Neural network and optical interference centering calibration method and system, and storage medium

Abstract

The present invention discloses a neural network and optical interference centering calibration method and system, and a storage medium. The method comprises: S1, measuring the optical distance between a camera and a target plate, and adjusting the initial focal length or position of the camera according to the measurement result; S2, identifying and determining the target plate and a first center coordinate in an image, and determining a first offset between the first center coordinate and an ideal center position; S3, obtaining a second center coordinate and a second offset, and determining and executing an adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset; S4, measuring and obtaining the distance and angle data between the camera and the target plate, and fine-tuning the focal length or position of the camera; S5, fine-tuning the optical path of the camera according to the first offset, the second offset, the distance and the angle data between the camera and the target plate. The present invention ensures a high degree of centering accuracy, has strong automatic adaptability, makes the entire centering calibration process rapid and stable, and has greater flexibility.

Description

Technical Field

[0001] The present invention relates to the field of centring technology, and in particular to a neural network and optical interference centring calibration method and system, and a storage medium. Background Art

[0002] Centring technology is widely used in various optical, imaging and measurement systems to ensure accurate alignment between the camera or optical sensor and the target object. Traditionally, such calibration mainly relies on mechanical fine-tuning or uses basic machine vision technology. However, with the rapid development of technology, especially in high-precision and complex environments, traditional methods face some difficulties in meeting the requirements of accuracy, speed and stability.

[0003] First of all, relying solely on mechanical fine-tuning or simple machine vision technology cannot meet the requirements of modern high-precision measurement equipment and is easily affected by environmental factors such as temperature and humidity changes, resulting in unstable measurement results.

[0004] Secondly, many complex application environments require the system to have higher adaptability and be able to respond to environmental changes and make adjustments in real time. Traditional technologies often lack this adaptability and are difficult to meet the needs of practical applications.

[0005] In addition, with the rise of deep learning and neural network technology, although great progress has been made in image recognition and processing, how to combine these advanced technologies with traditional optical measurement technology to achieve higher accuracy and faster centering calibration remains a challenge. Summary of the invention

[0006] In view of the many problems existing in the above-mentioned traditional mechanical fine adjustment when facing centering, the present invention provides a neural network and optical interference centering calibration method and system, and storage medium. The present invention uses a centering calibration method using machine vision, neural network and optical measurement technology to overcome the above-mentioned difficulties and provide more accurate and stable centering results.

[0007] A neural network and optical interference centering calibration method, comprising the following steps:

[0008] S1, measuring the optical distance between the camera and the target plate, and adjusting the initial focal length or position of the camera according to the measurement result;

[0009] S2, after the camera captures the image of the target plate, the target plate and the first center coordinates in the image are determined through machine vision and pattern recognition, and a first offset between the first center coordinates and the ideal center position is determined;

[0010] S3, inputting the image into a pre-trained neural network model for coordinate recognition, obtaining a second center coordinate and a second offset, and determining and executing an adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset;

[0011] S4. Based on the optical interferometry technology, the distance and angle data between the camera and the target plate are measured, and the focal length or position of the camera is fine-tuned;

[0012] S5. Based on adaptive optical technology, the optical path of the camera is fine-tuned according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

[0013] Preferably, in step S1, the measurement result is compared with the ideal distance to obtain an initial deviation value, and the initial focal length or position of the camera is adjusted so that the measurement result is close to a predetermined ideal distance range.

[0014] Preferably, step S1 further includes, when the measurement result reaches the ideal distance range, executing step S2.

[0015] Preferably, in step S2, after the camera captures the image of the target plate, the image is preprocessed, and the first center coordinate of the target plate in the image is identified by machine vision and pattern recognition, and the first center coordinate is compared with the coordinate of the ideal center position to obtain a first offset between the first center coordinate and the ideal center position;

[0016] Wherein, the preprocessing includes: denoising and enhancement.

[0017] Preferably, in step S3, the neural network model is pre-trained through positive samples, negative samples and perturbation samples, and the neural network model is used to identify features related to the target plate or the center position from the image, analyze the extracted features, locate the center of the target plate, obtain the second center coordinates, and compare them with the preset ideal center position to calculate the second offset;

[0018] The positive samples include: target images taken at different distances, angles, and lighting conditions;

[0019] The negative samples include: images of other objects with similar shapes or features to the target;

[0020] The perturbed sample comprises: transforming the original target plate image.

[0021] Preferably, in step S3, the adjustment mechanism includes:

[0022] Comparing the difference between the first center coordinate and the second center coordinate to obtain a first comparison result;

[0023] When the first comparison result is greater than the first threshold, adjusting the initial focal length or position of the camera according to the difference between the two;

[0024] Comparing the first offset and the second offset with a second threshold value respectively to obtain a second comparison result;

[0025] When the second comparison result is that the first offset is greater than the second threshold, adjusting the machine vision and pattern recognition;

[0026] When the second comparison result is that the first offset is greater than the second threshold, the training samples and parameters of the neural network model are adjusted;

[0027] When the second comparison result is that both the first offset and the second offset are greater than the second threshold, the initial position or focal length of the camera is adjusted;

[0028] Adjustments are made according to the first comparison result and the second comparison result, and steps S1-S3 are re-executed until the first comparison result is less than the first threshold and the second comparison result is that both the first offset and the second offset are less than the second threshold, and then step S4 is executed.

[0029] Preferably, in step S4, based on the optical interferometry technology, the distance and angle between the camera and the target plate are measured through the second center coordinate, the first center coordinate, the second center coordinate, the distance and angle between the camera and the target plate are analyzed, and the focal length or position of the camera is fine-tuned based on the analysis result;

[0030] The analysis results include: coordinate comparison, distance and angle verification, and technical accuracy assessment;

[0031] Among them, the fine-tuning corresponding to the analysis result includes: coordinate adjustment, distance and angle adjustment, and adjustment strategy suggestion.

[0032] Preferably, step S5 comprises:

[0033] S501, evaluating optical parameters that need to be adjusted based on the first offset, the second offset, and the distance and angle between the camera and the target plate;

[0034] S502, using adaptive optics technology to fine-tune the optical path of the camera according to the optical parameters;

[0035] S503, collect the image of the target plate, and determine its center coordinates through machine vision and pattern recognition to determine whether it is close to the expected ideal center position. If so, complete the fine-tuning; if not, loop through steps S501-S503.

[0036] A neural network and optical interference centering calibration system, comprising:

[0037] The hardware module is an integration of a camera and an optical interferometer. The camera is located at the front end of the hardware module and is responsible for capturing images in the scene, including the image of the target plate. The optical interferometer is closely connected to the camera and is located at the rear or side of the camera without blocking the camera's field of view. The interference beam emitted by the optical interferometer is consistent with the shooting direction of the camera. The camera maintains a certain vertical distance from the target plate and faces the center of the target plate.

[0038] A preparation module, the preparation module is used to measure the optical distance between the camera and the target plate, and adjust the initial focal length or position of the camera according to the measurement result;

[0039] A machine recognition module, wherein after the camera captures the image of the target plate, the machine recognition module is used to determine the target plate and the first center coordinates in the image through machine vision and pattern recognition, and determine a first offset between the first center coordinates and the ideal center position;

[0040] A neural network recognition module, the neural network recognition module is used to input the image into a pre-trained neural network model for coordinate recognition, obtain the second center coordinate and the second offset, and determine and execute the adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset;

[0041] A first fine-tuning module, which measures the distance and angle data between the camera and the target plate based on the optical interferometry technology, and fine-tunes the focal length or position of the camera;

[0042] The second fine-tuning module is based on adaptive optical technology and fine-tunes the optical path of the camera according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

[0043] A storage medium stores a computer program, which, when executed by a processor, implements the steps of the neural network and optical interference centering calibration method.

[0044] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0045] (1) The present invention combines machine vision, neural network technology and optical interferometry to achieve multiple calibrations and verifications, thereby ensuring a high degree of accuracy in centering;

[0046] (2) The present invention uses adaptive optics technology to fine-tune the optical path, so that the system can automatically adapt to various environments and conditions and provide the best imaging results;

[0047] (3) The continuous monitoring and feedback links in the system of the present invention allow real-time adjustment and optimization, making the entire centering calibration process faster and more stable;

[0048] (4) By setting thresholds and using positive samples, negative samples and perturbation samples for training, the neural network model can be fine-tuned and applied in a wide range of application scenarios, providing greater flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the process of the present invention;

[0050] Figure 2 Schematic diagram of recognition of the winning bid board image of the present invention;

[0051] Figure 3 It is a schematic diagram of the process of fine-tuning based on the optical interferometry technology in the present invention;

[0052] Figure 4 It is a schematic diagram of the process of fine-tuning based on adaptive optics technology in the present invention;

[0053] Figure 5 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0054] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0055] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0056] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0057] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0058] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some boxes or combinations thereof in the block diagrams and / or flow charts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions can create a device for implementing the functions / operations described in these block diagrams and / or flow charts when executed by the processor. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by an instruction execution system or used in conjunction with an instruction execution system.

[0059] like Figure 1 As shown, a neural network and optical interference centering calibration method includes the following steps:

[0060] S1, measuring the optical distance between the camera and the target plate, and adjusting the initial focal length or position of the camera according to the measurement result;

[0061] Preferably, in step S1, the measurement result is compared with the ideal distance to obtain an initial deviation value, and the initial focal length or position of the camera is adjusted so that the measurement result is close to a predetermined ideal distance range.

[0062] This S1 step relies on optical measurement technology, which uses a laser or other light source to send light to the target and receive the reflected light. By measuring the propagation time or other optical properties of the light, the distance between the camera and the target can be determined.

[0063] The basic principle of optical distance measurement is to use the constant speed of light (in a specific medium) to calculate the distance by measuring the time it takes for light to travel from one point to another. Of course, other factors will be considered in practical applications, such as refraction and reflection of light.

[0064] When the actual distance between the camera and the target is obtained, an initial deviation value can be obtained by comparing it with the preset ideal distance. This deviation value can be used to adjust the initial focal length or position of the camera to make it closer to the ideal working state.

[0065] This step is the initial setup phase of the entire calibration process. By accurately measuring and making preliminary adjustments, you can lay the foundation for more precise adjustments later. The focal length and position of the camera are two of the key parameters for the final image quality. Adjusting these parameters can ensure that the image provided by the camera is closer to the actual object, thereby improving the overall system accuracy.

[0066] In one embodiment, suppose there is an automatic welding robot in a manufacturing workshop, and its job is to dock various parts. In order to accurately position the parts, it is equipped with a camera. Before the robot starts working, engineers need to ensure that the position and focal length of the camera match its working distance. Therefore, they use a standard plate to measure the actual distance between the camera and the plate through optical measurement technology. Then, compare it with the ideal working distance. If there is a deviation, adjust the position or focal length of the camera until the actual distance is close to the ideal distance.

[0067] Preferably, step S1 further includes, when the measurement result reaches the ideal distance range, executing step S2.

[0068] Setting of judgment logic: The core principle of this step is to determine the subsequent operation process based on a judgment logic or threshold judgment. In technical operations, it is often necessary to ensure that certain conditions are met before proceeding to the next step, so as to ensure the accuracy and stability of the process. Here, the "ideal distance range" is this judgment logic or threshold. It can be a fixed value or a range value, such as within ±1mm.

[0069] Comparison and calibration: After optical distance measurement between the camera and the target, the actual distance is compared with the pre-set "ideal distance". The "ideal distance" here represents the distance that should be maintained between the camera and the target under optimal conditions to ensure image clarity, accuracy and other relevant parameters.

[0070] Decision making: If the difference between the actual measured distance and the ideal distance is within an acceptable range (i.e., it reaches the "ideal distance range"), the system or operator will determine that the task of this step has been completed and decide to proceed to the next step. This mechanism of making decisions based on the difference between the actual result and the expected result is a common practice in many automated or semi-automated processes.

[0071] This result-based automatic decision-making method can avoid unnecessary repeated operations and human intervention, thereby improving the overall process efficiency; only when the measurement results meet specific conditions or standards will it enter the next step, which can ensure that each step is based on accurate and reliable data, thereby improving the accuracy and reliability of the overall operation; this decision-making method also provides a certain degree of flexibility for subsequent processes. For example, if the "ideal distance range" needs to be modified in the future, only this parameter needs to be adjusted without changing the entire process.

[0072] S2, after the camera captures the image of the target plate, the target plate and the first center coordinates in the image are determined through machine vision and pattern recognition, and a first offset between the first center coordinates and the ideal center position is determined;

[0073] Preferably, Figure 2 As shown, in step S2, after the camera captures the image of the target plate, the image is preprocessed, and the first center coordinate of the target plate in the image is identified by machine vision and pattern recognition, and the first center coordinate is compared with the coordinate of the ideal center position to obtain a first offset between the first center coordinate and the ideal center position;

[0074] Machine vision is the use of computers to replace human eyes to perform machine vision tasks such as target recognition, tracking and measurement, while pattern recognition is the technology of classifying or labeling input data. In this scenario, machine vision technology first identifies the target plate from the image captured by the camera, and then pattern recognition technology further identifies the first center coordinates of the target plate.

[0075] The images captured by the camera will be affected by many factors, such as light, noise, blur, etc. These factors will interfere with subsequent recognition. Therefore, it is very necessary to preprocess the image, which can improve the quality of the image and provide more accurate data for subsequent recognition.

[0076] Once the first center coordinate is identified, the difference from the ideal center position is the offset, which is the key data for determining how to make adjustments in the subsequent steps.

[0077] Wherein, the preprocessing includes: denoising and enhancement.

[0078] Denoising refers to removing random noise in an image, while enhancement makes the target features in the image more obvious by adjusting parameters such as contrast and brightness of the image. Both operations are aimed at improving the recognition accuracy of the image.

[0079] Through preprocessing, noise and interference in the image are removed, so that machine vision and pattern recognition technology can more accurately identify the target plate and the first center coordinate on it.

[0080] In one embodiment, assuming a laboratory environment, the target plate image captured by the camera has noise points due to the interference of surrounding light. Without preprocessing, the pattern recognition technology will misjudge the noise points as feature points on the target plate, causing the first center coordinates of the recognition to deviate from the true position. However, after denoising and enhancement preprocessing, the noise points are effectively eliminated, the recognition results are greatly improved, and the first offset obtained is more accurate.

[0081] S3, inputting the image into a pre-trained neural network model for coordinate recognition, obtaining a second center coordinate and a second offset, and determining and executing an adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset;

[0082] Preferably, in step S3, the neural network model is pre-trained through positive samples, negative samples and perturbation samples, and the neural network model is used to identify features related to the target plate or the center position from the image, analyze the extracted features, locate the center of the target plate, obtain the second center coordinates, and compare them with the preset ideal center position to calculate the second offset;

[0083] A neural network is a computational model that mimics the working principle of the human brain. It consists of multiple layers, each containing multiple neurons. Through the connection weights between neurons, the neural network can transform and extract features from input data. Under the framework of deep learning, multi-layer neural networks (i.e., deep neural networks) can capture complex features and structures in data.

[0084] The neural network model needs to be trained with a large amount of data to optimize its weights so that it can accurately predict or classify. In the present invention, the model is trained with positive samples (real target images), negative samples (images similar to the target but not the target) and perturbation samples (transformed forms of the original target images). Such a training strategy can improve the robustness and accuracy of the model.

[0085] Once the model is trained, it can extract features related to the target or center position from new images. These features are used to determine the center position of the target, i.e., the second center coordinate. Thereafter, the second center coordinate is compared with the preset ideal center position to obtain the second offset.

[0086] Based on the first and second center coordinates, and their corresponding offsets, an adjustment strategy can be determined to better align the camera with the target.

[0087] Through the pre-trained neural network model, features related to the target plate or center position can be efficiently extracted from the image to obtain a more accurate second center coordinate; by comparing with the preset ideal center position, the calculation of the second offset can provide key information for subsequent adjustment strategies; the first center coordinate provided by machine vision and pattern recognition and the second center coordinate obtained by neural network recognition both provide references for the adjustment strategy. Combining the information of the two can make the adjustment more accurate.

[0088] The positive samples include: target images taken at different distances, angles, and lighting conditions;

[0089] Positive samples usually refer to samples expected by the target task, that is, in the present invention, real target images. Target images taken from different distances, angles, and lighting conditions can provide diverse data, allowing the neural network to identify the center of the target in various environments. Such diversity is to improve the generalization ability of the neural network, so that it has higher accuracy and robustness in real applications.

[0090] Positive samples enable the neural network to accurately identify the center of the target in a variety of environments.

[0091] The negative samples include: images of other objects with similar shapes or features to the target;

[0092] Negative samples refer to samples of non-target tasks. In this scenario, they refer to images that have similar shapes or features to the target but are not actually the target. The neural network needs to learn to distinguish between the target object (target) and non-target objects during training. If only positive samples are used for training, the network will be too optimistic and will mistakenly identify other similar objects as targets in actual applications. The purpose of introducing negative samples is to enhance the model's discrimination ability.

[0093] Negative samples improve the discriminative ability of the model and reduce the risk of misidentification.

[0094] The perturbed sample comprises: transforming the original target plate image.

[0095] Perturbation samples are images obtained by transforming the original target image (such as rotation, scaling, translation, etc.). Perturbation samples can simulate various situations encountered in practical applications, such as slight rotation or translation of the target. By introducing these perturbation samples, the neural network can learn to recognize targets in various states, further enhancing its robustness.

[0096] The perturbed samples enable the model to maintain high accuracy in the face of various minor changes in the target in practical applications.

[0097] Preferably, in step S3, the adjustment mechanism includes:

[0098] Comparing the difference between the first center coordinate and the second center coordinate to obtain a first comparison result;

[0099] When the first comparison result is greater than the first threshold, adjusting the initial focal length or position of the camera according to the difference between the two;

[0100] In step S3, the task of the neural network is to identify the second center coordinate of the target. In step S2, the first center coordinate has been obtained using machine vision and pattern recognition. In order to evaluate the accuracy and stability of the two technologies, it is necessary to compare these two center coordinates.

[0101] The essence of comparison is the mathematical calculation of coordinate differences, usually the Euclidean distance between two points in two-dimensional space (or three-dimensional, depending on the needs of the task). Specifically, if the two coordinates are (x 1 ,y 1 ) and (x 2 ,y 2 ), then the distance between them is

[0102] The threshold is pre-set and is an upper limit of the tolerable deviation. Any deviation beyond this threshold is considered unacceptable.

[0103] When the first comparison result is greater than this threshold, it indicates that there is a large difference in the center coordinates given by the two technologies, which may be due to a problem with the position or focal length of the camera, or an error in one of the technologies itself.

[0104] If the comparison result exceeds the threshold, it is necessary to make corresponding adjustments. "Adjustment" here means resetting the position or focus of the camera to reduce the difference between the two center coordinates.

[0105] The principle of adjustment is: if the deviation between the two is caused by the position or focal length of the camera, then adjusting them will reduce this deviation.

[0106] The comparison and adjustment steps are designed to ensure that the center coordinates provided by the two technologies are as close as possible, thereby enhancing the overall accuracy of the system. In real applications, the camera or environment will undergo slight changes. Through this adjustment mechanism, the system can adapt to these changes in real time and maintain high accuracy.

[0107] Comparing the first offset and the second offset with a second threshold value respectively to obtain a second comparison result;

[0108] When the second comparison result is that the first offset is greater than the second threshold, adjusting the machine vision and pattern recognition;

[0109] When the second comparison result is that the first offset is greater than the second threshold, the training samples and parameters of the neural network model are adjusted;

[0110] When the second comparison result is that both the first offset and the second offset are greater than the second threshold, the initial position or focal length of the camera is adjusted;

[0111] The offset here refers to the difference between the identified center coordinates and the ideal center coordinates. The first offset comes from machine vision and pattern recognition, while the second offset comes from the neural network model; the second threshold is a pre-set value used to define the acceptable range of the offset. Any offset outside this range needs to be adjusted.

[0112] If only the first offset exceeds the second threshold, it means that the results of machine vision and pattern recognition are not ideal. To solve this problem, it is necessary to fine-tune its parameters or algorithms; the goal of the adjustment is to reduce the influence of noise and other interference factors in the image, so as to more accurately identify the center of the target plate.

[0113] If the second comparison result shows that the first offset is greater than the second threshold, it means that the neural network model has not learned the features in the image well. In this case, it is necessary to adjust the training samples or parameters of the model; the adjustment includes adding more training data, changing the structure of the model, or adjusting hyperparameters such as the learning rate.

[0114] When both the first and second offsets exceed the second threshold, it means that there is a problem with the position or focus of the camera. Therefore, the camera needs to be fine-tuned to align it more accurately with the center of the target plate.

[0115] Adjusting the position or focus of a camera can change its field of view to better capture the target.

[0116] Whether adjusting machine vision, neural network models, or the position or focal length of the camera, the goal is to reduce the offset and thus improve the accuracy of the overall system. Different adjustment strategies are selected based on different comparison results, so that the system can adapt to different situations to ensure that the best results are always obtained.

[0117] Adjustments are made according to the first comparison result and the second comparison result, and steps S1-S3 are re-executed until the first comparison result is less than the first threshold and the second comparison result is that both the first offset and the second offset are less than the second threshold, and then step S4 is executed.

[0118] The adjustment strategy of the whole system adopts an iterative method. At the beginning, a preliminary adjustment is performed based on the first comparison result and the second comparison result. Then steps S1-S3 are re-executed to obtain a new comparison result.

[0119] Both the first comparison result and the second comparison result involve comparison with preset thresholds, which serve as a criterion to determine whether the output of the system has achieved the expected accuracy.

[0120] The loop will continue until both comparison results meet certain conditions: the first comparison result is less than the first threshold, and the first offset and the second offset are both less than the second threshold. Meeting these conditions means that the output of the system has reached an acceptable accuracy and no further adjustment is required.

[0121] Through an iterative approach, the system can gradually correct potential errors, thereby improving overall accuracy; this method allows the system to automatically adjust according to the actual output, thereby achieving adaptive optimization. This means that no matter what the initial state of the system is, it can automatically adjust to a better state.

[0122] In one embodiment, assuming that in an actual application scenario, when the camera is initially set up, the first comparison result is 0.08 (the first threshold is 0.05), the first offset is 0.03, and the second offset is 0.06 (the second threshold is 0.04). In the first iteration, after adjustment according to the comparison result, steps S1-S3 are re-executed to obtain new comparison results of 0.04 and 0.03. At this time, the first comparison result is already less than the first threshold, but the second offset is still greater than the second threshold, so it is necessary to continue iterating. After the second iteration, assuming that all comparison results meet the conditions, the iteration can be stopped and step S4 can be entered.

[0123] S4. Based on the optical interferometry technology, the distance and angle data between the camera and the target plate are measured, and the focal length or position of the camera is fine-tuned;

[0124] Preferably, Figure 3 As shown, in step S4, based on the optical interferometry technology, the distance and angle between the camera and the target plate are measured through the second center coordinate, the first center coordinate, the second center coordinate, the distance and angle between the camera and the target plate are analyzed, and the focal length or position of the camera is fine-tuned based on the analysis result;

[0125] Optical interferometry is a technique that uses the phenomenon of light wave interference to measure distance and angle. When two coherent light waves meet, they will superimpose on each other to form interference fringes. Based on the shape and position of these fringes, the path difference of the light can be accurately calculated, thereby inferring the distance and angle.

[0126] Once the distance and angle data between the camera and the target plate are obtained, the focal length or position of the camera can be further fine-tuned to make the camera's field of view more aligned with the target plate, thereby improving the accuracy of the measurement.

[0127] By analyzing the first center coordinate, the second center coordinate, and the distance and angle data between the camera and the target plate, the relative position and angle between the camera and the target plate can be determined. If these data are different from the expected data, the focal length or position of the camera can be further fine-tuned to make the measurement result closer to the expected one.

[0128] By utilizing optical interferometry technology and fine-tuning the camera, the accuracy and reliability of the measurement can be greatly improved; this solution can automatically fine-tune the camera based on the actual measurement data to achieve the best measurement effect.

[0129] In one embodiment, it is assumed that in actual application, the distance between the camera and the target plate obtained by optical interferometry technology is 100.5 mm, which is 0.5 mm different from the expected 100 mm. At the same time, the angle data obtained is 0.2° different from the expected data. In this case, the system will automatically adjust the position of the camera so that the distance and angle are closer to the expected data. After fine-tuning, the new measurement data obtained is 100.1 mm and the expected angle, which greatly improves the accuracy of the measurement.

[0130] In general, the present invention effectively solves the influence of the position and angle between the camera and the target plate on the measurement accuracy by combining optical interference measurement technology and automatic fine-tuning mechanism, thereby achieving high-precision measurement.

[0131] The analysis results include: coordinate comparison, distance and angle verification, and technical accuracy assessment;

[0132] Coordinate comparison: Comparison of the first and second center coordinates: whether the two center coordinates measured are close or deviated. If there is a deviation, record the specific deviation.

[0133] Distance and angle verification: Use the two center coordinate data to predict the distance and angle between the camera and the target through the geometric model; compare the predicted value with the actual distance and angle obtained by optical interferometry in step S4, and record the difference.

[0134] Technical accuracy evaluation: Determine which center coordinate data (first or second) is closer to the data of optical interferometry, thereby determining which technology is more accurate.

[0135] Determining which technology is more accurate requires quantitative analysis, which can be evaluated using the following methods:

[0136] Difference calculation: For each center coordinate data (first and second), calculate its distance and angle difference with the optical interferometry measurement result. This can be obtained by calculating the Euclidean distance between the coordinates or other related geometric distance measurement methods.

[0137] Likewise, for angle data, the difference between the predicted angle and the actual measured angle is calculated.

[0138] Difference comparison: Compare the difference between the first center coordinate and the optical interferometry measurement result with the difference between the second center coordinate and the optical interferometry measurement result.

[0139] The technology whose coordinate data differs less from the data obtained by optical interferometry is more accurate.

[0140] Statistical analysis: If multiple measurements and comparisons are performed, these differences can be statistically analyzed, such as calculating the mean, standard deviation, etc., to evaluate the stability and reliability of the two technologies.

[0141] Use t-test or other relevant statistical test methods to determine whether the difference between the two technologies is significant.

[0142] Conclusion and Decision: Based on the above analysis, decide which technique is more accurate. If the difference of one technique is much smaller than the other technique, then it can be considered more accurate.

[0143] If the difference between the two technologies is close, then further research or consideration of other factors is needed to make a decision, such as which technology is easier to implement, less expensive, more stable, etc.

[0144] Among them, the fine-tuning corresponding to the analysis result includes: coordinate adjustment, distance and angle adjustment, and adjustment strategy suggestion.

[0145] Coordinate adjustment: If there is a significant difference between the first and second center coordinates, and one of the coordinates is closer to the optical interferometry result, that coordinate is selected as the subsequent reference and the camera position or focal length is adjusted accordingly.

[0146] Distance and angle adjustment: If the predicted distance and angle between the camera and the target plate are significantly different from the data obtained by optical interferometry, the camera is fine-tuned according to the direction and amount of the difference.

[0147] Suggested adjustment strategy: If the data provided by the machine vision and pattern recognition technology in step S2 is more accurate, then we can consider enhancing the application of this part and reducing or simplifying the neural network recognition process; if the data recognized by the neural network is more accurate, we can consider further optimizing and training the neural network model to improve its accuracy; if both methods are significantly different from the data of optical interferometry, then we need to re-evaluate the entire measurement and calibration process, determine the source of the problem, and formulate corresponding optimization strategies.

[0148] S5. Based on adaptive optical technology, the optical path of the camera is fine-tuned according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

[0149] Preferably, Figure 4 As shown, step S5 includes:

[0150] S501, evaluating optical parameters that need to be adjusted based on the first offset, the second offset, and the distance and angle between the camera and the target plate;

[0151] S502, using adaptive optics technology to fine-tune the optical path of the camera according to the optical parameters;

[0152] S503, collect the image of the target plate, and determine its center coordinates through machine vision and pattern recognition to determine whether it is close to the expected ideal center position. If so, complete the fine-tuning; if not, loop through steps S501-S503.

[0153] Adaptive optics is a method of adjusting optical systems in real time to counteract wavefront distortion. This distortion may be caused by external factors (such as atmospheric disturbances) or internal factors (such as imperfections in optical components). In camera applications, adaptive optics can be used to adjust the optical path to improve image quality.

[0154] In order to make effective fine-tuning, the current optical state needs to be evaluated first. Based on the known first offset, second offset, and the distance and angle between the camera and the target, the optical parameters that need to be adjusted, such as the path of the light path, the position of the focus, etc., can be evaluated.

[0155] Adaptive optics techniques, such as liquid crystal phase modulators or deformable mirrors, are used to fine-tune the optical path of the camera to achieve the desired optical parameters. This process involves changing the propagation path of light or adjusting the focus of light.

[0156] By fine-tuning the optical path through adaptive optics technology, the imaging quality of the camera can be improved, making the imaging result closer to the expected ideal center position; the present invention can fine-tune the optical path of the camera in real time, thereby quickly responding to any external or internal influencing factors and ensuring that the camera always works in the best optical state.

[0157] In one embodiment, it is assumed that in an actual application, the center coordinates of the target plate image captured by the camera have a certain offset. Based on the first offset and the second offset and other relevant data, the evaluation result shows that the focus position of the optical path needs to be adjusted. Therefore, the optical path of the camera is fine-tuned using adaptive optics technology, and the image of the target plate is captured again after the adjustment. The result shows that the center coordinates are closer to the expected ideal center position. If there is still an offset, the system will continue to make fine adjustments until a satisfactory effect is achieved.

[0158] In general, the present invention effectively improves the imaging quality of the camera by combining adaptive optics technology and real-time evaluation and adjustment strategy, so that it can provide high-quality imaging results under various conditions.

[0159] like Figure 5 As shown, a neural network and optical interference centering calibration system comprises:

[0160] The hardware module is an integration of a camera and an optical interferometer. The camera is located at the front end of the hardware module and is responsible for capturing images in the scene, including the image of the target plate. The optical interferometer is closely connected to the camera and is located at the rear or side of the camera without blocking the camera's field of view. The interference beam emitted by the optical interferometer is consistent with the shooting direction of the camera. The camera maintains a certain vertical distance from the target plate and faces the center of the target plate.

[0161] A camera is an optical sensor that captures and converts a physical scene into a digital image. Its main task is to capture an image of a scene, especially a target. An optical interferometer measurement device that uses the interference of two or more light sources to measure physical properties of an object, such as distance or angle. When combined with a camera, they work together to ensure that the interfering light beams are in the same direction as the camera to accurately measure the distance and angle to the target.

[0162] A preparation module, the preparation module is used to measure the optical distance between the camera and the target plate, and adjust the initial focal length or position of the camera according to the measurement result;

[0163] This module is responsible for the initial distance measurement to ensure that the camera is correctly aligned with the center of the target. Based on this measurement, preliminary focus or position adjustments are required.

[0164] A machine recognition module, wherein after the camera captures the image of the target plate, the machine recognition module is used to determine the target plate and the first center coordinates in the image through machine vision and pattern recognition, and determine a first offset between the first center coordinates and the ideal center position;

[0165] Using traditional machine vision and pattern recognition techniques, the module identifies the target from the image captured by the camera and locates its center coordinates. Further, it calculates the offset between this coordinate and the ideal center position.

[0166] A neural network recognition module, the neural network recognition module is used to input the image into a pre-trained neural network model for coordinate recognition, obtain the second center coordinate and the second offset, and determine and execute the adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset;

[0167] Using deep learning and machine learning techniques, the module further identifies the center coordinates of the target from the image. It uses a pre-trained neural network model that can perform more complex and in-depth analysis of the image, identify the second center coordinates, and calculate its offset from the ideal position.

[0168] A first fine-tuning module, which measures the distance and angle data between the camera and the target plate based on the optical interferometry technology, and fine-tunes the focal length or position of the camera;

[0169] Using optical interferometry, the module provides precise distance and angle data between the camera and the target. This data is then used to fine-tune the position or focus of the camera.

[0170] The second fine-tuning module is based on adaptive optical technology and fine-tunes the optical path of the camera according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

[0171] Using adaptive optics technology, the module fine-tunes the optical path of the camera to ensure that the captured image is optimal.

[0172] The system of the present invention achieves highly accurate centering calibration by combining cameras, optical interferometry equipment, traditional machine vision technology and advanced neural network technology. This system is designed to work in a variety of environments and conditions to ensure that high-quality imaging results are always obtained. This multi-stage calibration method ensures verification and adjustment at multiple levels to minimize errors.

[0173] In one embodiment, assuming that in an actual factory environment, the system is used in a critical part of a robotic assembly line to ensure that components are properly aligned. After the camera captures the image of the component, the system can determine the exact position of the component through the machine recognition module and the neural network recognition module. Then, based on the feedback from the optical interferometry device, the system can accurately know the distance and angle between the camera and the component. The system will then make the necessary fine-tuning to ensure perfect alignment between the component and the robotic assembly arm, thereby ensuring the quality and accuracy of the assembly.

[0174] Ultimately, through continued fine-tuning and calibration, the camera is able to precisely center the target, providing highly accurate and consistent results.

[0175] A storage medium stores a computer program, which, when executed by a processor, implements the steps of the neural network and optical interference centering calibration method.

[0176] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0177] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0178] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0179] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0180] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0181] Memory includes non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0182] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0183] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0184] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A neural network and optical interference centering calibration method, characterized in that: The following steps are involved: S1, measuring the optical distance between the camera and the target plate, and adjusting the initial focal length or position of the camera according to the measurement result; S2, after the camera captures the image of the target plate, the target plate and the first center coordinates in the image are determined through machine vision and pattern recognition, and a first offset between the first center coordinates and the ideal center position is determined; S3, input the image into a pre-trained neural network model for coordinate recognition, obtain the second center coordinate and the second offset, and determine and execute an adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset; the adjustment mechanism includes: Comparing the difference between the first center coordinate and the second center coordinate to obtain a first comparison result; Comparing the first offset and the second offset with a second threshold value respectively to obtain a second comparison result; Adjust according to the first comparison result and the second comparison result, and re-execute steps S1-S3 until the first comparison result is less than the first threshold and the second comparison result is that the first offset and the second offset are both less than the second threshold, then execute step S4; S4. Based on the optical interferometry technology, the distance and angle data between the camera and the target plate are measured, and the focal length or position of the camera is fine-tuned; S5. Based on adaptive optical technology, the optical path of the camera is fine-tuned according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

2. The neural network and optical interference centering calibration method according to claim 1, characterized in that: In step S1, the measurement result is compared with the ideal distance to obtain an initial deviation value, and the initial focal length or position of the camera is adjusted so that the measurement result reaches a predetermined ideal distance range.

3. The neural network and optical interference centering calibration method according to claim 2, characterized in that: Step S1 also includes, when the measurement result reaches the ideal distance range, executing step S2.

4. The neural network and optical interference centering calibration method according to claim 1, characterized in that: In step S2, after the camera captures the image of the target plate, the image is preprocessed, and the first center coordinate of the target plate in the image is identified by machine vision and pattern recognition, and the first center coordinate is compared with the coordinate of the ideal center position to obtain a first offset between the first center coordinate and the ideal center position; Wherein, the preprocessing includes: denoising and enhancement.

5. The neural network and optical interference centering calibration method according to claim 1, characterized in that: In step S3, the neural network model is pre-trained through positive samples, negative samples and perturbation samples. The neural network model is used to identify features related to the target plate or the center position from the image, analyze the extracted features, locate the center of the target plate, obtain the second center coordinates, and compare them with the preset ideal center position to calculate the second offset; The positive samples include: target images taken at different distances, angles, and lighting conditions; The negative samples include: images of other objects with similar shapes or features to the target; The perturbed sample comprises: transforming the original target plate image.

6. The neural network and optical interference centering calibration method according to claim 1, characterized in that: In step S3, the adjustment mechanism specifically includes: Comparing the difference between the first center coordinate and the second center coordinate to obtain a first comparison result; When the first comparison result is greater than the first threshold, adjusting the initial focal length or position of the camera according to the difference between the two; Comparing the first offset and the second offset with the second threshold respectively to obtain a second comparison result; When the second comparison result is that the first offset is greater than the second threshold, adjusting the machine vision and pattern recognition; When the second comparison result is that the first offset is greater than the second threshold, the training samples and parameters of the neural network model are adjusted; When the second comparison result is that both the first offset and the second offset are greater than the second threshold, the initial position or focal length of the camera is adjusted; Adjustments are made according to the first comparison result and the second comparison result, and steps S1-S3 are re-executed until the first comparison result is less than the first threshold and the second comparison result is that both the first offset and the second offset are less than the second threshold, and then step S4 is executed.

7. The neural network and optical interference centering calibration method according to claim 1, characterized in that: In step S4, based on the optical interferometry technology, the distance and angle between the camera and the target plate are measured through the second center coordinate, the first center coordinate, the second center coordinate, the distance and angle between the camera and the target plate are analyzed, and the focal length or position of the camera is fine-tuned based on the analysis result; The analysis results include: coordinate comparison, distance and angle verification, and technical accuracy assessment; Among them, the fine-tuning corresponding to the analysis result includes: coordinate adjustment, distance and angle adjustment, and adjustment strategy suggestion.

8. The neural network and optical interference centering calibration method according to claim 7, characterized in that: Step S5 includes: S501, evaluating optical parameters that need to be adjusted based on the first offset, the second offset, and the distance and angle between the camera and the target plate; S502, using adaptive optics technology to fine-tune the optical path of the camera according to the optical parameters; S503, collect the image of the target plate, and determine its center coordinates through machine vision and pattern recognition to determine whether it is close to the expected ideal center position. If so, complete the fine-tuning; if not, loop through steps S501-S503.

9. A neural network and optical interference centering calibration system, characterized in that: include: The hardware module is an integration of a camera and an optical interferometer. The camera is located at the front end of the hardware module and is responsible for capturing images in the scene, including the image of the target plate. The optical interferometer is closely connected to the camera and is located at the rear or side of the camera without blocking the camera's field of view. The interference beam emitted by the optical interferometer is consistent with the shooting direction of the camera. The camera maintains a certain vertical distance from the target plate and faces the center of the target plate. A preparation module, the preparation module is used to measure the optical distance between the camera and the target plate, and adjust the initial focal length or position of the camera according to the measurement result; A machine recognition module, wherein after the camera captures the image of the target plate, the machine recognition module is used to determine the target plate and the first center coordinates in the image through machine vision and pattern recognition, and determine a first offset between the first center coordinates and the ideal center position; A neural network recognition module is used to input the image into a pre-trained neural network model for coordinate recognition, obtain the second center coordinate and the second offset, and determine and execute an adjustment mechanism according to the first center coordinate, the first offset, the second center coordinate and the second offset; the adjustment mechanism includes: Comparing the difference between the first center coordinate and the second center coordinate to obtain a first comparison result; Comparing the first offset and the second offset with the second threshold respectively to obtain a second comparison result; Adjust according to the first comparison result and the second comparison result, and re-execute steps S1-S3 until the first comparison result is less than the first threshold and the second comparison result is that the first offset and the second offset are both less than the second threshold, then execute step S4; A first fine-tuning module, which measures the distance and angle data between the camera and the target plate based on the optical interferometry technology, and fine-tunes the focal length or position of the camera; The second fine-tuning module is based on adaptive optical technology and fine-tunes the optical path of the camera according to the first offset, the second offset, the distance between the camera and the target plate, and the angle data.

10. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the neural network and optical interference centering calibration method described in any one of claims 1 to 8 are implemented.

Images