CCD camera imaging center calibration method and system based on direct optical method

Abstract

The present invention discloses a method and system for calibrating the imaging center of a CCD camera based on a direct optical method. The calibration method includes the following steps: S100, obtaining a first interference ring image library and performing preprocessing; S200, performing a convolution operation, learning, training, and correction to obtain a trained calculation model and a third interference ring image library; S300, installing a dedicated calibration device, coarsely adjusting until interference rings appear on the light screen, then fine-tuning, and capturing interference ring images in real time when the interference rings are close to overlapping when observed by the naked eye; S400, inputting the calculation model for calculation and recognition, comparing and determining with the third interference ring image library, obtaining a type result and operation suggestions; S500, performing fine-tuning according to the operation suggestions until all interference rings are determined to overlap; S600, calculating the coordinates of the imaging center point. Compared with the existing technology, the present invention can more accurately determine whether the interference ring centers overlap. The image recognition technology can identify dim light spots that are difficult to detect, thereby improving calibration accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of camera imaging calibration, and in particular to a CCD camera imaging center calibration method and system based on a direct optical method. Background Art

[0002] In recent years, computer vision and visual detection have been widely used in many fields. A three-dimensional computer vision system should be able to calculate the position, shape and other geometric information of three-dimensional environment objects based on the image information obtained by the camera, and thereby identify objects in the environment.

[0003] One of the main research objectives in computer vision is to calculate the geometric information of objects in three-dimensional space from image information captured by a camera. This involves projecting a three-dimensional scene onto a two-dimensional plane, and reconstructing and identifying objects based on this information. The primary challenge in this area is camera calibration. This involves establishing a relationship between image pixel positions and scene point positions. Because each pixel is generated through projection, it corresponds to a ray formed between the optical center and the scene point. Camera calibration involves determining the equation of this ray in the scene's absolute coordinate system.

[0004] Camera calibration plays a very important role in computer vision and is the basis for conducting other research in computer vision. What method can be used to make camera calibration more accurate has become the central issue in current camera calibration research.

[0005] With the improvement of camera imaging resolution, image sampling rate and computer image processing speed, in addition to general monitoring, computer vision and other fields, cameras are increasingly widely used in three-dimensional stereo measurement, visual inspection, motion measurement and other fields, and the measurement accuracy is getting higher and higher. They are playing an increasingly important role in important departments such as aerospace and military.

[0006] The lens optical center (camera imaging center) is the intersection of the optical axis through the lens and the CCD image plane. Typically, the lens optical center serves as the origin for image processing. In high-precision applications, it also serves as the center of the lens radial and tangential distortion calibration model. Methods for calibrating the camera imaging center include the zoom method, radial collimation method, and direct optical method. Direct optical methods offer high accuracy and repeatability.

[0007] In calibration experiments, the camera imaging center coordinates are typically calculated after the human eye determines the overlap of the interference ring centers. However, this is subject to significant subjective influence. When adjusting the camera lens position, it is important to ensure that all interference spots (including some very dim spots that are difficult to detect) are aligned as much as possible. Otherwise, the calibration deviation of the lens optical center and its repeatability will be significantly affected. Introducing these deviations into the calculation will cause errors, thus affecting the calibration accuracy. Summary of the Invention

[0008] The purpose of the present invention is to provide a CCD camera imaging center calibration method and system based on direct optical method, which uses image recognition technology to determine the overlap of interference rings, can identify some very dim light spots that are difficult to detect with the naked eye, reduce calculation errors, and improve calibration accuracy.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] The CCD camera imaging center calibration method based on the direct optical method includes the following steps:

[0011] S100, obtaining a first interference ring image library and preprocessing the first interference ring image library, wherein the first interference ring image library includes images of overlapping interference rings and images with overlapping interference ring centers, each of which is labeled with a certain type. The overlapping interference ring images are labeled as non-overlapping interference rings, and the images with overlapping interference ring centers are labeled as overlapping interference rings.

[0012] S200, establishing a convolutional neural computing model, performing a convolution operation on the preprocessed first interference ring image library, performing learning, training, and correction, and obtaining a trained computing model and a third interference ring image library;

[0013] S300: Install the dedicated calibration equipment for the CCD camera imaging center and perform coarse adjustments until as many interference rings as possible appear on the screen. Then, perform fine adjustments using the five-dimensional adjustment frame. When the interference rings are close to overlapping, add the laser light intensity attenuator and open the CCD camera aperture to capture the interference ring image on the screen in real time.

[0014] S400: Input the real-time acquired image as the image to be recognized into the trained computing model, perform computational recognition on the image to be recognized, and compare and determine it with the third interference ring image library to obtain a type result and corresponding operation suggestions;

[0015] S500, fine-tuning is performed according to the operation suggestions until it is determined that all interference loops coincide;

[0016] S600: Calculate the coordinates of the imaging center point using a centroid positioning method.

[0017] Step S100 further includes:

[0018] S110, image grayscale processing: Gaussian blurring is performed on the image in the first interference ring image library to reduce image noise, and then weighted grayscale processing is performed on the R, G, and B channels of the image;

[0019] S120, image edge analysis: detecting edges in the grayscale image and outputting image information;

[0020] S130, Image Vectorization: Vectorize the image and convert it into a two-dimensional array. The grayscale value of each pixel represents the information of a feature in the vector. The processed data is used as training data and input into the convolutional neural network computing model.

[0021] Step S200 further includes:

[0022] S210, convolution operation and prediction: input the training data into the convolutional neural computing model, which includes two initial convolution layers A and B, and perform convolution operation to obtain prediction results. The prediction results include A type C A , A score S A , B type C B , B score S B , and stipulate that the initial weights of convolutional network A and convolutional network B are ω respectively A and ω B ,ω A =ω B =0.5;

[0023] S220, modify the calculation model according to the initial prediction result: verify and judge the type A and type B in each prediction result respectively with the type marked in the corresponding training data. If both type A and type B are judged to be correct or both are judged to be wrong, then ω A ′=ω A ,ω B ′=ω B ; If type A is correct and type B is wrong, ω A ′=ω A +γ,ω B ′=ω B -γ, when the B type judgment is correct and the A type judgment is wrong, ω A ′=ω A -γ,ω B ′=ω B +γ; where γ is the corrected excitation value, γ=0.01, ω A ′ is the modified weight of the convolutional network A, ω B ′ is the corrected weight of the convolutional network B.

[0024] S230, compare type A and type B, and the corresponding weights, and output the type result R and the score result P: A =C B When R=C A , P = ω A ′*S A +ω B ′*S B When C A ≠C B And ω A ′>ω B 'When R=C A , P=S A When C A ≠C B And ω A ′<ω B ', R=C B , P=S B When C A ≠C B And ω A ′=ω B ′, the corresponding initial interference ring image enters the manual verification and annotation library;

[0025] S240, classify the interference ring image according to the scoring result: set the scoring threshold P m , when P>P m When P≤P m When , the corresponding initial interference ring image enters the manual verification and annotation library;

[0026] S250, manual verification and correction: manually verify and annotate the images in the manual verification and annotation library and then output them to the correction interference ring image library. A ′、ω B 'Make corrections;

[0027] S260, re-learning: Based on the manually annotated corrected interference ring image library, the convolutional neural network calculation model is input again for prediction calculation, and the convolutional neural network calculation model is repeatedly trained and corrected to obtain a trained calculation model and a third interference ring image library.

[0028] The calculation and determination process of step S400 is:

[0029] When C A '=C B ', R'=C A ', when C A '≠C B 'And ω A ′>ωB ', R'=C A ', when C A '≠C B 'And ω A ′<ω B ', R'=C B ', where C A ' and C B ' is the classification result calculated in real time;

[0030] When R' is an interference ring that does not overlap, the operator recommends that the experimenter continue fine-tuning; when R' is an interference ring that overlaps, the operator recommends that the experimenter stop fine-tuning and output the currently acquired corresponding light screen image.

[0031] The present invention also provides a CCD camera imaging center calibration system based on a direct optical method, which is used to implement the above-mentioned method, including a data input module, a preprocessing module, a calculation module and an output display module;

[0032] The data input module is used to input the first interference ring image library and the interference ring image collected in real time;

[0033] The preprocessing module is connected to the data input module and is used to preprocess the first interference ring image library and the interference ring images collected in real time;

[0034] The calculation module is connected to the pre-processing module and is used to perform calculation, identification, analysis and judgment on the pre-processed data;

[0035] The output display module is connected to the calculation module and is used to display the calculation results and corresponding operation suggestions.

[0036] Compared with the prior art, the beneficial technical effects of the present invention are:

[0037] 1. Using image recognition technology to determine whether the interference ring centers coincide is more accurate than using the human eye. Image recognition technology can identify very dim light spots that are difficult to detect with the naked eye, thereby reducing calculation errors and improving the calibration accuracy of the lens optical center.

[0038] 2. Using the convolution results as training data for the computational model can effectively improve the image annotation accuracy. The computational model can be verified and corrected by reserving training data and manually correcting annotations, thereby reducing the impact of factors such as labeling errors or input errors on the computational accuracy, and ultimately achieving the expected accuracy of the computational model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a flow chart of the CCD camera imaging center calibration method based on the direct optical method in Example 1;

[0041] Figure 2 Schematic diagram of the overall structure of the dedicated calibration equipment for the CCD camera imaging center in Example 1;

[0042] Figure 3 This is an example image of the overlapping interference rings in Example 1;

[0043] Figure 4 This is an example image of the overlapped interference ring centers in Example 1;

[0044] Figure 5 This is a structural block diagram of the CCD camera imaging center calibration system based on the direct optical method in Example 2;

[0045] Figure 6 This is a structural block diagram of the computer device in Example 3.

[0046] Reference numerals:

[0047] 1. Processor; 2. Memory; 3. Communication interface; 4. Network adapter; 5. External device;

[0048] 11. Data input module; 12. Preprocessing module; 13. Calculation module; 14. Output display module;

[0049] 21. Helium-neon laser; 22. Laser bracket; 23. Light screen; 24. Optical platform; 25. CCD camera; 26. Five-dimensional adjustment frame. DETAILED DESCRIPTION

[0050] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0051] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0052] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0053] Example 1

[0054] Reference Figure 1 , is a CCD camera imaging center calibration method based on a direct optical method disclosed in the present invention, comprising the following steps:

[0055] S100: Obtain a first interference ring image library and preprocess the first interference ring image library, wherein the first interference ring image library includes images of overlapping interference rings of a certain type and images of interference rings with coincident centers. The overlapping interference ring images are labeled as non-overlapping interference rings, and the images of interference rings with coincident centers are labeled as coincident interference rings.

[0056] Image preprocessing is the process of separating each image for analysis and recognition. In image analysis, it is the processing performed on the input image before feature extraction, segmentation, and matching. Its main purpose is to eliminate irrelevant information in the image, restore useful real information, enhance the detectability of relevant information, and maximize data simplification, thereby improving the reliability of feature extraction, matching, and recognition.

[0057] The image acquisition in the first interference ring image library requires the use of a dedicated calibration device for the CCD camera imaging center. Figure 2The dedicated calibration equipment for the CCD camera imaging center includes a helium-neon laser 21 and a laser bracket 22, a light screen 23 with small holes, an optical platform 24, a CCD camera 25 and a five-dimensional adjustment frame 26. First, install the helium-neon laser 21 on the laser bracket 22, and install the CCD camera 25 on the five-dimensional adjustment frame 26. The laser bracket 22, the five-dimensional adjustment frame 26 and the light screen 23 are all placed on the optical platform 24, so that the light source of the helium-neon laser 21 and the lens center of the CCD camera 25 are basically on the same horizontal line. The dotted line in the figure indicates the light source direction of the helium-neon laser. Before turning on the power of the helium-neon laser 21, close the camera aperture and roughly adjust the position of the five-dimensional adjustment frame 26 so that the interference rings formed by the reflected light from each surface of the camera lens appear on the light screen 23, as shown in FIG. Figure 3 As shown, the image of the overlapping interference rings is recorded. Then, the direction of the camera is fine-tuned by the knob of the five-dimensional adjustment frame 26 so that the centers of all interference spots coincide with the center of the exit aperture. The images of the overlapping interference rings during the adjustment process and the final image of the overlapping interference ring centers are recorded. The final image of the overlapping interference ring centers is shown as Figure 4 As shown. At this point, the incident laser beam and the reflected light coincide with each other and are coaxial with the lens optical axis. The image point of the laser beam on the CCD image plane is the optical imaging center of the lens. The above method can be used to obtain images of overlapping interference rings and images of coincident interference ring centers.

[0058] It is worth mentioning that the technical features of the helium-neon laser 21, laser bracket 22, optical platform 23, five-dimensional adjustment frame 26, etc. involved in this patent application should be regarded as prior art. The specific structure, working principle and possible control method and spatial arrangement method of these technical features can adopt the conventional selection in the field and should not be regarded as the invention point of this patent. This patent will not be further elaborated.

[0059] Step S100 further includes:

[0060] S110, image grayscale processing: Gaussian blurring is performed on the image in the first interference ring image library to reduce image noise, and then weighted grayscale processing is performed on the R, G, and B channels of the image.

[0061] It should be noted that in the color space RGB model, if the red, green and blue components are equal, the color represents a grayscale color, and then this same value is the grayscale value. Therefore, in the computer, each pixel of the grayscale image can be stored in one byte, and its range is 0-255.

[0062] S120, image edge analysis: detect edges in the grayscale image and output image information.

[0063] S130, image vectorization: vectorizing the image and converting it into a two-dimensional array. The grayscale value of each pixel represents information about a feature in the vector. The processed data is input into the convolutional neural network computing model as training data.

[0064] S200, using a convolutional neural computing model to perform convolution operations on the training data, perform learning, training and correction, and obtain a trained computing model and a third interference ring image library.

[0065] Step S200 further includes:

[0066] S210, initial calculation prediction: input the training data into the convolutional neural computing model, which includes two initial convolutional layers A and B. The initial convolutional layers A and B are followed by convolutional layers + pooling layers respectively. The combination of convolutional layers + pooling layers can appear multiple times, and the number of occurrences can be set according to the needs of the model. The purpose of the convolution operation is to extract different features of the input. The first convolutional layer may only be able to extract some low-level features such as edges, lines, and corners. Model structures with more layers can iteratively extract more complex features from low-level features to meet actual computing needs. The initial convolutional layer A is a standardized vector convolutional network with Gaussian blur and grayscale processing, and the convolutional network B is a standardized vector convolutional network with Gaussian blur and RGB full channels. Then the convolution operation is performed to obtain the prediction results. The prediction results include type A, type C, and type C. A , A score S A , B type C B , B score S B , where C A and C B The predicted values ​​include two types: overlapped and non-overlapped, and the initial weights of convolutional network A and convolutional network B are specified as ω respectively. A and ω B ,ω A =ω B =0.5.

[0067] S220: Since the initial calculation accuracy of the calculation model is not ideal, it is necessary to modify the calculation model according to the initial prediction results: the type A and type B in each prediction result are respectively verified and judged with the types (overlapping and non-overlapping) marked in the corresponding training data. If type A and type B are both judged to be correct or both judged to be wrong, the weight remains unchanged, that is, ω A ′=ω A ,ω B ′=ω B ; If only one of type A and type B is judged correctly, the correct one is given positive incentives, and the wrong one is given negative incentives. That is, when type A is judged correctly and type B is judged incorrectly, ω A ′=ω A+γ,ω B ′=ω B -γ, when the B type judgment is correct and the A type judgment is wrong, ω A ′=ω A -γ,ω B ′=ω B +γ; where γ is the corrected excitation value, ω A ′ is the modified weight of the convolutional network A, ω B ' is the modified weight of the convolutional network B. For example, γ = 0.001.

[0068] Since the initial entry into the first interference ring image library may be affected by factors such as labeling errors or entry errors, manual verification and correction are required.

[0069] S230, compare type A and type B, and the corresponding weights, and output the type result R and the score result P: A =C B When R=C A , P = ω A ′*S A +ω B ′*S B When C A ≠C B And ω A ′>ω B 'When R=C A , P=S A When C A ≠C B And ω A ′<ω B ', R=C B , P=S B When C A ≠C B And ω A ′=ω B ′, the corresponding initial interference ring image enters the manual verification and annotation library.

[0070] S240, classify the interference ring image according to the scoring result: set the scoring threshold P m , when P>P m When P≤P m When , the corresponding initial interference ring image enters the manual verification and annotation library.

[0071] S250, manual verification and correction: manually verify and annotate the images in the manual verification and annotation library and then output them to the correction interference ring image library. A ′、ωB ’ to make corrections.

[0072] S260: Relearning: Based on the manually annotated corrected interference ring image library, the convolutional neural network computing model is again input for prediction calculation. The convolutional neural network computing model is repeatedly trained and corrected to obtain a trained computing model and a third interference ring image library. The images in the third interference ring image library are stored in the form of an annotation type and a two-dimensional vector group.

[0073] Using the convolution results as training data for the computational model can effectively improve the accuracy of image annotation. The computational model can be verified and corrected by reserving training data and manually correcting annotations, thereby reducing the impact of factors such as labeling errors or input errors on the computational accuracy, and ultimately ensuring that the computational accuracy of the computational model reaches the expected results.

[0074] S300, install the dedicated calibration equipment for the CCD camera imaging center, perform coarse adjustment until as many interference rings as possible appear on the light screen, then perform fine adjustment using the five-dimensional adjustment frame. When the interference rings are close to overlapping when observed with the naked eye, add a laser light intensity attenuator, open the CCD camera aperture, and capture the interference ring image on the light screen in real time.

[0075] S400: Input the real-time captured image as the image to be identified into the trained computing model, perform computational identification on the image to be identified, and compare and determine it with the third interference ring image library, thereby obtaining a type result R', and outputting the type result R' and corresponding operation suggestions.

[0076] The specific process is: when C A '=C B ', R'=C A ';When C A '≠C B 'And ω A ′>ω B ', R'=C A ';When C A '≠C B 'And ω A ′<ω B ', R'=C B ', where C A ' and C B ' is the classification result calculated in real time.

[0077] When the type result R' is a non-overlapping interference ring, the operator recommends that the experimenter continue fine-tuning; when the type result R' is an overlapping interference ring, the operator recommends that the experimenter stop fine-tuning and output the currently corresponding collected and pre-processed light screen image at the same time.

[0078] S500. The experimenter makes fine adjustments according to the operation suggestions until it is determined that all interference loops coincide.

[0079] S600: Calculate the coordinates of the imaging center point using a centroid positioning method.

[0080] After obtaining the grayscale image of the imaging center point (i.e., the screen image output after fine-tuning), its coordinates need to be determined using the centroid positioning method. Centroid positioning is a method that uses the target's regional moment to find the location of the regional centroid.

[0081] The target region moment is a description of the region shape. The region moment of the target image f(x,y) is expressed as:

[0082]

[0083] Where p,q are the orders, and x,y are contained in the target region s.

[0084] The centroid of the target region is located by using the zero-order moment and the first-order moment of the region. When locating a region, the region is usually treated as a binary image, that is, f(x,y) = 1. In this way, the zero-order moment m of the target region is 00 For:

[0085] m 00 =∑∑1

[0086] The first moment is:

[0087] m 10 =∑∑x

[0088] m 01 =∑∑y

[0089] Then the center of mass of the region (x c ,y c ) is expressed as:

[0090] x c =m 10 / m 00

[0091] y c =m 01 / m 00 (Formula 2)

[0092] Calculate the number of pixels m in the centroid area 00 , and then calculate according to formula 2 to get the coordinate value of the CCD camera imaging center.

[0093] The beneficial effects of this method are:

[0094] The use of image recognition methods to determine whether the interference ring centers coincide can be more accurately determined than by the human eye. Image recognition technology can identify very dim light spots that are difficult to detect with the naked eye, thereby reducing calculation errors and improving the calibration accuracy of the lens optical center.

[0095] Example 2

[0096] Reference Figure 5 , which is a CCD camera imaging center calibration system based on the direct optical method disclosed in the present invention, includes a data input module 11, a preprocessing module 12, a calculation module 13 and an output display module 14, wherein the data input module 11 is used to input the first interference ring image library and the interference ring image collected in real time.

[0097] The preprocessing module 12 is connected to the data input module 11 and is used to preprocess the first interference ring image library and the interference ring images collected in real time.

[0098] The calculation module 13 is connected to the pre-processing module 12 and is used to perform calculation, identification, analysis and judgment on the pre-processed data.

[0099] The output display module 14 is connected to the calculation module 13 and is used to display the calculation results and corresponding operation suggestions.

[0100] This system can implement the corresponding methods in the aforementioned method embodiments. The specific implementation process can be found in the aforementioned method embodiments and will not be repeated here.

[0101] For example, the following are experimental results obtained by using the above system to perform a calibration experiment:

[0102] To ensure accuracy, this measurement was adjusted 10 times, and an image was recorded each time the interference rings overlapped. The image is a 756*574 BMP image. The center results obtained using the centroid positioning method are shown in the table below:

[0103] 1st time 2nd time 3rd time 4th time 5th 6th X direction 367.2 366.4 366.8 366.4 366.8 367.2 Y direction 288.0 287.8 287.6 288.2 287.8 287.7 7th 8th 9th 10th average value Standard deviation S X direction 367.2 367.0 367.2 367.0 366.92 0.0896 Y direction 287.7 288.4 287.8 287.4 287.84 0.0644

[0104] The experimental results show that the coordinates of the CCD camera imaging center in the image are (366.92, 287.84), not the center of the captured 756*574 image (378, 287). This imaging center should be used as the calibration center and the origin of image processing distortion correction.

[0105] Example 3

[0106] Reference Figure 6, is an embodiment of a computer device provided by the present invention, the computer device includes: one or more processors 1, and a memory 2 for storing executable instructions of the processor, and the memory 2 and the processor 1 are electrically connected.

[0107] Among them, the processor 1 is configured to execute the method corresponding to the aforementioned method embodiment by executing the executable instructions. The specific implementation process can be found in the aforementioned method embodiment and will not be repeated here.

[0108] Optionally, the computer device may further include: a communication interface 3, through which the computer device can be communicatively connected to one or more external devices 4 (such as a keyboard, a pointing device, a display, etc.).

[0109] Optionally, the computer device may further include: a network adapter 5, through which the computer device can communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN) and / or a public network, such as the Internet).

[0110] The present invention also provides a computer-readable storage medium on which a computer program (or computer-executable instructions) is stored. When the computer program is executed, the corresponding method in the aforementioned method embodiment can be implemented. The specific implementation process can be referred to the aforementioned method embodiment and will not be repeated here.

[0111] The computer-readable storage medium is a disk, an optical disk, a read-only memory, a random access memory, or the like.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A CCD camera imaging center calibration method based on direct optical method, characterized in that: The steps include: S100, obtaining a first interference ring image library and preprocessing the first interference ring image library, wherein the first interference ring image library includes images of overlapping interference rings and images with overlapping interference ring centers, each of which is labeled with a certain type. The overlapping interference ring images are labeled as non-overlapping interference rings, and the images with overlapping interference ring centers are labeled as overlapping interference rings. S200, establishing a convolutional neural computing model, performing a convolution operation on the preprocessed first interference ring image library, performing learning, training, and correction, and obtaining a trained computing model and a third interference ring image library; S300: Install the dedicated calibration equipment for the CCD camera imaging center and perform coarse adjustments until as many interference rings as possible appear on the screen. Then, perform fine adjustments using the five-dimensional adjustment frame. When the interference rings are close to overlapping, add the laser light intensity attenuator and open the CCD camera aperture to capture the interference ring image on the screen in real time. S400: Input the real-time acquired image as the image to be recognized into the trained computing model, perform computational recognition on the image to be recognized, and compare and determine it with the third interference ring image library to obtain a type result and corresponding operation suggestions; S500, fine-tuning is performed according to the operation suggestions until it is determined that all interference loops coincide; S600: Calculate the coordinates of the imaging center point using a centroid positioning method.

2. The CCD camera imaging center calibration method based on direct optical method according to claim 1, characterized in that: Step S100 further includes: S110, image grayscale processing: Gaussian blurring is performed on the image in the first interference ring image library to reduce image noise, and then weighted grayscale processing is performed on the R, G, and B channels of the image; S120, image edge analysis: detecting edges in the grayscale image and outputting image information; S130, image vectorization: vectorize the image and convert it into a two-dimensional array. The grayscale value information of each pixel represents the information of a feature in the vector. The data after the above processing is input into the convolutional neural computing model as training data.

3. The CCD camera imaging center calibration method based on direct optical method according to claim 2, characterized in that: Step S200 further includes: S210, convolution operation and prediction: input the training data into the convolutional neural computing model, which includes two initial convolution layers A and B, and perform convolution operation to obtain prediction results. The prediction results include A type C A , A score S A , B type C B , B score S B , and stipulate that the initial weights of convolutional network A and convolutional network B are ω respectively A and ω B ,ω A =ω B =0.5; S220, modify the calculation model according to the initial prediction result: verify and judge the type A and type B in each prediction result respectively with the type marked in the corresponding training data. If both type A and type B are judged to be correct or both are judged to be wrong, then ω A ′=ω A ,ω B ′=ω B ; If type A is correct and type B is wrong, ω A ′=ω A +γ,ω B ′=ω B -γ, when the B type judgment is correct and the A type judgment is wrong, ω A ′=ω A -γ,ω B ′=ω B +γ; where γ is the corrected excitation value, γ=0.01, ω A ′ is the modified weight of the convolutional network A, ω B ′ is the corrected weight of the convolutional network B.

4. The CCD camera imaging center calibration method based on direct optical method according to claim 3, characterized in that: After step S220, the method further includes: S230, compare type A and type B, and the corresponding weights, and output the type result R and the score result P: A =C B When R=C A , P = ω A ′*S A +ω B ′*S B When C A ≠C B And ω A ′>ω B 'When R=C A , P=S A When C A ≠C B And ω A ′<ω B ', R=C B , P=S B When C A ≠C B And ω A ′=ω B ′, the corresponding initial interference ring image enters the manual verification and annotation library; S240, classify the interference ring image according to the scoring result: set the scoring threshold P m , when P>P m When P≤P m When , the corresponding initial interference ring image enters the manual verification and annotation library; S250, manual verification and correction: manually verify and annotate the images in the manual verification and annotation library and then output them to the correction interference ring image library. A ′、ω B 'Make corrections; S260, re-learning: Based on the manually annotated corrected interference ring image library, the convolutional neural network calculation model is input again for prediction calculation, and the convolutional neural network calculation model is repeatedly trained and corrected to obtain a trained calculation model and a third interference ring image library.

5. The CCD camera imaging center calibration method based on direct optical method according to claim 1, characterized in that: The calculation and determination process of step S400 is: When C A '=C B ', R'=C A ', when C A '≠C B 'And ω A ′>ω B ', R'=C A ', when C A '≠C B 'And ω A ′<ω B ', R'=C B ', where C A ' and C B ' is the classification result calculated in real time; When R' is an interference ring that does not overlap, the operator recommends that the experimenter continue fine-tuning; when R' is an interference ring that overlaps, the operator recommends that the experimenter stop fine-tuning and output the currently acquired corresponding light screen image at the same time.

6. The CCD camera imaging center calibration method based on direct optical method according to claim 1, characterized in that: The calculation process of the imaging center point coordinates is as follows: The regional moment of the target image f(x, y) is expressed as: Where p,q are orders, x,y are contained in the target area s, The centroid of the target area is located by using the zero-order moment and the first-order moment of the area. When locating the area, the area is usually treated as a binary image, that is, f(x,y)=1. In this way, the zero-order moment m of the target area is 00 For: m 00 ＝∑∑1 The first moment is: m 10 =∑∑x m 01 =∑∑y Then the center of mass of the region (x c ,y c ) is expressed as: x c =m 10 / m 00 y c =m 01 / m 00 (Formula 2) Calculate the number of pixels m in the centroid area 00 , and then calculate according to formula 2 to get the coordinate value of the CCD camera imaging center.

7. The method for calibrating the imaging center of a CCD camera based on direct optical method according to claim 3, characterized in that: The initial convolution layer A is a standardized vector convolution network with Gaussian blur and grayscale processing, and the convolution network B is a standardized vector convolution network with Gaussian blur and RGB full channels.

8. A CCD camera imaging center calibration system based on direct optical method, used to implement the method according to any one of claims 1 to 7, characterized in that: It includes a data input module (11), a pre-processing module (12), a calculation module (13) and an output display module (14); The data input module (11) is used to input the first interference ring image library and the interference ring image collected in real time; The preprocessing module (12) is connected to the data input module (11) and is used to preprocess the first interference ring image library and the interference ring image collected in real time; The calculation module (13) is connected to the pre-processing module (12) and is used to perform calculation, identification, analysis and judgment on the pre-processed data; The output display module (14) is connected to the calculation module (13) and is used to display the calculation results and corresponding operation suggestions.

9. A computer device, characterized in that: The invention comprises a processor (1) and a memory (2), wherein the memory (2) is electrically connected to the processor (1), and the memory (2) is used to store executable instructions of the processor (1); wherein the processor (1) is configured to execute the method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, it is used to implement the method according to any one of claims 1 to 7.

Images