A method for judging spot quality under dynamic background light and a related device thereof

Abstract

This application discloses a method and related apparatus for judging the quality of light spots under dynamic background light. The method includes: vectorizing good and bad light spot images acquired under different background lights to obtain good and bad light spot image matrices under different background lights; vectorizing multiple light spot images to be detected acquired under dynamic background light to obtain a light spot image matrix to be tested under dynamic background light; inputting the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the light spot image matrix to be tested under dynamic background light into a semi-supervised multi-view learning model for processing to obtain a predicted label matrix of the light spot image matrix to be tested under dynamic background light; judging the quality of each light spot image to be tested under dynamic background light by the values ​​in the predicted label matrix, thereby improving the technical problem that existing technologies are difficult to effectively detect the edge contour of light spots under the influence of dynamic background light, thus affecting the light spot quality judgment results.

Description

Technical Field

[0001] This application relates to the field of light spot quality discrimination technology, and in particular to a method and related apparatus for light spot quality discrimination under dynamic background light. Background Technology

[0002] Lasers have wide and important applications in fields requiring high precision, such as medical technology, military detection, satellite communications, and industrial processing. A key aspect is the detection of laser spot quality, which is the prerequisite and foundation for monitoring changes in laser position and a crucial factor determining the reliable application of laser technology.

[0003] Existing methods typically require segmenting the spot edge contour to detect spot quality. However, these methods are susceptible to the influence of background light. Although some techniques exist to address the effects of background light, existing techniques still struggle to effectively detect the spot edge contour under complex and variable background light (i.e., dynamic background light), thus affecting the spot quality judgment results. Summary of the Invention

[0004] This application provides a method and related apparatus for judging the quality of light spots under dynamic background light, which is used to improve the technical problem that the existing technology is difficult to effectively detect the edge contour of light spots under the influence of dynamic background light, thus affecting the result of light spot quality judgment.

[0005] In view of this, the first aspect of this application provides a method for judging the quality of light spots under dynamic background light, including:

[0006] Good and bad spot images acquired under different background lights are vectorized to obtain good and bad spot image matrices under different background lights. Multiple spot images to be detected acquired under dynamic background lights are vectorized to obtain spot image matrices to be detected under dynamic background lights.

[0007] The good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light are input into a semi-supervised multi-view learning model for processing to obtain the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0008] The quality of each of the test spot images under dynamic background light is determined by the values ​​in the predicted label matrix.

[0009] Optionally, the semi-supervised multi-view learning model is:

[0010]

[0011] st(W i ) T Wi =E1,Q T Q = E2, G ∈ {0, 1}

[0012] In the formula, X i Let Z be the good and bad spot image matrix under the i-th background light, and W be the spot image matrix to be tested. 1 ,…,W i ,…,W p Let W be the basis matrix under p different background lights. i ∈R m×k Let be the basis matrix under the i-th background light. Let k be the consensus representation matrix that is insensitive to background light, and k be the subspace dimension. F Let f(t) denote the Frobenius norm, and st denote compliance with constraints; E1∈R k×k E2∈R 2×2 It is the identity matrix; Q∈R k×2 For the projection matrix, Y∈{0,1} is the label matrix corresponding to the known good and bad spot image matrices under different background lights. is the predicted label matrix of the image matrix of the light spot to be tested under dynamic background light; β is the regularization parameter.

[0013] Optionally, the step of inputting the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into a semi-supervised multi-view learning model for processing to obtain the predicted label matrix of the spot image matrix to be tested under dynamic background light includes:

[0014] Initialize the target parameters of the semi-supervised multi-view learning model. The target parameters include the basis matrix, the consensus representation matrix that is insensitive to background light, the projection matrix, and the predicted label matrix of the test spot image matrix under dynamic background light.

[0015] Label the good and bad spot image matrices under different background lights to generate label matrices corresponding to the good and bad spot image matrices under different background lights. Input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into the semi-supervised multi-view learning model.

[0016] The semi-supervised multi-view learning model iteratively updates the target parameters based on the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light, until the semi-supervised multi-view learning model converges, thereby obtaining the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0017] Optionally, determining the quality of each of the test spot images under dynamic background light based on the values ​​in the predicted label matrix includes:

[0018] Each predicted label vector in the predicted label matrix is ​​compared with the good quality label vector and the bad quality label vector in the label matrix corresponding to the good and bad spot image matrix.

[0019] If the predicted label vector is the same as the good quality label vector, then the quality of the test spot image corresponding to the label vector is determined to be good.

[0020] If the predicted label vector is the same as the label vector indicating poor quality, then the quality of the light spot image to be tested corresponding to the label vector is determined to be poor.

[0021] A second aspect of this application provides a light spot quality discrimination device under dynamic background light, comprising:

[0022] The vectorization unit is used to vectorize good and bad spot images acquired under different background lights to obtain a matrix of good and bad spot images under different background lights. It also vectorizes multiple spot images to be detected acquired under dynamic background lights to obtain a matrix of spot images to be tested under dynamic background lights.

[0023] The prediction unit is used to input the good and bad spot image matrix under different background lights, the label matrix corresponding to the good and bad spot image matrix, and the spot image matrix to be tested under dynamic background light into a semi-supervised multi-view learning model for processing, so as to obtain the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0024] The discrimination unit is used to determine the quality of each of the test spot images under dynamic background light based on the values ​​in the prediction label matrix.

[0025] Optionally, the semi-supervised multi-view learning model is:

[0026]

[0027] st(W i ) T W i =E1,Q T Q = E2, G ∈ {0, 1}

[0028] In the formula, X i Let Z be the good and bad spot image matrix under the i-th background light, and W be the spot image matrix to be tested. 1 ,…,W i ,…,W p Let W be the basis matrix under p different background lights. i ∈Rm×k Let be the basis matrix under the i-th background light. Let k be the consensus representation matrix that is insensitive to background light, and k be the subspace dimension. F Let f(t) denote the Frobenius norm, and st denote compliance with constraints; E1∈R k×k E2∈R 2×2 It is the identity matrix; Q∈R k×2 For the projection matrix, Y∈{0,1} is the label matrix corresponding to the known good and bad spot image matrices under different background lights. is the predicted label matrix of the image matrix of the light spot to be tested under dynamic background light; β is the regularization parameter.

[0029] Optionally, the prediction unit is specifically used for:

[0030] Initialize the target parameters of the semi-supervised multi-view learning model. The target parameters include the basis matrix, the consensus representation matrix that is insensitive to background light, the projection matrix, and the predicted label matrix of the test spot image matrix under dynamic background light.

[0031] Label the good and bad spot image matrices under different background lights to generate label matrices corresponding to the good and bad spot image matrices under different background lights. Input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into the semi-supervised multi-view learning model.

[0032] The semi-supervised multi-view learning model iteratively updates the target parameters based on the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light, until the semi-supervised multi-view learning model converges, thereby obtaining the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0033] Optionally, the discrimination unit is specifically used for:

[0034] Each predicted label vector in the predicted label matrix is ​​compared with the good quality label vector and the bad quality label vector in the label matrix corresponding to the good and bad spot image matrix.

[0035] If the predicted label vector is the same as the good quality label vector, then the quality of the test spot image corresponding to the label vector is determined to be good.

[0036] If the predicted label vector is the same as the label vector indicating poor quality, then the quality of the light spot image to be tested corresponding to the label vector is determined to be poor.

[0037] A third aspect of this application provides a light spot quality discrimination device under dynamic background light, the device including a processor and a memory;

[0038] The memory is used to store program code and transmit the program code to the processor;

[0039] The processor is used to execute the light spot quality discrimination method under dynamic background light as described in the first aspect according to the instructions in the program code.

[0040] A fourth aspect of this application provides a computer-readable storage medium for storing program code, which, when executed by a processor, implements the light spot quality discrimination method under dynamic background light as described in any of the first aspects.

[0041] As can be seen from the above technical solutions, this application has the following advantages:

[0042] This application provides a method for judging the quality of light spots under dynamic background light, including: vectorizing good and bad light spot images acquired under different background lights to obtain good and bad light spot image matrices under different background lights; vectorizing multiple light spot images to be detected acquired under dynamic background light to obtain a light spot image matrix to be tested under dynamic background light; inputting the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the light spot image matrix to be tested under dynamic background light into a semi-supervised multi-view learning model for processing to obtain a predicted label matrix of the light spot image matrix to be tested under dynamic background light; and judging the quality of each light spot image to be tested under dynamic background light by the values ​​in the predicted label matrix.

[0043] In this application, known good and bad spot images under different background lights are acquired to obtain good and bad spot image matrices under different background lights. Multiple spot images to be detected under dynamic background lights are acquired to obtain a spot image matrix to be tested under dynamic background lights. A semi-supervised multi-view learning model is used to learn the background light insensitivity features of good and bad spot images under different background lights and their corresponding label matrices, as well as the projection matrix associated with the feature labels, based on the known good and bad spot image matrices under different background lights and the spot image matrix to be tested under dynamic background lights. This allows for the prediction of the label matrix of the spot image matrix to be tested under dynamic background lights. By utilizing the complementary information of spot images under different background lights, robustness to dynamic background lights is improved. Furthermore, the quality of each spot image to be tested is determined by the value of the predicted label matrix, without the need to segment the spot edge contour. This allows for batch determination of spot image quality, improving the technical problem in existing technologies where spot edge contour detection is difficult to perform effectively under the influence of dynamic background lights, thus affecting the spot quality determination results. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A schematic flowchart illustrating a method for judging the quality of light spots under dynamic background light, provided in an embodiment of this application;

[0046] Figure 2 This is a schematic diagram of a light spot quality discrimination device under dynamic background light provided in an embodiment of this application. Detailed Implementation

[0047] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0048] For easier understanding, please refer to Figure 1 This application provides a method for judging the quality of light spots under dynamic background light, including:

[0049] Step 101: Vectorize the good and bad spot images acquired under different background lights to obtain the good and bad spot image matrix under different background lights. Vectorize the multiple spot images to be detected acquired under dynamic background light to obtain the spot image matrix to be tested under dynamic background light.

[0050] For n1 known good and bad light spots, acquire p sets of good and bad light spot images (including good and bad light spot images) under different background lights, where each set of images is obtained under the same background light; vectorize each set of good and bad light spot images to obtain a matrix X of known good and bad light spot images under p different background lights. 1 ,…,X i ,…,X p ,in, This represents the good and bad light spot image matrix under the i-th background light. Let represent the vector of the n1-th good or bad light spot image under the i-th background light, where m is the number of features of the image, and R represents the real number domain. For n2 light spots to be detected under dynamic background light (complex and variable background light), the images of the light spots to be detected are acquired and vectorized to obtain a matrix of light spot images to be detected. Among them, z n2 ∈R m×1 This represents the vector of the n2th light spot image to be tested.

[0051] Step 102: Input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into the semi-supervised multi-view learning model for processing to obtain the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0052] X represents the image matrix of good and bad light spots under p different background lights. 1 ,…,X i ,…,X p The matrix Z of the test spot image under dynamic background light is input into the semi-supervised multi-view learning model to predict the label of each test spot image under dynamic background light.

[0053] In this embodiment of the application, the constructed semi-supervised multi-view learning model is as follows:

[0054]

[0055] st(W i ) T W i =E1,Q T Q = E2, G ∈ {0, 1}

[0056] In the formula, X i Let Z be the good and bad spot image matrix under the i-th background light, and W be the spot image matrix to be tested. 1 ,…,W i ,…,W p Let W be the basis matrix under p different background lights. i ∈R m×k Let be the basis matrix under the i-th background light. Let k be the consensus representation matrix that is insensitive to background light, and k be the subspace dimension. F Let f(t) denote the Frobenius norm, and st denote compliance with constraints; E1∈R k×k E2∈R 2×2 It is the identity matrix; Q∈R k×2 For the projection matrix, Y∈{0,1} is the label matrix corresponding to the known good and bad spot image matrices under different background lights. is the predicted label matrix of the image matrix of the light spot to be tested under dynamic background light; β is the regularization parameter.

[0057] After constructing the semi-supervised multi-view learning model, initialize the target parameters of the semi-supervised multi-view learning model. The target parameters include the basis matrix, the consensus representation matrix that is insensitive to background light, the projection matrix, and the prediction label matrix of the test spot image matrix under dynamic background light.

[0058] Label the good and bad spot image matrices under different background lights to generate label matrices corresponding to the good and bad spot image matrices under different background lights; input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into a semi-supervised multi-view learning model;

[0059] The target parameters are iteratively updated using a semi-supervised multi-view learning model based on the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light, until the semi-supervised multi-view learning model converges, thus obtaining the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0060] W 1 ,…,W i ,…,W p W in the basis matrix i The update formula is:

[0061] W i =B i C i ;

[0062] In the formula, B i C i The matrices [X] are respectively i ,Z]H T The left and right singular value matrices, where T is the matrix transpose;

[0063] The update formula for the consensus representation matrix H, which is insensitive to background light, is:

[0064]

[0065] In the formula, ∑ represents summation;

[0066] The update formula for the projection matrix Q is:

[0067] Q = DF;

[0068] In the formula, D and F represent matrices H[Y,G], respectively. T The left and right singular value matrices;

[0069] The predicted label matrix G of the image matrix of the light spot under dynamic background light is updated by updating each column of the update matrix, assuming g i Let the i-th column of G be... Let a1 be the (n1+i)th column of H, and let a1 = [1, 0]. T and a2 = [0,1] T ; calculate in sequence and The value of , where || ||2 represents the vector 2 norm;

[0070] like The value is less than The value of g is then i =a1;

[0071] like The value is greater than The value of g is then i =a2;

[0072] The update of G is completed by updating each column of G through the process described above.

[0073] Iterative update W 1 ,…,W i ,…,W p The process continues until the model converges, outputting the predicted label matrix G of the final updated image matrix of the light spot under dynamic background light. If the number of iterations exceeds the preset maximum number of iterations, the semi-supervised multi-view learning model can be considered converged. Alternatively, if the value of the semi-supervised multi-view learning model converges to a certain value, the semi-supervised multi-view learning model can be considered converged.

[0074] Step 103: Determine the quality of each test spot image under dynamic background light by predicting the values ​​in the label matrix.

[0075] Each predicted label vector in the predicted label matrix is ​​compared with the good quality label vector and the bad quality label vector in the label matrix corresponding to the good and bad spot image matrices. If the predicted label vector is the same as the good quality label vector, the quality of the spot image to be tested corresponding to the label vector is determined to be good; if the predicted label vector is the same as the bad quality label vector, the quality of the spot image to be tested corresponding to the label vector is determined to be bad.

[0076] Specifically, each column of the prediction label matrix G represents the predicted label vector of a test spot image under dynamic background lighting. Assuming that the label vector corresponding to good quality in the label matrix Y is s1, and the label vector corresponding to poor quality is s2, for the i-th (1≤i≤n2) test spot image, its predicted label vector is the i-th column of the prediction label matrix G, i.e., g i If g i If s1 is the same, then the quality of the corresponding i-th image of the light spot to be tested is judged to be good; if g i If the result is the same as s2, then the quality of the corresponding i-th image of the light spot to be tested is judged to be bad.

[0077] In this embodiment, known good and bad spot images under different background lights are acquired to obtain a matrix of good and bad spot images under different background lights. Multiple spot images to be detected under dynamic background lights are acquired to obtain a matrix of spot images to be tested under dynamic background lights. A semi-supervised multi-view learning model is used to learn the background light insensitivity features of good and bad spot images under different background lights and their corresponding label matrices and the matrix of spot images to be tested under dynamic background lights. The projection matrix associated with the feature labels can then be predicted. This can predict the label matrix of the spot image matrix to be tested under dynamic background lights. By utilizing the complementary information of spot images under different background lights, the robustness to dynamic background lights is improved. Furthermore, the quality of each spot image to be tested is determined by the value of the predicted label matrix. There is no need to segment the spot edge contours. The quality of spot images can be determined in batches. This improves the technical problem of existing technologies that are difficult to effectively detect the edge contours of spots under the influence of dynamic background lights, thus affecting the spot quality judgment results.

[0078] The above is an embodiment of a method for judging the quality of light spots under dynamic background light provided by this application. The following is an embodiment of a device for judging the quality of light spots under dynamic background light provided by this application.

[0079] Please refer to Figure 2 This application provides a light spot quality discrimination device under dynamic background light, comprising:

[0080] The vectorization unit is used to vectorize good and bad spot images acquired under different background lights to obtain a matrix of good and bad spot images under different background lights. It also vectorizes multiple spot images to be detected acquired under dynamic background lights to obtain a matrix of spot images to be tested under dynamic background lights.

[0081] The prediction unit is used to input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into the semi-supervised multi-view learning model for processing, so as to obtain the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0082] The discrimination unit is used to determine the quality of each test spot image under dynamic background light by predicting the values ​​in the label matrix.

[0083] Optionally, a semi-supervised multi-view learning model is:

[0084]

[0085] st(W i ) T W i =E1,Q TQ = E2, G ∈ {0, 1}

[0086] In the formula, X i Let Z be the good and bad spot image matrix under the i-th background light, and W be the spot image matrix to be tested. 1 ,…,W i ,…,W p Let W be the basis matrix under p different background lights. i ∈R m×k Let be the basis matrix under the i-th background light. Let k be the consensus representation matrix that is insensitive to background light, and k be the subspace dimension. F Let f(t) denote the Frobenius norm, and st denote compliance with constraints; E1∈R k×k E2∈R 2×2 It is the identity matrix; Q∈R k×2 For the projection matrix, Y∈{0,1} is the label matrix corresponding to the known good and bad spot image matrices under different background lights. is the predicted label matrix of the image matrix of the light spot to be tested under dynamic background light; β is the regularization parameter.

[0087] Optional, prediction unit, specifically used for:

[0088] Initialize the target parameters of the semi-supervised multi-view learning model. The target parameters include the basis matrix, the consensus representation matrix that is insensitive to background light, the projection matrix, and the predicted label matrix of the test spot image matrix under dynamic background light.

[0089] Label the good and bad spot image matrices under different background lights to generate label matrices corresponding to the good and bad spot image matrices under different background lights. Input the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light into the semi-supervised multi-view learning model.

[0090] The target parameters are iteratively updated using a semi-supervised multi-view learning model based on the good and bad spot image matrices under different background lights, the label matrices corresponding to the good and bad spot image matrices, and the spot image matrix to be tested under dynamic background light, until the semi-supervised multi-view learning model converges, thus obtaining the predicted label matrix of the spot image matrix to be tested under dynamic background light.

[0091] Optional, the discrimination unit is specifically used for:

[0092] Compare each predicted label vector in the predicted label matrix with the good quality label vector and the bad quality label vector in the label matrix corresponding to the good and bad spot image matrix, respectively;

[0093] If the predicted label vector is the same as the high-quality label vector, then the quality of the test spot image corresponding to the label vector is determined to be good.

[0094] If the predicted label vector is the same as the label vector of poor quality, then the quality of the light spot image to be tested corresponding to the label vector is determined to be poor.

[0095] In this embodiment, known good and bad spot images under different background lights are acquired to obtain a matrix of good and bad spot images under different background lights. Multiple spot images to be detected under dynamic background lights are acquired to obtain a matrix of spot images to be tested under dynamic background lights. A semi-supervised multi-view learning model is used to learn the background light insensitivity features of good and bad spot images under different background lights and their corresponding label matrices and the matrix of spot images to be tested under dynamic background lights. The projection matrix associated with the feature labels can then be predicted. This can predict the label matrix of the spot image matrix to be tested under dynamic background lights. By utilizing the complementary information of spot images under different background lights, the robustness to dynamic background lights is improved. Furthermore, the quality of each spot image to be tested is determined by the value of the predicted label matrix. There is no need to segment the spot edge contours. The quality of spot images can be determined in batches. This improves the technical problem of existing technologies that are difficult to effectively detect the edge contours of spots under the influence of dynamic background lights, thus affecting the spot quality judgment results.

[0096] This application embodiment also provides a light spot quality discrimination device under dynamic background light, the device including a processor and a memory;

[0097] The memory is used to store program code and transfer the program code to the processor;

[0098] The processor is used to execute the light spot quality discrimination method under dynamic background light in the aforementioned method embodiment according to the instructions in the program code.

[0099] This application also provides a computer-readable storage medium for storing program code, which, when executed by a processor, implements the light spot quality discrimination method under dynamic background light in the aforementioned method embodiments.

[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0101] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0102] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0103] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0104] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0105] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0106] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the various embodiments of this application through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0107] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for determining the quality of a light spot under dynamic background light, characterized in that, The method comprises the steps of: vectorizing good and bad light spot images collected under different background lights to obtain good and bad light spot image matrices under different background lights, vectorizing a plurality of to-be-detected light spot images collected under dynamic background light to obtain a to-be-detected light spot image matrix under dynamic background light; inputting the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the to-be-detected light spot image matrix under dynamic background light into a semi-supervised multi-view learning model for processing to obtain a predicted label matrix of the to-be-detected light spot image matrix under dynamic background light; the semi-supervised multi-view learning model is: s.t. (W i ) T W i = E1,Q T Q = E2,G ∈ {0,1} where X i is the good-bad spot image matrix under the i-th background light, Z is the to-be-detected spot image matrix, W 1 ,…,W i ,…,W p are the base matrices under p different background lights, W i ∈R m×k is the base matrix under the i-th background light, is the consensus representation matrix insensitive to the background light, k is the subspace dimension, || || F represents the Frobenius norm, s.t. represents subject to constraints; E1∈R k×k , E2∈R 2×2 is the unit matrix; Q∈R k×2 is the projection matrix, Y∈{0,1} is the label matrix corresponding to the good-bad spot image matrix under different known background lights, is the predicted label matrix of the to-be-detected spot image matrix under the dynamic background light; β is a regularization parameter; determining the quality of each to-be-detected light spot image under dynamic background light by values in the predicted label matrix.

2. The method of claim 1, wherein the method further comprises: The step of inputting the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the to-be-detected light spot image matrix under dynamic background light into the semi-supervised multi-view learning model for processing to obtain the predicted label matrix of the to-be-detected light spot image matrix under dynamic background light comprises the steps of: initializing target parameters of the semi-supervised multi-view learning model, wherein the target parameters comprise a base matrix, a background-light-insensitive consensus representation matrix, a projection matrix, and a predicted label matrix of the to-be-detected light spot image matrix under dynamic background light; labeling the good and bad light spot image matrices under different background lights to generate label matrices corresponding to the good and bad light spot image matrices under different background lights, and inputting the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the to-be-detected light spot image matrix under dynamic background light into the semi-supervised multi-view learning model; iteratively updating the target parameters according to the good and bad light spot image matrices under different background lights, the label matrices corresponding to the good and bad light spot image matrices, and the to-be-detected light spot image matrix under dynamic background light by the semi-supervised multi-view learning model until the semi-supervised multi-view learning model converges, to obtain the predicted label matrix of the to-be-detected light spot image matrix under dynamic background light.

3. The method of claim 1, wherein the method further comprises: The step of determining the quality of each to-be-detected light spot image under dynamic background light by values in the predicted label matrix comprises the steps of: respectively comparing each predicted label vector in the predicted label matrix with a good-quality label vector and a bad-quality label vector in the label matrix corresponding to the good and bad light spot image matrices; if the predicted label vector is the same as the good-quality label vector, determining that the quality of the to-be-detected light spot image corresponding to the label vector is good; if the predicted label vector is the same as the bad-quality label vector, determining that the quality of the to-be-detected light spot image corresponding to the label vector is bad.

4. A device for determining the quality of a light spot under dynamic background light, characterized in that The method comprises the steps of: vectorizing good and bad light spot images collected under different background lights to obtain good and bad light spot image matrices under different background lights, and vectorizing a plurality of to-be-detected light spot images collected under dynamic background light to obtain a to-be-detected light spot image matrix under dynamic background light; The prediction unit is configured to input the good and bad spot image matrix under different background lights, the label matrix corresponding to the good and bad spot image matrix, and the to-be-tested spot image matrix under dynamic background light into a semi-supervised multi-view learning model for processing to obtain a predicted label matrix of the to-be-tested spot image matrix under dynamic background light. s.t. (W i ) T W i = E1, Q T Q = E2, G ∈ {0, 1} In the formula, X i Let Z be the good and bad spot image matrix under the i-th background light, and W be the spot image matrix to be tested. 1 ,…,W i ,…,W p Let W be the basis matrix under p different background lights. i ∈R m×k Let be the basis matrix under the i-th background light. Let k be the consensus representation matrix that is insensitive to background light, and k be the subspace dimension. F Let f(t) denote the Frobenius norm, and st denote compliance with constraints; E1∈R k×k E2∈R 2×2 It is the identity matrix; Q∈R k×2 Let be the projection matrix. Y∈{0,1} is the label matrix corresponding to the known good and bad spot image matrices under different background lights. is the predicted label matrix of the image matrix of the light spot to be tested under dynamic background light; β is the regularization parameter; The discrimination unit is configured to discriminate the quality of each to-be-tested spot image under dynamic background light according to the value in the predicted label matrix.

5. The apparatus of claim 4, wherein the apparatus is configured to determine the quality of the light spot based on the intensity of the light spot and the intensity of the dynamic background light. The prediction unit is specifically configured to: initialize target parameters of the semi-supervised multi-view learning model, wherein the target parameters include a basis matrix, a background light-insensitive consensus representation matrix, a projection matrix, and a predicted label matrix of the to-be-tested spot image matrix under dynamic background light; perform label annotation on the good and bad spot image matrix under different background lights to generate a label matrix corresponding to the good and bad spot image matrix under different background lights, and input the good and bad spot image matrix under different background lights, the label matrix corresponding to the good and bad spot image matrix, and the to-be-tested spot image matrix under dynamic background light into the semi-supervised multi-view learning model; perform iterative update on the target parameters according to the good and bad spot image matrix under different background lights, the label matrix corresponding to the good and bad spot image matrix, and the to-be-tested spot image matrix under dynamic background light through the semi-supervised multi-view learning model until the semi-supervised multi-view learning model converges, so as to obtain the predicted label matrix of the to-be-tested spot image matrix under dynamic background light.

6. The apparatus of claim 4, wherein the apparatus further comprises a light source configured to provide a light source signal to the light source driver. The discrimination unit is specifically configured to: respectively compare each predicted label vector in the predicted label matrix with a good-quality label vector and a bad-quality label vector in the label matrix corresponding to the good and bad spot image matrix; if the predicted label vector is the same as the good-quality label vector, it is determined that the quality of the to-be-tested spot image corresponding to the label vector is good; if the predicted label vector is the same as the bad-quality label vector, it is determined that the quality of the to-be-tested spot image corresponding to the label vector is bad.

7. A device for determining the quality of a light spot under dynamic background light, characterized in that The device comprises a processor and a memory. The memory is configured to store program code and transmit the program code to the processor. The processor is configured to execute the instructions in the program code to perform the spot quality discrimination method under dynamic background light according to any one of claims 1-3.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium is configured to store program code, and the program code is executed by the processor to implement the spot quality discrimination method under dynamic background light according to any one of claims 1-3.

Images