Contact lens defect detection method and system

Abstract

The invention discloses a contact lens defect detection method and system, which is applied to the field of contact lens detection, and comprises the following steps: constructing a lens image acquisition system of a contact lens, and setting parameters of the lens image acquisition system according to the type of the contact lens and the position of a defect; wherein the lens image acquisition system comprises a bright-field illumination light source and a dark-field illumination light source which are not on the same optical axis, and a light beam splitter for synthesizing the bright-field illumination light source and the dark-field illumination light source and transmitting the synthesized light to the surface of the contact lens; and inputting a lens image acquired by the lens image acquisition system into a pre-trained neural network model for defect detection and identification. According to the method, the lens image acquisition efficiency is effectively improved, the defect identification efficiency and precision are improved, the professional requirement of contact lens defect identification is reduced, and the contact lens defect identification efficiency is effectively improved on the basis of ensuring the subsequent normal utilization rate and reducing the defect return rate.

Description

Technical Field

[0001] The present invention relates to the field of contact lens detection, and in particular to a contact lens defect detection method and system. Background Art

[0002] As one of the important links in the manufacture of contact lenses, the quality inspection of contact lenses not only determines the normal use rate and defective return rate of subsequent contact lenses, but also restricts the manufacturing efficiency of contact lenses. Therefore, on the basis of ensuring the normal use rate and reducing the defective return rate, effectively improving the manufacturing efficiency of contact lenses is of great practical significance for the quality assurance of contact lenses and their widespread popularization in the future.

[0003] At present, the existing contact lens defect detection method is mainly completed in two parts. The first part is based on the fusion processing of bright field light source images and dark field light source images, and the second part is defect recognition based on the fusion processing image. Among them, the fusion processing based on bright field light source images and dark field light source images requires obtaining the images under the bright field light source image and the dark field light source image respectively, and using relevant algorithms for image fusion. The operation process is relatively cumbersome; while the defect recognition based on the fusion processing image relies on manual recognition, which is not only inefficient, but also has accuracy affected by subjectivity, and has high requirements for professionalism. In summary, the current contact lens defect detection method has the disadvantages of cumbersome lens image acquisition process, low defect recognition efficiency and accuracy, and high professional requirements, which seriously restricts the manufacturing efficiency of contact lenses and hinders the widespread popularization of contact lenses.

[0004] To this end, how to provide a contact lens defect detection method and system that can effectively improve the efficiency of lens image acquisition, improve defect recognition efficiency and accuracy, and reduce the professional requirements for contact lens defect recognition, while ensuring the subsequent normal usage rate and reducing the defect return rate, effectively improve the contact lens defect recognition efficiency, improve the manufacturing efficiency of contact lenses, and promote the widespread popularization of contact lenses is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0005] In view of this, the present invention proposes a contact lens defect detection method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A contact lens defect detection method, comprising:

[0008] Step 1: construct a lens image acquisition system for contact lenses, and set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system includes: a bright field illumination light source and a dark field illumination light source that are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens;

[0009] Step 2: Input the lens image collected by the lens image acquisition system into the pre-trained neural network model for defect detection and identification.

[0010] Optionally, in step 1, the beam splitter is selected to be 50% transmission or 50% reflection.

[0011] Optionally, in step 1, the bright field illumination light source is a planar array light emitting diode, and the dark field illumination light source is a circular array light emitting diode, and the circular array light emitting diode has a tilt angle deviating from 0 degrees.

[0012] Optionally, in step 1, parameters of the lens image acquisition system are set according to the contact lens type and the location of the defect, specifically:

[0013] The tilt angle value range of the annular array light emitting diode is set according to the position of the defect;

[0014] According to the type of contact lens, the tilt angle of the ring array light emitting diode is changed within the tilt angle value range;

[0015] Based on the determined tilt angle after the change, the distance value from the synthetic light source to the contact lens is determined.

[0016] Optionally, the tilt angle value range of the annular array light emitting diode is set according to the location of the defect, specifically:

[0017] When the defect is located in the first peripheral arc area and the second peripheral arc area, the tilt angle of the annular array light emitting diode is greater than the first preset angle threshold;

[0018] When the defect is located in the optical zone, the tilt angle of the annular array light emitting diode is less than a second preset angle threshold;

[0019] Among them, the first preset angle threshold is greater than the second preset angle threshold.

[0020] Optionally, according to the type of contact lens, the tilt angle of the annular array light emitting diode is changed within the tilt angle value range, specifically:

[0021] A relationship function between the contact lens power and the tilt angle value is constructed over the tilt angle value range, specifically:

[0022] When the contact lens type is myopia contact lens, the relationship function is as follows:

[0023]

[0024] Among them, y is the tilt angle value; x is the degree of the contact lens; a is the relationship coefficient between the tilt angle and degree of the myopic contact lens;

[0025] When the contact lens type is a hyperopic contact lens, the relationship function is as follows:

[0026] y = bx;

[0027] Where b is the coefficient of the relationship between the tilt angle and degree of the hyperopic contact lens;

[0028] Based on the relationship function, the tilt angles corresponding to different contact lens types are changed within the tilt angle value range.

[0029] Optionally, based on the tilt angle determined after the change, the distance value from the synthetic light source to the contact lens is determined, specifically:

[0030] Construct the relationship function between the tilt angle and the distance from the synthetic light source to the contact lens as follows:

[0031]

[0032] Among them, z is the distance from the light source to the contact lens; y is the tilt angle; c is the relationship coefficient between the distance from the light source to the contact lens and the tilt angle.

[0033] Optionally, in step 2, the neural network model includes: a first neural network sub-model and a second neural network sub-model, wherein the first neural network sub-model performs spatial transformation processing on the lens image to obtain a corresponding first image; the second neural network sub-model performs dimensionality reduction processing on the lens image to obtain a corresponding second image; the first image and the second image are fused to obtain a corresponding third image; the Softmax function is used to classify and predict image features of the third image to obtain classification probabilities corresponding to preset contact lens defect types, and the defects of the lens image are identified according to the classification probabilities.

[0034] Optionally, in step 2, the pre-training process of the neural network model is specifically as follows:

[0035] Constructing training samples: collecting lens images, dividing the lens images into grids according to the image pyramid, and constructing a lens image database containing multi-scale lens image blocks; marking defective areas of the lens images, and amplifying the lens image blocks in the lens image database to obtain an amplified lens image database;

[0036] Constructing a convolutional neural network: A semantic segmentation network composed of a fully convolutional neural network structure is used. The fully convolutional neural network structure consists of only multiple convolutional layers and deconvolutional layers, and does not include a fully connected layer. It is a convolutional neural network structure from image to image.

[0037] Training convolutional neural network: Use training samples to train the parameters of the convolutional neural network using the back propagation algorithm to obtain a trained neural network model.

[0038] The present invention also provides a contact lens defect detection system using a contact lens defect detection method, comprising:

[0039] Lens image acquisition system construction module: used to construct a lens image acquisition system for contact lenses, and set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system includes: a bright field illumination light source and a dark field illumination light source that are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens;

[0040] Defect detection and identification module: used to input the lens image collected by the lens image acquisition system into a pre-trained neural network model for defect detection and identification.

[0041] It can be seen from the above technical solutions that compared with the prior art, the present invention proposes a contact lens defect detection method and system. By constructing a lens image acquisition system including a bright field illumination light source and a dark field illumination light source that are not on the same optical axis and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens, and setting the parameters of the lens image acquisition system according to the type of contact lens and the location of the defect, it is achieved that only one imaging is required to obtain a high-precision contact lens mirror image, and on the basis of ensuring the subsequent normal usage rate and reducing the defect return rate, the contact lens defect recognition efficiency is effectively improved, the contact lens manufacturing efficiency is improved, and the contact lens is widely popularized; the pre-trained neural network model is used instead of manual lens image defect detection and recognition, which effectively reduces the accuracy impact caused by human subjectivity and improves the defect detection efficiency, while reducing the professional requirements, effectively improving the contact lens defect recognition efficiency, improving the contact lens manufacturing efficiency, and promoting the widespread popularization of contact lenses. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0043] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Embodiment 1:

[0046] Embodiment 1 of the present invention discloses a contact lens defect detection method, comprising:

[0047] Step 1: Construct a lens image acquisition system for contact lenses, and set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system includes: a bright field illumination light source and a dark field illumination light source that are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens.

[0048] The optical axes of the initial outgoing light between the bright field illumination light source and the dark field illumination light source are respectively the first optical axis and the second optical axis, and the first optical axis and the second optical axis are perpendicular to each other. After being synthesized by the beam splitter, they are incident on the target along the first optical axis.

[0049] In order to ensure that the relative strength of the two light sources remains unchanged after passing through the beam splitter, the beam splitter is selected to be 50% transmittance or 50% reflection.

[0050] The bright field illumination light source is a planar array light emitting diode, and the dark field illumination light source is a circular array light emitting diode, and the circular array light emitting diode has a tilt angle deviating from 0 degrees.

[0051] Among them, the tilt angle is the angle between the normal direction of the luminous surface and the direction of the optical axis.

[0052] According to the contact lens type and defect location, the parameters of the lens image acquisition system are set as follows:

[0053] The tilt angle value range of the annular array light emitting diode is set according to the position of the defect;

[0054] According to the type of contact lens, the tilt angle of the ring array light emitting diode is changed within the tilt angle value range;

[0055] Based on the determined tilt angle after the change, the distance value from the synthetic light source to the contact lens is determined.

[0056] The tilt angle value range of the annular array light emitting diode is set according to the location of the defect, specifically:

[0057] The structural characteristics of the lens mainly refer to the angle between the normal direction of the lens surface and the optical axis direction perpendicular to the horizontal plane; wherein, compared with the first peripheral arc and the second peripheral arc, the angle between the surface normal direction of the lens optical zone and the optical axis direction is smaller, so a ring array light emitting diode with a smaller tilt angle can be selected. Therefore, when the defect is located in the first peripheral arc and the second peripheral arc area, the tilt angle of the ring array light emitting diode is greater than the first preset angle threshold;

[0058] When the defect is located in the optical zone, the tilt angle of the annular array of light emitting diodes is set to be smaller than a second preset angle threshold;

[0059] The first preset angle threshold is greater than the second preset angle threshold. Both the first preset angle threshold and the second preset angle threshold can be set according to actual needs.

[0060] According to the type of contact lens, the tilt angle of the annular array light emitting diode is changed within the tilt angle value range, specifically:

[0061] A relationship function between the contact lens power and the tilt angle value is constructed over the tilt angle value range, specifically:

[0062] For lenses with focal length, since the higher the myopia, the stronger the light deflection ability, a ring array LED with a smaller tilt angle can be selected. Therefore, when the contact lens type is a myopia contact lens, the relationship function is as follows:

[0063]

[0064] Among them, y is the tilt angle value; x is the degree of the contact lens; a is the relationship coefficient between the tilt angle and degree of the myopic contact lens;

[0065] On the contrary, the higher the degree of hyperopia, the stronger the light convergence ability, so a ring array LED with a larger tilt angle can be selected. Therefore, when the contact lens type is a hyperopic contact lens, the relationship function is as follows:

[0066] y = bx;

[0067] Where b is the coefficient of the relationship between the tilt angle and degree of the hyperopic contact lens;

[0068] Based on the relationship function, the tilt angles corresponding to different contact lens types are changed within the tilt angle value range.

[0069] Based on the determined tilt angle after the change, the distance value from the synthetic light source to the contact lens is determined, specifically:

[0070] Due to the limitation of lateral space, in order to make the light evenly and fully illuminate the contact lens to be tested, the ring array LED with a smaller tilt angle can be placed slightly away from the contact lens to be tested; on the contrary, the ring array LED with a larger tilt angle needs to be placed slightly closer to the contact lens to be tested. Therefore, the relationship function between the tilt angle and the distance from the synthetic light source to the contact lens is constructed as follows:

[0071]

[0072] Among them, z is the distance from the light source to the contact lens; y is the tilt angle; c is the relationship coefficient between the distance from the light source to the contact lens and the tilt angle.

[0073] Step 2: Input the lens image collected by the lens image acquisition system into the pre-trained neural network model for defect detection and identification.

[0074] The neural network model includes: a first neural network sub-model and a second neural network sub-model, wherein the first neural network sub-model performs spatial transformation processing on the lens image to obtain a corresponding first image; the second neural network sub-model performs dimensionality reduction processing on the lens image to obtain a corresponding second image; the first image and the second image are image-fused to obtain a corresponding third image; a Softmax function is used to classify and predict the image features of the third image to obtain classification probabilities corresponding to preset contact lens defect types, and defects in the lens image are identified according to the classification probabilities.

[0075] The Softmax function is a function used by the convolutional neural network for classification. The Softmax function is used to map the image features of the third image obtained after image fusion to the interval (0, 1) according to the preset contact lens defect type, and obtain the classification probabilities corresponding to the contact lens defect type. Then, the defects of the contact lens mirror image are identified according to the high and low classification probabilities corresponding to the contact lens defect type to determine the defect type corresponding to the contact lens mirror image, where the contact lens mirror image includes but is not limited to: scratch defects, edge damage defects, etc.

[0076] The pre-training process of the neural network model is as follows:

[0077] Constructing training samples: collecting lens images, dividing the lens images into grids according to the image pyramid, and constructing a lens image database containing multi-scale lens image blocks; marking defective areas of the lens images, and amplifying the lens image blocks in the lens image database to obtain an amplified lens image database;

[0078] Constructing a convolutional neural network: A semantic segmentation network composed of a fully convolutional neural network structure is used. The fully convolutional neural network structure consists of only multiple convolutional layers and deconvolutional layers, and does not include a fully connected layer. It is a convolutional neural network structure from image to image.

[0079] Training convolutional neural network: Use training samples to train the parameters of the convolutional neural network using the back propagation algorithm to obtain a trained neural network model.

[0080] Embodiment 2:

[0081] Embodiment 2 of the present invention further provides a contact lens defect detection system using a contact lens defect detection method, comprising:

[0082] Lens image acquisition system construction module: used to construct a lens image acquisition system for contact lenses, and set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system includes: a bright field illumination light source and a dark field illumination light source that are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens;

[0083] Defect detection and identification module: used to input the lens image collected by the lens image acquisition system into a pre-trained neural network model for defect detection and identification.

[0084] The embodiment of the present invention discloses a contact lens defect detection method and system. By constructing a lens image acquisition system including a bright field illumination light source and a dark field illumination light source that are not on the same optical axis and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens, and setting the parameters of the lens image acquisition system according to the type of contact lens and the location of the defect, it is achieved that only one imaging is required to obtain a high-precision contact lens mirror image, and on the basis of ensuring the subsequent normal usage rate and reducing the defect return rate, the contact lens defect recognition efficiency is effectively improved, the contact lens manufacturing efficiency is improved, and the contact lens is widely popularized; by using a pre-trained neural network model instead of manual lens image defect detection and recognition, the accuracy impact caused by human subjectivity is effectively reduced, and the defect detection efficiency is improved, while reducing the professional requirements, effectively improving the contact lens defect recognition efficiency, improving the contact lens manufacturing efficiency, and promoting the widespread popularization of contact lenses.

[0085] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0086] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A contact lens defect detection method, characterized in that: include: Step 1: construct a lens image acquisition system for contact lenses, and set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system includes: a bright field illumination light source and a dark field illumination light source that are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens; Step 2: Input the lens image acquired by the lens image acquisition system into a pre-trained neural network model for defect detection and identification.

2. A contact lens defect detection method according to claim 1, characterized in that: In step 1, the beam splitter is selected to be 50% transmission or 50% reflection.

3. A contact lens defect detection method according to claim 1, characterized in that: In step 1, the bright field illumination light source is a planar array light emitting diode, and the dark field illumination light source is a circular array light emitting diode, and the circular array light emitting diode has a tilt angle deviating from 0 degrees.

4. A contact lens defect detection method according to claim 3, characterized in that: In step 1, the parameters of the lens image acquisition system are set according to the contact lens type and the location of the defect, specifically: The tilt angle value range of the annular array light emitting diode is set according to the position of the defect; According to the type of contact lens, the tilt angle of the annular array light emitting diode is changed within the tilt angle value range; Based on the determined tilt angle after the change, a distance value from the synthetic light source to the contact lens is determined.

5. A contact lens defect detection method according to claim 4, characterized in that: The tilt angle value range of the annular array light emitting diode is set according to the location of the defect, specifically: When the defect is located in the first peripheral arc area and the second peripheral arc area, the inclination angle of the annular array light emitting diode is greater than a first preset angle threshold; When the defect is located in the optical zone, the tilt angle of the annular array of light emitting diodes is less than a second preset angle threshold; Wherein, the first preset angle threshold is greater than the second preset angle threshold.

6. A contact lens defect detection method according to claim 4, characterized in that: According to the type of contact lens, the tilt angle of the annular array light emitting diode is changed within the tilt angle value range, specifically: A relationship function between the contact lens power and the tilt angle value is constructed over the tilt angle value range, specifically: When the contact lens type is a myopic contact lens, the relationship function is as follows: Wherein, y is the tilt angle value; x is the degree of the contact lens; a is the relationship coefficient between the tilt angle and degree of the myopic contact lens; When the contact lens type is a hyperopic contact lens, the relationship function is as follows: y = bx; Wherein, b is the relationship coefficient between the tilt angle and degree of the hyperopic contact lens; Based on the relationship function, the tilt angles corresponding to different contact lens types are changed within the tilt angle value range.

7. A contact lens defect detection method according to claim 4, characterized in that: Based on the determined tilt angle after the change, the distance value from the synthetic light source to the contact lens is determined, specifically: The relationship function between the tilt angle and the distance from the synthetic light source to the contact lens is constructed as follows: Among them, z is the distance value from the light source to the contact lens; y is the tilt angle value; c is the relationship coefficient between the distance value from the light source to the contact lens and the tilt angle.

8. A contact lens defect detection method according to claim 1, characterized in that: In step 2, the neural network model includes: a first neural network sub-model and a second neural network sub-model, wherein the first neural network sub-model performs spatial transformation processing on the lens image to obtain a corresponding first image; the second neural network sub-model performs dimensionality reduction processing on the lens image to obtain a corresponding second image; the first image and the second image are fused to obtain a corresponding third image; a Softmax function is used to classify and predict image features of the third image to obtain classification probabilities corresponding to preset contact lens defect types, and defects of the lens image are identified according to the classification probabilities.

9. A contact lens defect detection method according to claim 1, characterized in that: In step 2, the pre-training process of the neural network model is specifically as follows: Constructing training samples: collecting lens images, and dividing the lens images into grids according to the image pyramid, and constructing a lens image database containing multi-scale lens image blocks; wherein, areas where defects exist in the lens images are marked, and the lens image blocks in the lens image database are amplified to obtain an amplified lens image database; Constructing a convolutional neural network: A semantic segmentation network consisting of a fully convolutional neural network structure is used. The fully convolutional neural network structure consists only of multiple convolutional layers and deconvolutional layers, and does not include a fully connected layer. It is a convolutional neural network structure from image to image. Training the convolutional neural network: using the training samples to train the parameters of the convolutional neural network using a back propagation algorithm to obtain the trained neural network model.

10. A contact lens defect detection system using a contact lens defect detection method according to any one of claims 1 to 9, characterized in that: include: Lens image acquisition system construction module: used to construct a lens image acquisition system for contact lenses, and to set parameters of the lens image acquisition system according to the type of contact lenses and the location of defects; wherein the lens image acquisition system comprises: a bright field illumination light source and a dark field illumination light source which are not on the same optical axis, and a beam splitter for synthesizing the bright field illumination light source and the dark field illumination light source and transmitting them to the surface of the contact lens; Defect detection and identification module: used to input the lens image collected by the lens image acquisition system into a pre-trained neural network model for defect detection and identification.

Images