A printed circuit board defect detection system based on computer vision

Abstract

The present invention discloses a printed circuit board defect detection system based on computer vision, relates to the technical field of defect detection, and improves the accuracy of the defect detection process. The present invention extracts a defect ultrasonic signal reflection spectrum from a test ultrasonic reflection signal spectrum, establishes a test three-dimensional circuit board model according to the test ultrasonic reflection signal spectrum, and maps the corresponding generated characteristic defect ultrasonic reflection signal spectrum at different angles in a three-dimensional coordinate system, thereby obtaining a characteristic defect signal distribution network, maps the test three-dimensional circuit board model to the corresponding characteristic defect signal distribution network, thereby obtaining characteristic reflection signal spectrum identification intervals of different defect types at different acquisition angles at various positions, compares and matches the target ultrasonic reflection signal spectrum with the corresponding standard ultrasonic reflection signal spectrum and the characteristic reflection signal spectrum identification interval in turn, thereby determining the defect position and type of the target printed circuit board.

Description

Technical Field

[0001] The invention relates to the technical field of defect detection, and in particular to a printed circuit board defect detection system based on computer vision. Background Art

[0002] With the rapid development of the electronics industry, printed circuit boards, as an indispensable basic component of electronic equipment, have a direct impact on the performance and reliability of the entire electronic product. In order to ensure the quality of PCBs, defect detection has become a crucial part of the production process.

[0003] The existing printed circuit board defect detection technology field has the following defects:

[0004] Poor adaptability: PCBs of different types and specifications have different characteristics and defect types. Existing systems are often optimized for specific types of PCBs, lack versatility and adaptability, and are unable to cope with diverse inspection needs.

[0005] Strong dependence on data sets: Many machine learning-based detection methods require a large amount of labeled data for training. However, the acquisition and labeling costs of PCB defect data sets are high, and the defect characteristics in different production environments may vary greatly, resulting in insufficient generalization capabilities of the model.

[0006] Therefore, how to improve the diversity of the defect detection process while improving the accuracy of defect detection at different test angles is a difficulty in the prior art. To this end, a printed circuit board defect detection system based on computer vision 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 printed circuit board defect detection system based on computer vision.

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

[0009] A printed circuit board defect detection system based on computer vision includes a cloud computing platform, wherein the cloud computing platform is communicatively connected with a test data acquisition module, a defect feature data analysis module and a defect detection module;

[0010] The test data acquisition module is used to install several groups of phased array probes on the surface of the test printed circuit board and the standard printed circuit board, and each group of phased array probes transmits multiple groups of ultrasonic beams to the target printed circuit board, thereby collecting several test ultrasonic reflection signal spectra and standard ultrasonic reflection signal spectra at different angles;

[0011] The defect feature data analysis module is used to filter the standard ultrasonic reflection signal spectrum and the test ultrasonic reflection signal spectrum, and to separate the defect ultrasonic signal reflection spectrum from the test ultrasonic reflection signal spectrum, and then to establish a test three-dimensional circuit board model according to the test ultrasonic reflection signal spectrum;

[0012] Then, several defect positions are marked on the three-dimensional circuit board model according to the defect ultrasonic signal reflection spectrum, and the characteristic defect ultrasonic reflection signal spectrum generated at different angles is mapped in the three-dimensional coordinate system to obtain the characteristic defect signal distribution network;

[0013] The three-dimensional circuit board model to be tested is mapped to the corresponding characteristic defect signal distribution network, thereby obtaining the characteristic reflection signal spectrum identification interval of different defect types at different acquisition angles at each position;

[0014] The defect detection module is used to collect the target ultrasonic reflection signal spectrum at different angles at various positions of the target printed circuit board, compare and match the target ultrasonic reflection signal spectrum with the corresponding standard ultrasonic reflection signal spectrum and the characteristic reflection signal spectrum identification interval in turn, and then determine the defect position and type of the target printed circuit board.

[0015] Furthermore, the acquisition process of the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum includes:

[0016] Several test printed circuit boards and a standard printed circuit board are set, and several groups of phased array probes are installed on the front and back sides of the test printed circuit board and the standard printed circuit board, wherein the phased array probes are composed of four groups of ultrasonic transmitting devices and ultrasonic receiving devices, and each phased array probe is numbered;

[0017] Setting a number of data collection periods, placing the test printed circuit board and the standard printed circuit board in the same test environment, so that within the same data collection period, multiple ultrasonic signals emitted at different angles pass through various positions of the printed circuit board;

[0018] When all data collection periods are completed, the ultrasonic reflection signal spectra generated by each phased array probe are integrated to generate corresponding test signal spectrum data sets and standard test signal spectrum data sets, and the corresponding phased array probe numbers are marked.

[0019] Furthermore, the process of filtering the standard ultrasonic reflection signal spectrum and the test ultrasonic reflection signal spectrum includes:

[0020] A two-dimensional coordinate system is established, and all standard ultrasonic reflection signal spectra are extracted from each standard test signal spectrum data set, and the standard ultrasonic reflection signal spectra generated by the phased array probes corresponding to the front and back sides of the standard printed circuit board are mapped onto the same two-dimensional coordinate system;

[0021] The standard ultrasonic reflection signal spectrum generated by the phased array probe on the back side is reversed, and the standard ultrasonic reflection signal spectrum generated by the front side and the reversed standard ultrasonic reflection signal spectrum on the back side are subtracted at opposite transmission angles by the ultrasonic emitting devices corresponding to the front and back sides, so as to obtain the background signal of each data collection period;

[0022] Each background signal is inverted, and then the inverted background signal is subtracted from the test ultrasonic reflection signal spectrum generated on the back of each test printed circuit board during the corresponding data collection period, and the test ultrasonic reflection signal spectrum generated on the front is subtracted, thereby completing the filtering of the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum in each data collection period.

[0023] Furthermore, the generation process of the defect ultrasonic signal reflection spectrum includes:

[0024] According to the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum, a corresponding local test two-dimensional circuit board image and a local standard two-dimensional circuit board image are established;

[0025] A plurality of image regions of the same size are respectively divided in the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, and then the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image corresponding to the same position and the same angle are overlapped and mapped, and then the pixel values ​​of each pixel in the overlapped image region are compared to determine whether there is a difference between the two corresponding image regions;

[0026] According to the judgment result, the corresponding local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, the corresponding test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum are decreased in chronological order, and the corresponding defect ultrasonic signal reflection spectrum is obtained according to the decrease result.

[0027] Furthermore, the generation process of the test three-dimensional circuit board model includes:

[0028] The defect ultrasonic signal reflection spectrum of each defect is generated by combining the defect ultrasonic signal reflection spectrum with its adjacent position, and the characteristic defect ultrasonic reflection signal spectrum of different angles at the position of each phased array probe is obtained in descending order of time;

[0029] A three-dimensional coordinate system is established, and the spectrum of ultrasonic reflection signals of each characteristic defect is mapped onto the three-dimensional coordinate system to obtain a characteristic defect signal distribution network of the corresponding test printed circuit board;

[0030] According to the local test two-dimensional circuit board image generated by the same phased array probe, a local test three-dimensional circuit board model of the corresponding test printed plate part is generated, and according to the distribution of each phased array probe, each local test three-dimensional circuit board model is spliced ​​in sequence to obtain a test three-dimensional circuit board model of the corresponding test circuit board, and the corresponding defect position is marked on the test three-dimensional circuit board model according to the spectrum of the ultrasonic reflection signal of each characteristic defect.

[0031] Furthermore, the process of generating the characteristic reflection signal spectrum identification interval includes:

[0032] The three-dimensional circuit board model to be tested is mapped in the characteristic defect signal distribution network, and the characteristic defect ultrasonic reflection signal spectra corresponding to the same defect position but the same acquisition angle are arranged in parallel according to the depth, and a number of depth points are set on the coordinate axis corresponding to the depth, and the spectrum of each characteristic defect ultrasonic reflection signal is divided into a number of spectrum points through the depth points;

[0033] Perform a normal distribution on each spectrum point at the same depth point, obtain the standard reflection signal spectrum interval at each depth point according to the normal distribution result, and sequentially splice the standard reflection signal spectrum intervals at each depth point to obtain the characteristic reflection signal spectrum identification intervals of the corresponding position and defect type at different angles.

[0034] Furthermore, the process of determining whether the target printed circuit board is abnormal includes:

[0035] Each phased array probe collects the target ultrasonic reflection signal spectrum at different angles at its location, sets the deviation amplitude threshold and the abnormal node number threshold, and then reduces the target ultrasonic reflection signal spectrum and the corresponding standard ultrasonic reflection signal spectrum in chronological order to obtain the target ultrasonic reflection signal detection spectrum, and judges the relationship between the amplitude value of the target ultrasonic reflection signal detection spectrum at each time node and the deviation amplitude threshold, and judges whether the target printed circuit board has an abnormality based on the judgment result.

[0036] Furthermore, the process of determining the defect location and type of the target printed circuit board includes:

[0037] The target ultrasonic reflection signal detection spectrum is decremented from the target ultrasonic reflection signal detection spectrum at the adjacent acquisition position, and then a plurality of target characteristic ultrasonic reflection signal spectra are divided from the target ultrasonic reflection signal detection spectrum;

[0038] Then, the spectrum of each target characteristic ultrasonic reflection signal is matched with the characteristic reflection signal spectrum identification interval at the corresponding position and different angles, and the defect type and defect depth of each position of the target printed board are obtained according to the matching results;

[0039] According to the target ultrasonic reflection signal spectrum at different angles of the target printed circuit board, the corresponding target printed circuit board model is established, and according to the defect types and defect depths at various positions of the target printed circuit board, the defect positions are filled and marked at the corresponding positions of the target printed circuit board model.

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

[0041] The present invention extracts the defect ultrasonic signal reflection spectrum from the test ultrasonic reflection signal spectrum, establishes a test three-dimensional circuit board model according to the test ultrasonic reflection signal spectrum, maps the corresponding characteristic defect ultrasonic reflection signal spectrum generated at different angles in a three-dimensional coordinate system to obtain a characteristic defect signal distribution network, maps the test three-dimensional circuit board model to the corresponding characteristic defect signal distribution network, and then obtains the characteristic reflection signal spectrum recognition interval of different defect types at various positions at different acquisition angles, compares and matches the target ultrasonic reflection signal spectrum with the corresponding standard ultrasonic reflection signal spectrum and the characteristic reflection signal spectrum recognition interval in turn, and then determines the defect position and type of the target printed circuit board, thereby improving the accuracy of defect detection at different test angles while improving the diversity of the defect detection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0043] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0044] like Figure 1 As shown, a printed circuit board defect detection system based on computer vision includes a cloud computing platform, wherein the cloud computing platform is communicatively connected with a test data acquisition module, a defect feature data analysis module, and a defect detection module;

[0045] The test data acquisition module is used to install several groups of phased array probes on the surface of the test printed circuit board and the standard printed circuit board, and each group of phased array probes transmits multiple groups of ultrasonic beams to the target printed circuit board, thereby collecting several test ultrasonic reflection signal spectra and standard ultrasonic reflection signal spectra at different angles;

[0046] The defect feature data analysis module is used to filter the standard ultrasonic reflection signal spectrum and the test ultrasonic reflection signal spectrum, extract the defect ultrasonic signal reflection spectrum from the test ultrasonic reflection signal spectrum, and establish a test three-dimensional circuit board model according to the test ultrasonic reflection signal spectrum;

[0047] Then, several defect positions are marked on the three-dimensional circuit board model according to the defect ultrasonic signal reflection spectrum, and the characteristic defect ultrasonic reflection signal spectrum generated at different angles is mapped in the three-dimensional coordinate system to obtain the characteristic defect signal distribution network;

[0048] The three-dimensional circuit board model to be tested is mapped to the corresponding characteristic defect signal distribution network, thereby obtaining the characteristic reflection signal spectrum identification interval of different defect types at different acquisition angles at each position;

[0049] The defect detection module is used to collect the target ultrasonic reflection signal spectrum at different angles at various positions of the target printed circuit board, compare and match the target ultrasonic reflection signal spectrum with the corresponding standard ultrasonic reflection signal spectrum and the characteristic reflection signal spectrum identification interval in turn, and then determine the defect position and type of the target printed circuit board.

[0050] Further, the working principle of the present invention is specifically described below through embodiments:

[0051] Several test printed circuit boards and a standard printed circuit board are provided, wherein the test printed circuit boards have defects such as breakage, inclusions, cracks, etc., and the defects and defect positions of the test printed circuit boards are not exactly the same;

[0052] The standard printed circuit board refers to a printed circuit board without any defects;

[0053] i sets of phased array probes are installed on the front and back sides of the test printed circuit board and the standard printed circuit board, wherein the phased array probes are composed of four sets of ultrasonic transmitting devices and ultrasonic receiving devices, each phased array probe is provided with a fixed data collection range, and the data collection ranges of the phased array probes at adjacent positions partially overlap;

[0054] In the phased array probe, the transmitting angle and receiving angle of each group of ultrasonic transmitting devices and ultrasonic receiving devices are different, and the installation positions of each phased array probe of each test printed circuit board and the standard printed circuit board are the same, and i is a natural number greater than 0;

[0055] The test data acquisition module communicates with the phased array probes on each test printed circuit board and the standard printed circuit board, and sets a number a for each phased array probe 1，1 、a 1，2 ,……,am，j , where m and n are natural numbers greater than 0;

[0056] Several data collection periods are set, the test printed circuit board and the standard printed circuit board are placed in the same test environment, and fixed transmission ranges and reception ranges are set for each ultrasonic transmitter and ultrasonic receiver in the phased array probe, so that the ultrasonic signal transmission angles of each ultrasonic transmitter in the phased array probe in the same data collection period are staggered, so that for any position of the test printed circuit board and the standard printed circuit board, multiple ultrasonic signals transmitted at different angles pass through in the same data collection period;

[0057] It should be noted that, in the same data collection period, the ultrasonic signal frequencies emitted by each ultrasonic emitting device in the same phased array probe are different, and the edge positions of the ultrasonic signals emitted by the ultrasonic emitting devices at adjacent positions partially overlap, and the data collection process of the phased array probes in the phased array probes located on the front and back sides of the test printed circuit board and the standard printed circuit board is in an alternating state, that is, in the same data collection period, when the front phased array probe is working, the back phased array probe is in a standby state, and the data collection time of the two is equal;

[0058] When all data collection periods are over, the test data collection module collects data from each ultrasonic receiving device to generate a test ultrasonic reflection signal spectrum and a standard ultrasonic reflection signal spectrum, integrates the ultrasonic reflection signal spectra of the same phased array probe to generate corresponding test signal spectrum data sets and standard test signal spectrum data sets, and marks the corresponding phased array probe number.

[0059] Further, the test data acquisition module sends all test signal spectrum data sets and standard test signal spectrum data sets to the defect feature data analysis module;

[0060] The defect feature data analysis module establishes a two-dimensional coordinate system and extracts all standard ultrasonic reflection signal spectra from each standard test signal spectrum data set. The standard ultrasonic reflection signal spectra generated by the phased array probes corresponding to the front and back sides of the standard printed circuit board are mapped onto the same two-dimensional coordinate system.

[0061] The spectrum of the standard ultrasonic reflection signal generated by the phased array probe on the back side is reversed, that is, the amplitude value at each time node is taken opposite, and then the spectrum of the standard ultrasonic reflection signal generated by the front side and the reversed spectrum of the standard ultrasonic reflection signal on the back side are subtracted at opposite transmitting angles by the ultrasonic transmitting devices corresponding to the front and back sides, so as to obtain the background signal of each data collection period;

[0062] Each background signal is inverted, and then the inverted background signal is subtracted from the test ultrasonic reflection signal spectrum generated on the back of each test printed circuit board during the corresponding data collection period, and the test ultrasonic reflection signal spectrum generated on the front is subtracted, thereby completing the filtering of the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum in each data collection period.

[0063] Further, the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum generated at the same position and the same acquisition angle are mapped onto the same two-dimensional coordinate system, and at the same time, a corresponding local test two-dimensional circuit board image and a local standard two-dimensional circuit board image are established according to the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum;

[0064] A plurality of image regions of the same size are respectively divided in the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, and then the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image corresponding to the same position and the same angle are overlapped and mapped, and a pixel difference threshold and a difference quantity threshold are set, and then the pixel values ​​of each pixel in the overlapping image region are compared to determine whether there is a difference between the corresponding two image regions;

[0065] If the pixel value difference of the pixels at the same position in the two overlapping image regions is greater than the pixel difference threshold, then it is judged that the pixels at the corresponding positions are different, otherwise they are judged to be the same;

[0066] The total number of pixels with differences is counted as the difference pixel amount. If the difference pixel amount is less than the difference amount threshold, it is determined that there is no difference between the corresponding two image regions.

[0067] If the number of difference pixels is greater than or equal to the difference number threshold, it is determined that there is a difference between the corresponding two image areas, that is, there is a defect at the position of the test printed circuit board corresponding to the corresponding test ultrasonic reflection signal spectrum;

[0068] Then, the corresponding local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, the corresponding test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum are decreased in chronological order, and the corresponding defect ultrasonic signal reflection spectrum is obtained according to the decreasing result.

[0069] Further, the process of obtaining the defect ultrasonic signal reflection spectrum is repeated to obtain the defect ultrasonic signal reflection spectrum in each test ultrasonic reflection signal spectrum;

[0070] Since the acquisition angle and acquisition position of each test ultrasonic reflection signal spectrum are different, each test ultrasonic reflection signal spectrum records the defect data of the position where the adjacent phased array probe is located;

[0071] Then, the defect ultrasonic signal reflection spectrum of each defect is generated by combining the defect ultrasonic signal reflection spectrum with its adjacent position, and characteristic defect ultrasonic reflection signal spectra of different angles at the position of each phased array probe are obtained in descending order of time;

[0072] A three-dimensional coordinate system is established, wherein the three axes of the three-dimensional coordinate system represent time, depth, and amplitude respectively, and then the spectrum of each characteristic defect ultrasonic reflection signal is mapped onto the three-dimensional coordinate system according to the position of the corresponding phased array probe and the shooting angle to obtain a characteristic defect signal distribution network of the corresponding test printed circuit board;

[0073] Generate a local test three-dimensional circuit board model of the corresponding test printed plate part according to the local test two-dimensional circuit board image generated by the same phased array probe, and splice each local test three-dimensional circuit board model in sequence according to the distribution of each phased array probe, so as to obtain a test three-dimensional circuit board model of the corresponding test circuit board, and mark the corresponding defect position on the test three-dimensional circuit board model according to the spectrum of each characteristic defect ultrasonic reflection signal;

[0074] The three-dimensional circuit board model to be tested is mapped into the characteristic defect signal distribution network, since the characteristic defect ultrasonic reflection signal spectrum in each characteristic defect signal distribution network is in a one-to-many correspondence with the defect position on the three-dimensional circuit board model to be tested;

[0075] Therefore, the characteristic defect ultrasonic reflection signal spectra corresponding to the same defect position but the same acquisition angle are arranged in parallel according to the depth, and a number of depth points are set on the coordinate axis of the corresponding depth. The spectrum of each characteristic defect ultrasonic reflection signal is divided into a number of spectrum points through the depth points;

[0076] Perform normal distribution on each spectrum point at the same depth point, select 30% of the center of the normal distribution relative to the whole as the standard reflection signal spectrum interval at the corresponding depth point, and sequentially splice the standard reflection signal spectrum intervals of each depth point to obtain the characteristic reflection signal spectrum identification intervals at the corresponding position and defect type at different angles;

[0077] The process of identifying the characteristic reflection signal spectrum interval is repeated to obtain the characteristic reflection signal spectrum identification interval at different angles when various defects exist at various positions of the test printed circuit board.

[0078] Further, a plurality of phased array probes are installed on the target printed circuit board at the same positions and in the same number as those on the standard printed circuit board;

[0079] Each phased array probe collects the target ultrasonic reflection signal spectrum at different angles at its location, sets a deviation amplitude threshold and an abnormal node number threshold, and then reduces the target ultrasonic reflection signal spectrum and the corresponding standard ultrasonic reflection signal spectrum in chronological order to obtain the target ultrasonic reflection signal detection spectrum, and determines the size relationship between the amplitude value of the target ultrasonic reflection signal detection spectrum at each time node and the deviation amplitude threshold;

[0080] If there is an amplitude value greater than or equal to the deviation amplitude threshold, it is judged that there is an abnormality at the spectrum point at the corresponding time node, otherwise it is judged that there is no abnormality;

[0081] The total number of spectrum points with abnormalities is counted. If the total number of spectrum points is less than or equal to the threshold of the number of abnormal nodes, it is determined that the spectrum of the target ultrasonic reflection signal at the corresponding shooting angle is normal.

[0082] If the total number of spectrum points is greater than the threshold value of the number of abnormal nodes, the target ultrasonic reflection signal detection spectrum and the target ultrasonic reflection signal detection spectrum of the adjacent collection position are decremented, and then several target characteristic ultrasonic reflection signal spectra are divided from the target ultrasonic reflection signal detection spectrum;

[0083] Then, the spectrum of each target characteristic ultrasonic reflection signal is matched with the characteristic reflection signal spectrum identification interval at the corresponding position and different angles, and the defect type and defect depth of each position of the target printed board are obtained according to the matching results;

[0084] According to the target ultrasonic reflection signal spectrum at different angles of the target printed circuit board, the corresponding target printed circuit board model is established, and according to the defect types and defect depths at various positions of the target printed circuit board, the defect positions are filled and marked at the corresponding positions of the target printed circuit board model.

[0085] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A printed circuit board defect detection system based on computer vision, comprising a cloud computing platform, characterized in that: The cloud computing platform is communicatively connected with a test data acquisition module, a defect feature data analysis module and a defect detection module; The test data acquisition module is used to install several groups of phased array probes on the surface of the test printed circuit board and the standard printed circuit board, and each group of phased array probes transmits multiple groups of ultrasonic beams to the target printed circuit board, thereby collecting several test ultrasonic reflection signal spectra and standard ultrasonic reflection signal spectra at different angles; The acquisition process of the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum includes: A plurality of test printed circuit boards and standard printed circuit boards are provided, a plurality of groups of phased array probes are installed on the front and back sides of the test printed circuit boards and the standard printed circuit boards, and each phased array probe is numbered, wherein the phased array probe is composed of an ultrasonic transmitting device and an ultrasonic receiving device; Setting a number of data collection periods, placing the test printed circuit board and the standard printed circuit board in the same test environment, and allowing multiple ultrasonic signals emitted at different angles to pass through various positions of the printed circuit board within the same data collection period; When all data collection periods are completed, the ultrasonic reflection signal spectra generated by each phased array probe are integrated to generate a corresponding test signal spectrum data set and a standard test signal spectrum data set, and the corresponding phased array probe numbers are marked; The defect feature data analysis module is used to filter the standard ultrasonic reflection signal spectrum and the test ultrasonic reflection signal spectrum, and to separate the defect ultrasonic signal reflection spectrum from the test ultrasonic reflection signal spectrum, and then to establish a test three-dimensional circuit board model according to the test ultrasonic reflection signal spectrum; Then, several defect positions are marked on the three-dimensional circuit board model according to the defect ultrasonic signal reflection spectrum, and the characteristic defect ultrasonic reflection signal spectrum generated at different angles is mapped in the three-dimensional coordinate system to obtain the characteristic defect signal distribution network; The three-dimensional circuit board model to be tested is mapped to the corresponding characteristic defect signal distribution network, thereby obtaining the characteristic reflection signal spectrum identification interval of different defect types at different acquisition angles at each position; The process of generating the characteristic reflection signal spectrum identification interval includes: The three-dimensional circuit board model to be tested is mapped in the characteristic defect signal distribution network, and the characteristic defect ultrasonic reflection signal spectra corresponding to the same defect position but the same acquisition angle are arranged in parallel according to the depth, and a number of depth points are set on the coordinate axis corresponding to the depth, and the spectrum of each characteristic defect ultrasonic reflection signal is divided into a number of spectrum points through the depth points; Perform normal distribution on each spectrum point at the same depth point, obtain the standard reflection signal spectrum interval at each depth point according to the normal distribution result, and sequentially splice the standard reflection signal spectrum intervals at each depth point to obtain the characteristic reflection signal spectrum identification intervals at the corresponding position and defect type at different angles; The defect detection module is used to collect the target ultrasonic reflection signal spectrum at different angles at various positions of the target printed circuit board, compare and match the target ultrasonic reflection signal spectrum with the corresponding standard ultrasonic reflection signal spectrum and the characteristic reflection signal spectrum identification interval in turn, and then determine the defect position and type of the target printed circuit board; The process of determining whether the target printed circuit board has an abnormality includes: The deviation amplitude threshold is set, and the target ultrasonic reflection signal spectrum and the corresponding standard ultrasonic reflection signal spectrum are decremented in chronological order to obtain the target ultrasonic reflection signal detection spectrum, and the magnitude relationship between the amplitude value of the target ultrasonic reflection signal detection spectrum at each time node and the deviation amplitude threshold is determined, and whether the target printed circuit board is abnormal is determined according to the judgment result; The process of determining the defect location and type of the target printed circuit board includes: The target ultrasonic reflection signal detection spectrum is decremented from the target ultrasonic reflection signal detection spectrum at the adjacent acquisition position, and then a plurality of target characteristic ultrasonic reflection signal spectra are divided from the target ultrasonic reflection signal detection spectrum; Match the spectrum of each target characteristic ultrasonic reflection signal with the characteristic reflection signal spectrum identification interval at the corresponding position and different angles, and obtain the defect type and defect depth at each position of the target printed board according to the matching results; According to the target ultrasonic reflection signal spectrum at different angles of the target printed circuit board, a target printed circuit board model is established, and according to the defect types and defect depths at various positions of the target printed circuit board, the defect positions are filled and marked at the corresponding positions of the target printed circuit board model.

2. A printed circuit board defect detection system based on computer vision according to claim 1, characterized in that: The process of filtering the standard ultrasonic reflection signal spectrum and the test ultrasonic reflection signal spectrum includes: A two-dimensional coordinate system is established, and all standard ultrasonic reflection signal spectra are extracted from each standard test signal spectrum data set, and the standard ultrasonic reflection signal spectra generated by the phased array probes corresponding to the front and back sides of the standard printed circuit board are mapped onto the same two-dimensional coordinate system; The standard ultrasonic reflection signal spectrum generated by the phased array probe on the back side is inverted, and the ultrasonic emitting devices corresponding to the front and back sides are set at opposite emitting angles. The standard ultrasonic reflection signal spectrum generated by the front side is subtracted from the inverted standard ultrasonic reflection signal spectrum on the back side to obtain the background signal of each data collection period; The background signal is inverted and the test ultrasonic reflection signal spectrum generated positively is reduced, thereby completing the filtering of the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum in each data collection period.

3. A printed circuit board defect detection system based on computer vision according to claim 2, characterized in that: The generation process of the defect ultrasonic signal reflection spectrum includes: Establishing a local test two-dimensional circuit board image and a local standard two-dimensional circuit board image according to the test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum; A plurality of image regions of the same size are respectively divided in the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, and then the local test two-dimensional circuit board image and the local standard two-dimensional circuit board image corresponding to the same position and the same angle are overlapped and mapped, and the pixel values ​​of each pixel in the overlapped image region are compared to determine whether there is a difference between the two corresponding image regions; According to the judgment result, the corresponding local test two-dimensional circuit board image and the local standard two-dimensional circuit board image, the corresponding test ultrasonic reflection signal spectrum and the standard ultrasonic reflection signal spectrum are decreased in chronological order, and the corresponding defect ultrasonic signal reflection spectrum is obtained according to the decrease result.

4. A printed circuit board defect detection system based on computer vision according to claim 3, characterized in that: The generation process of the test three-dimensional circuit board model includes: The defect ultrasonic signal reflection spectrum of each defect is generated by combining the defect ultrasonic signal reflection spectrum with its adjacent position, and the characteristic defect ultrasonic reflection signal spectrum of different angles at the position of each phased array probe is obtained in descending order of time; A three-dimensional coordinate system is established, and the spectrum of ultrasonic reflection signals of each characteristic defect is mapped onto the three-dimensional coordinate system to obtain a characteristic defect signal distribution network of the corresponding test printed circuit board; Based on the local test two-dimensional circuit board image generated by the same phased array probe, a local test three-dimensional circuit board model of the corresponding test printed plate part is generated, and according to the distribution of each phased array probe, each local test three-dimensional circuit board model is spliced ​​in sequence to obtain a test three-dimensional circuit board model of the corresponding test circuit board, and the defect position is marked on the test three-dimensional circuit board model.

Images