normal multi-spectral raw data and the real-time multi-spectral raw data after filtering each spectrum, a partial normal state image model and a partial real-time state image model of the target shielding cover in the corresponding data acquisition range are established;

[0072] Since the data collection ranges of each multispectral imager are cross-related, and the normal multispectral raw data and real-time multispectral raw data generated by different plane target shields are also correlated;

[0073] Therefore, each part of the normal state image model and the part of the real-time state image model are arranged according to the serial number sequence of the corresponding multi-spectral imager, and the shielding cover frame model is set according to the actual size of the target shielding cover;

[0074] Inputting each part of the normal state image model and the part of the real-time state image model into the shielding cover frame model respectively, dividing a number of three-dimensional space pixels of equal size from the part of the normal state image model and the part of the real-time state image model, and graying each surface of the three-dimensional space pixel, thereby obtaining the pixel value of each surface of the three-dimensional space pixel, and marking it on each surface of the three-dimensional space pixel;

[0075] The pixel values ​​of each surface of the three-dimensional space pixels from different parts of the normal state image model at the same position in the shielding cover frame model are fuzzy processed, wherein the calculation formula of the fuzzy processing is: Where Num represents the number of normal state image models associated with the three-dimensional space pixels at any position in the shield frame model, η j represents the pixel value of the jth face pixel after blurring, β i，j represents the pixel value of the jth face of the three-dimensional space pixel corresponding to the i-th partial normal state image model, where j=1, ..., 6;

[0076] The three-dimensional pixels after the fuzzy processing are sequentially spliced ​​to obtain a normal state image model. At the same time, the process of establishing the normal state image model is adopted to establish a real-time state image model according to the partial real-time state image model.

[0077] Furthermore, the real-time state image model and the normal state image model are overlapped and mapped, and the difference rate of each three-dimensional space pixel in the real-time state image model and the normal state image model is obtained, wherein the calculation formula of the difference rate is: where p k and P k They represent the pixel rate of the kth face of a three-dimensional space pixel in the real-time state image model and the normal state image model respectively;

[0078] When the metal surface or content of the target shielding cover is cracked or dirty, the absorption and refraction of the spectrum at the corresponding position will change, and then the difference threshold will be set, and the difference rate of each three-dimensional space pixel will be compared with the difference threshold;

[0079] If the difference rate is less than the difference threshold, no action is taken;

[0080] If the difference rate is greater than or equal to the difference threshold, it is determined that the target shield portion corresponding to the three-dimensional space pixel is abnormal, and a label is set on the corresponding three-dimensional space pixel of the real-time state image model.

[0081] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A shielding cover defect detection system based on multispectral imaging, including a cloud computing platform, characterized in that: The cloud computing platform is communicatively connected with a shielding cover data acquisition module, a spectrum filtering module and a defect detection module; The shielding cover data acquisition module is used to set a plurality of multispectral imagers, set the same data acquisition range for each multispectral imager, and arrange the multispectral imagers around the target shielding cover at a fixed spatial distance, so that the data acquisition ranges of each multispectral imager intersect with each other, thereby acquiring normal multispectral raw data and real-time multispectral raw data of the target shielding cover; The shielding cover data acquisition module is provided with n multi-spectral imagers, and each multi-spectral imager is marked with a number s1, s2, s3, ..., s n , where n is a natural number greater than 0; The same data collection range and fixed spatial distance are set for each multispectral imager, and then the multispectral imagers are distributed around the target shielding cover, and the spatial distance between the multispectral imagers at adjacent spatial positions is equal to the fixed spatial distance, and the data collection ranges of the multispectral imagers at adjacent spatial positions are crossed with each other; The acquisition interface of each multi-spectral imager is perpendicular to the surface of the target shielding cover, and the vertical distance between each multi-spectral imager and the target shielding cover is equal; The spectral filtering module is used to extract spectral signal spectra of multiple different bands from normal multi-spectral raw data at different acquisition angles, map spectral signal spectra of the same band in normal multi-spectral raw data with cross-relationship in acquisition ranges to each other, and obtain ambient light spectra at different acquisition angles, and then perform spectral filtering on the corresponding normal multi-spectral raw data and real-time multi-spectral raw data through the ambient light spectra at different acquisition angles; The generation process of the ambient light spectrum includes: Establish several two-dimensional coordinate systems, extract the spectrum signal spectrum at each wavelength from each normal multi-spectral raw data, map the spectrum signal spectrum of the same wavelength in the normal multi-spectral raw data with adjacent numbers onto the same two-dimensional coordinate system, and take the second half of the first normal spectrum signal spectrum on the two-dimensional coordinate system and the first half of the third spectrum signal spectrum on the two-dimensional coordinate system in the order of numbers according to the data acquisition overlap range between the multi-spectral imagers with adjacent numbers, and splice the spectrum signal spectrum of the above two parts to obtain the reference spectrum signal spectrum; A plurality of coordinate intervals of equal length are set on the two-dimensional coordinate system, and then the reference spectral signal spectrum and the spectral signal spectrum with intermediate numbers are divided into reference spectral segments and spectral signal spectrum segments according to the two-dimensional plane range corresponding to the coordinate interval; Map and compare the spectrum segments of the spectrum signals of each wavelength with the reference spectrum segments in each coordinate interval. The completely overlapping parts of the spectrum segments of the spectrum signals and the reference spectrum segments are automatically ignored. For the parts of the spectrum segments of the spectrum signals and the reference spectrum segments that cannot be completely overlapped, the interval parts composed of the curves that do not overlap between the two are retained. The various interval parts are arranged in the order of the corresponding coordinates, and then the difference spectrum intervals of the corresponding coordinate intervals are obtained. Numbers γ1, ..., γ6 are sequentially set for each plane of the target shielding cover, and then the difference spectrum intervals corresponding to the same coordinate interval of the spectrum signal of the same wavelength in the normal multi-spectral raw data generated on the same plane of the target shielding cover are overlapped with each other, and then the ambient light spectrum of the spectrum signal of the corresponding wavelength under each plane of the target shielding cover is obtained, and the plane number of the target shielding cover is marked; The process of performing spectral filtering on the normal multi-spectral raw data and the real-time multi-spectral raw data includes: According to the ambient light spectrum of the spectral signals of each wavelength under each plane of the target shielding cover, the spectrum is overlapped with the corresponding real-time spectral signal spectrum in the real-time multi-spectral raw data, and then the ambient light spectrum in the corresponding real-time spectral signal spectrum is eliminated; Repeat the above process of removing the ambient light spectrum from the real-time spectrum signal spectrum, and then perform spectrum filtering on the normal multi-spectral raw data and the real-time multi-spectral raw data; The defect detection module is used to establish a normal state image model and a real-time state image model of the target shielding cover for the normal multi-spectral raw data and the real-time multi-spectral raw data after spectral filtering, respectively, overlap and map the normal state image model and the real-time state image model, and then determine whether the target shielding cover currently has defects.

2. The shielding cover defect detection system based on multispectral imaging according to claim 1 is characterized in that: The acquisition process of the normal multispectral raw data and the real-time multispectral raw data includes: A data acquisition cycle is set. Whenever a data acquisition cycle starts, the shielding cover data acquisition module generates a data acquisition instruction and sends it to each multi-spectral imager. Each multi-spectral imager simultaneously sends spectral signals of several wavelengths to the target shielding cover, thereby generating real-time multi-spectral raw data within the corresponding data acquisition range. When the data collection cycle is over, each multispectral imager uploads the real-time multispectral raw data it has collected to the shielding cover data collection module, and then the shielding cover data collection module marks the corresponding multispectral imager number and collection time for each real-time multispectral raw data; At the same time, the shielding cover data acquisition module collects normal multispectral raw data of the target shielding cover according to the process of collecting real-time multispectral raw data, and marks the corresponding multispectral imager number; The normal multi-spectral raw data and the real-time multi-spectral raw data are respectively composed of a plurality of spectral signal spectra.

3. The shielding cover defect detection system based on multispectral imaging according to claim 1 is characterized in that: The process of establishing the real-time state image model and the normal state image model includes: According to normal multi-spectral raw data and real-time multi-spectral raw data after filtering each spectrum, a partial normal state image model and a partial real-time state image model of the target shielding cover in the corresponding data acquisition range are established; Arrange the normal state image models and the real-time state image models according to the serial number sequence of the corresponding multi-spectral imagers, and set the shielding cover frame model according to the actual size of the target shielding cover; Inputting each part of the normal state image model and the part of the real-time state image model into the shielding cover frame model respectively, dividing a number of three-dimensional space pixels of equal size from the part of the normal state image model and the part of the real-time state image model, and graying each surface of the three-dimensional space pixel, thereby obtaining the pixel value of each surface of the three-dimensional space pixel, and marking it on each surface of the three-dimensional space pixel; The pixel values ​​of each surface of the three-dimensional space pixels in the same position in the shielding cover frame model but coming from different parts of the normal state image model are blurred, and the three-dimensional space pixels after the blurring are sequentially spliced ​​to obtain the normal state image model. At the same time, the process of establishing the normal state image model is adopted to establish the real-time state image model according to the partial real-time state image model.

4. The shielding cover defect detection system based on multispectral imaging according to claim 3 is characterized in that: The calculation formula for the fuzzy processing is: , where Num represents the number of partial normal state image models associated with the three-dimensional space pixels at any position in the shield frame model, represents the pixel value of the jth face pixel after blurring, Represents the pixel value of the jth surface of the three-dimensional space pixel corresponding to the i-th partial normal state image model, where j=1,...,6.

5. The shielding cover defect detection system based on multispectral imaging according to claim 4 is characterized in that: The process of determining whether the target shield is defective includes: Overlapping and mapping the real-time state image model and the normal state image model, and obtaining the difference rate of each three-dimensional space pixel in the real-time state image model and the normal state image model; Then, a difference threshold is set, and the difference rate of each three-dimensional space pixel is compared with the difference threshold, and whether the target shielding cover has defects is determined based on the comparison result.

Images