Method and device for detecting the field of view of a window of a protective clothing

Abstract

The application discloses a kind of detection method and device for the field of view of protective clothing window, it is related to field of view detection technique field, the present application, ideal field of view projection is formed using the light emitted by fiber optic light source, and the head mould simulates the position of human eye pupil, the deviation of pupil and light intersection is monitored in combination with photoelectric encoder, dynamic calibration pupil position, adapt to different pupil distance;Three-dimensional surface model of the window is generated by introducing laser scanner or structured light equipment, and the complete three-dimensional model of the occlusion area is constructed by combining ray tracing technology to correct the light propagation path;In addition, the occlusion area is extracted using a deep learning assisted occlusion recognition algorithm, and it is projected into the surface model, the model parameters are adjusted dynamically to adapt to different pupil distance, so that the occlusion calibration is more accurate;On this basis, the field of view preservation rate and horizontal, vertical viewing angle are calculated, and the detection report containing field of view integrity analysis and occlusion influence evaluation is generated;Significantly enhance the adaptability of detection device to dynamic change.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual field detection, and in particular to a detection method and device for the visual field of a protective clothing window. BACKGROUND

[0002] The visual field range of the window of the protective clothing is particularly important, and the visual field detection of the window can effectively guarantee the safety of the user. The existing detection method simulates the visual field of the human eye by using a head model and a fiber optic light source, analyzes the visual field change before and after wearing the window by using projection technology, and has good effects in the detection of the protective window with a plane and a regular shape. However, as the application scenarios of the protective equipment increase, the design of the protective equipment gradually tends to diversification, and higher requirements are put forward for the detection means.

[0003] In actual use scenarios, the protective window not only needs to meet the basic protective performance, but also needs to provide a wider visual field coverage in a high-risk environment. This requirement has given birth to more protective equipment with arc surfaces or irregular designs, which reduces edge obstruction and improves wearing comfort by optimizing the shape. However, the window with such a structure has some problems in the detection process. The assumption of the light propagation of the traditional detection method is usually based on a plane or a regular shape, which will cause projection distortion or misjudgment of the blocked area when applied to the arc surface design. In addition, the optical characteristics of the multi-arc surface structure, such as refraction or reflection caused by the curvature, will also interfere with the image collection of the CCD camera and affect the accuracy of the detection result.

[0004] Although some traditional methods have certain adjustment measures to deal with the above problems, such as parameter calibration or background light difference technology to improve the detection adaptability, these improvements are still limited to correcting a single error source. Therefore, on the basis of the existing detection system, how to optimize the method to adapt to more complex protective window designs has become a problem to be solved. SUMMARY

[0005] In view of the above existing problems, the present application is proposed.

[0006] The present application provides a detection method and device for the visual field of a protective clothing window to solve the problem that the assumption of the light propagation of the traditional detection method is usually based on a plane or a regular shape, which will cause projection distortion or misjudgment of the blocked area when applied to the arc surface design. In addition, the optical characteristics of the multi-arc surface structure, such as refraction or reflection caused by the curvature, will also interfere with the image collection of the CCD camera and affect the accuracy of the detection result.

[0007] To solve the above technical problems, the present application provides the following technical scheme:

[0008] In a first aspect, the present application embodiment provides a detection method for the visual field of a protective clothing window, which comprises,

[0009] Step S1, set up a detection device in a darkroom, including a head model, a fiber optic light source, a light guide screen, a CCD camera and a computer processing module, the light emitted by the fiber optic light source forms an initial visual field area image without wearing a protective window on the light guide screen;

[0010] Step S2, wear a multi-arc surface protective window on the head model, observe the projection change of the light fiber light source on the light guide screen after passing through the window, and capture the projection image by the CCD camera;

[0011] Step S3, use a laser scanner or a structured light device to perform three-dimensional geometric modeling on the multi-arc surface window, generate a window curved surface model, and superimpose the window curved surface model and the shielding area image collected by the CCD to construct a complete shielding area model,

[0012] Step S4, use a computer image processing module, use a deep learning assisted shielding recognition algorithm to extract the shielding area, project it into the shielding area model, calibrate the range and position of the shielding, and quantify the shielding degree;

[0013] The shielding recognition algorithm adjusts the shielding recognition parameters for different interpupillary distances;

[0014] Step S5, according to the calibrated shielding area, calculate the field of view preservation rate and the horizontal and vertical viewing angles, generate a detection report, the detection report includes field of view integrity analysis and shielding influence evaluation; and provide specific data comparison for different interpupillary distances.

[0015] As a preferred scheme of the protective clothing window field of view detection method, the head model is used to simulate the position of the pupil of the human eye, and the simulated pupil position on the head model is adjustable to adapt to different interpupillary distances.

[0016] The fiber optic light source is used to emit light, and the light forms a visual field area projection image on the light guide screen, which is an ideal visual field without wearing a window.

[0017] As a preferred scheme of the protective clothing window field of view detection method, in step S2, the positions of the head model and the light source are dynamically calibrated, and the light always uniformly covers the arc surface of the window.

[0018] Meanwhile, the relative position change of the head model pupil and the light intersection point is monitored through the photoelectric encoder, and the simulated pupil position is automatically adjusted.

[0019] As a preferred scheme of the protective clothing window field of view detection method, the step of monitoring the relative position change of the head model pupil and the light intersection point through the photoelectric encoder and automatically adjusting the simulated pupil position is,

[0020] Define the light path of the light emitted by the fiber optic light source as a straight line, and the position of the head model pupil is defined as coordinates (xh ,y h ), the projection point of the light ray on the light guide screen is (x p ,y p ), the deviation of the intersection point of the pupil and the light ray is monitored by the photoelectric encoder, and the deviation calculation formula is:

[0021] Δ x =x h -x p , Δ y =y h -y p ,

[0022] Based on the deviation value (Δ x ,Δ y ), the pupil position of the head model is updated, and the update formula is:

[0023]

[0024] Wherein, x h is the horizontal position coordinate of the current head model pupil, y h is the vertical position coordinate of the current head model pupil, is the adjusted horizontal position coordinate of the head model pupil, is the adjusted vertical position coordinate of the head model pupil, k x ,k y respectively represent the step adjustment coefficients in horizontal and vertical directions, used to control the amplitude of each position update,

[0025] Δ x ,Δ y is the deviation value of the intersection point of the pupil and the light ray;

[0026] The position change rate of each time step t is recorded by the photoelectric encoder, and the rate calculation formula is:

[0027]

[0028] If the following conditions are met, it indicates that dynamic deviation is detected:

[0029] |v x |>∈ x or |v y |>∈ y ,

[0030] Wherein, v x is the horizontal pupil movement rate, v y is the vertical pupil movement rate, Δt is the time step, used to calculate the rate, ∈ x ,∈ y is the rate threshold in horizontal and vertical directions,

[0031] When the above conditions are met, the automatic calibration program is triggered, the head mold position is corrected, and the light uniformly covers the curved surface of the view window.

[0032] As a preferred scheme of the detection method for the view field of the window of the protective clothing, in step S3, the light propagation path on the curved surface is simulated by combining the light ray tracing technology, and the deviation caused by refraction or reflection is corrected.

[0033] In step S3, the angle of the head mold is adjusted by the mechanical arm to simulate the change of the view field in different wearing postures.

[0034] As a preferred scheme of the detection method for the view field of the window of the protective clothing, in step S3, the light propagation path on the curved surface is simulated by combining the light ray tracing technology, and the deviation caused by refraction or reflection is corrected.

[0035] The surface data of the multi-arc surface window is collected by the laser scanner to obtain a point cloud set, which is set as P:

[0036] P={(x i ,y i ,z i )∣i=1,2,…,n},

[0037] Wherein, x i ,y i ,z i is the three-dimensional coordinates of the i-th sampling point in the point cloud, and n is the total number of sampling points in the point cloud.

[0038] The surface fitting is performed on the point cloud data by using a quadratic surface equation to obtain a window surface model, and the model formula is:

[0039] z=ax 2 +by 2 +cxy+dx+ey+f,

[0040] Wherein, z is the height coordinate of any point on the window surface, a, b, c, d, e, and f are undetermined coefficients of the fitting equation, and are solved based on the least square method.

[0041] The error function E is defined as:

[0042]

[0043] Solve:

[0044] To obtain the optimal coefficients;

[0045] The occlusion area image M(x,y) collected by the CCD camera is mapped to the surface model is expressed as:

[0046]

[0047] According to Snell's law, the light propagation path is corrected, assuming that the light incidence angle is θ i , the refraction angle is θ t , and the direction vector of the refraction path is calculated as:

[0048]

[0049] wherein, is the incident direction vector, is the surface normal vector, and n1 and n2 are the refractive indices of air and the window material, respectively.

[0050] As a preferred scheme of the protective clothing window field of view detection method, the angle adjustment mode of the mechanical arm is:

[0051] Assuming that the head model angle controlled by the mechanical arm is α, β, and γ, the rotation matrix is:

[0052]

[0053] wherein, α is the rotation angle of the head model around the x-axis, β is the rotation angle of the head model around the y-axis, and γ is the rotation angle of the head model around the z-axis.

[0054] As a preferred scheme of the protective clothing window field of view detection method, the step of extracting the occlusion area by using the deep learning assisted occlusion recognition algorithm, projecting it into the occlusion area model, calibrating the range and position of the occlusion, and quantifying the occlusion degree is:

[0055] The input image is preprocessed, the original image I(x, y) collected by the CCD camera is converted into a gray image I g (x, y), and Gaussian filtering is applied for denoising, and the conversion denoising formula is:

[0056]

[0057] wherein, I(x, y) is the original image, each pixel position is represented by x, y, I g (x, y) is the denoised image after gray scaling, σ is the standard deviation of the Gaussian filter, exp is the exponential function, indicating the kernel calculation in the Gaussian function, and * is the convolution operator;

[0058] The convolutional neural network CNN is used to extract the image features φ(I g ), and the output is the occlusion area mask M(x, y), and the extraction formula is:

[0059] M(x, y) = σ(Wφ(I g )+b),

[0060] where M(x, y) is a binary mask image of the occlusion region, with value 1 indicating occlusion and value 0 indicating no occlusion, φ(I g ) is the feature representation extracted by the deep learning model for the preprocessed image, W is the weight matrix of the neural network for linear transformation, b is the bias term, and σ is the activation function;

[0061] The occlusion mask M(x, y) is mapped to a three-dimensional window model to obtain a three-dimensional representation of the occlusion region, and the mapping formula is:

[0062] M occlusion = {(x, y, z) | M(x, y) = 1 and z = ax 2 + by 2 + cxy + dx + ey + f},

[0063] where M occlusion is the mapped three-dimensional occlusion region model, z is the height of the three-dimensional surface model, and a, b, c, d, e, f are the coefficients of the quadratic surface fitting equation;

[0064] The deviation between the occlusion region and the actual region is calculated, and the calculation formula is:

[0065]

[0066] By minimizing the deviation, the range and position of the occlusion region are optimized, where (x i ,y i ) detected is the i-th occlusion region point coordinate detected by deep learning, (x i ,y i ) actual is the true coordinate of the i-th point in the actual three-dimensional model, n is the total number of points in the occlusion region,

[0067] Δ x,y is the error value of the occlusion region detection;

[0068] The degree of occlusion is quantified, and the degree of occlusion is defined as the proportion of occluded pixels to the total number of pixels, and the definition formula is:

[0069]

[0070] where η is the degree of occlusion, which is used to quantitatively describe the influence of occlusion on the field of view, and ∑ x, yM(x, y) is the number of pixels in the occlusion region, and ∑ x,y 1 is the total number of pixels;

[0071] Adjusting the parameter for different pupil distances, adjusting the convolution kernel size of the occlusion recognition model as k according to the input pupil distance d d , the adjustment formula is:

[0072]

[0073] Wherein, d is the current pupil distance, d0 is the standard pupil distance, k d is the convolution kernel size adjusted according to the current pupil distance, k0 is the initial convolution kernel size under the standard pupil distance.

[0074] As a preferred scheme of the detection method of the window field of view of the protective clothing, wherein: the step of calculating the field of view preservation rate and the horizontal and vertical viewing angles according to the calibrated occlusion area, generating a detection report, and the detection report including field of view integrity analysis and occlusion influence evaluation; and providing specific data comparison for different pupil distances,

[0075] The horizontal and vertical viewing angles are defined, and the definition formula is:

[0076]

[0077] Wherein, θ h is the horizontal viewing angle, θ v is the vertical viewing angle, w is the window width, h is the window height, and d is the distance from the pupil to the window.

[0078] The field of view preservation rate is calculated, which is defined as the ratio of the area of the unoccluded area to the total window area, and the calculation formula is:

[0079]

[0080] Wherein, ρ is the field of view preservation rate, A unobstructed is the unoccluded area, and A total is the total window area.

[0081] The field of view preservation rates of different pupil distances d i are compared, and the influence is quantified, and the difference calculation formula is:

[0082] Δρ=ρ(d max )-ρ(d min ),

[0083] Wherein, Δρ is the difference of the field of view preservation rate caused by the change of the pupil distance, d max , d min and the minimum pupil distance.

[0084] Finally, a detection report is generated, and the detection report includes:

[0085] Field of view integrity analysis: through ρ, θ h , θv numerical evaluation window,

[0086] Shading impact evaluation: quantitative analysis of shading changes under different pupil distances.

[0087] In a second aspect, the present application provides a detection device for the visual field of a protective eyewear window, comprising a head model, a fiber optic light source, a light guide screen, a CCD camera and a computer processing module,

[0088] The head model is used to simulate the position of the pupil of the human eye and the range of the visual field, and the position of the pupil on the head model can be adjusted.

[0089] The fiber optic light source is used to emit light to form an initial visual field area projection image, and the light from the fiber optic light source is projected onto the light guide screen through the simulated pupil of the head model eye, providing a reference for the ideal visual field when the window is not worn. The light intensity and angle of the fiber optic light source can be adjusted to uniformly cover the multi-arc surface structure.

[0090] The light guide screen is used to receive the light projection from the fiber optic light source after passing through the window to form a visual field area image.

[0091] The CCD camera is used to capture the projection image on the light guide screen, record the visual field changes before and after the protective window in real time, and transmit the image to the computer processing module.

[0092] The computer processing module is used to receive and process the image data collected by the CCD camera, extract the shading area, calibrate the range and position, and quantify the shading degree through the built-in image processing algorithm. The module combines deep learning algorithm and ray tracing technology to simulate the actual use environment of the window, calculate the visual field preservation rate, horizontal and vertical angles, and generate a detection report.

[0093] The present application has the following advantages: the detection device comprising a head model, a fiber optic light source, a light guide screen, a CCD camera and a computer processing module is built in a dark room, the ideal visual field projection is formed by the light emitted by the fiber optic light source, and the position of the pupil of the human eye is simulated by the head model. The deviation of the intersection of the pupil and the light is monitored by the photoelectric encoder, the position of the pupil is dynamically calibrated, and different pupil distances are adapted. A three-dimensional curved surface model of the window is generated by introducing a laser scanner or a structured light device, and the light propagation path is corrected by combining the ray tracing technology to construct a complete three-dimensional model of the shading area. In addition, a shading recognition algorithm assisted by deep learning is used to extract the shading area and project it into the curved surface model. By dynamically adjusting the model parameters to adapt to different pupil distances, the shading calibration is more accurate. On this basis, the visual field preservation rate and the horizontal and vertical angles are calculated, a detection report containing visual field integrity analysis and shading impact evaluation is generated, and comparative analysis is performed on different pupil distances. The detection device significantly enhances the adaptability to dynamic changes and the processing accuracy of complex multi-arc surface structures. BRIEF DESCRIPTION OF DRAWINGS

[0094] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort based on these drawings.

[0095] Figure 1 The flowchart of the method for detecting the field of view of the window of the protective clothing.

[0096] Figure 2 The frame diagram of the device for detecting the field of view of the window of the protective clothing. DETAILED DESCRIPTION

[0097] In order to make the above-mentioned objects, features and advantages of the present application more apparent and understandable, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0098] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0099] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment that excludes other embodiments.

[0100] Embodiment 1, reference Figure 1 and Figure 2 The embodiment provides a method for detecting the field of view of the window of the protective clothing, comprising the following steps:

[0101] Step S1, a detection device is built in a darkroom, comprising a head model, a fiber optic light source, a light guide screen, a CCD camera and a computer processing module, the light emitted by the fiber optic light source forms an initial image of the field of view area without wearing the protective window on the light guide screen;

[0102] The head model is used for simulating the position of the pupil of the human eye, and the simulated pupil position on the head model is adjustable to adapt to different interpupillary distances;

[0103] The fiber optic light source is used for emitting light to form a projection image of the field of view area on the light guide screen, which is the ideal field of view without wearing the window;

[0104] Step S2, wearing the multi-arc surface protective window on the head model, observing the projection change of the optical fiber light source on the light guide screen after passing through the window, and capturing the projection image by the CCD camera;

[0105] In step S2, the positions of the head model and the light source are dynamically calibrated, and the light rays always uniformly cover the arc surface of the window;

[0106] Meanwhile, the relative position change of the head model pupil and the light intersection point is monitored by the photoelectric encoder, and the simulated pupil position is automatically adjusted;

[0107] The step of monitoring the relative position change of the head model pupil and the light intersection point by the photoelectric encoder and automatically adjusting the simulated pupil position is,

[0108] The path of the light emitted by the optical fiber light source is defined as a straight line, and the position of the head model pupil is defined as coordinates (x h ,y h ), and the projection point of the light and the light guide screen is (x p ,y p ), the deviation of the pupil and the light intersection point is monitored by the photoelectric encoder, and the deviation calculation formula is:

[0109] Δ x =x h -x p , Δ y =y h -y p ,

[0110] Based on the deviation value (Δ x ,Δ y ), the position of the head model pupil is updated, and the update formula is:

[0111]

[0112] Wherein, x h is the horizontal position coordinate of the current head model pupil, y h is the vertical position coordinate of the current head model pupil, is the adjusted horizontal position coordinate of the head model pupil, is the adjusted vertical position coordinate of the head model pupil, k x ,k y respectively represent the step adjustment coefficients in horizontal and vertical directions, which are used to control the amplitude of each position update,

[0113] Δ x ,Δ y is the deviation value of the pupil and the light intersection point;

[0114] The position change rate of each time step t is recorded by the photoelectric encoder, and the rate calculation formula is:

[0115]

[0116] If the following condition is met, it indicates that a dynamic deviation is detected:

[0117] |v x |>∈ x Or |v y |>∈ y ,

[0118] Where v x is the horizontal pupil movement rate, v y is the vertical pupil movement rate, Δt is the time step, for calculating the rate, ∈ x , ∈ y is the rate threshold in the horizontal and vertical directions,

[0119] When the above conditions are met, the automatic calibration procedure is triggered, the head module position is corrected, and the light rays uniformly cover the arc surface of the viewing window;

[0120] In the field of view detection of multi-arc protective viewing windows, interpupillary distance is a key influencing parameter; different interpupillary distances will cause the angle of light rays contacting the viewing window curve to change after entering the viewing window, thereby affecting the refraction and reflection path of the light rays, interpupillary distance affects the position of the human eye in the horizontal field of view, and different interpupillary distances will change the size of the field of view shielding area, interpupillary distance is directly related to the overlapping part of the binocular common field of view, and the size thereof will affect the field of view preservation rate; therefore, the present application introduces the interpupillary distance parameter to further improve the detection accuracy.

[0121] Specifically, the change of the head module pupil position is monitored in real time by the photoelectric encoder, and the position is corrected based on the dynamic deviation adjustment rule, so that the light rays always uniformly cover the multi-arc surface structure during the test process, and the adjustment process is more accurate and efficient by combining dynamic deviation and speed monitoring.

[0122] Step S3, using a laser scanner or a structured light device to perform three-dimensional geometric modeling on the multi-arc viewing window, generating a viewing window curve model, and superimposing the viewing window curve model and the shielding area image collected by the CCD to construct a complete shielding area model,

[0123] In step S3, the light propagation path on the curve is simulated by combining the ray tracing technology, and the deviation caused by refraction or reflection is corrected;

[0124] In step S3, the angle of the head module is adjusted by a mechanical arm to simulate the change of the field of view under different wearing postures;

[0125] The step of using a laser scanner or a structured light device to perform three-dimensional geometric modeling on the multi-arc viewing window, generating a viewing window curve model, and superimposing the viewing window curve model and the shielding area image collected by the CCD to construct a complete shielding area model is,

[0126] Collect the surface data of multi-arc window through laser scanner, obtain point cloud set, set as P:

[0127] P={(x i ,y i ,z i )∣i=1,2,…,n},

[0128] Wherein, x i ,y i ,z i is the three-dimensional coordinates of the i-th sampling point in the point cloud, n is the total number of sampling points in the point cloud,

[0129] The surface fitting is carried out on the point cloud data by using quadratic surface equation, and the window surface model is obtained, and the model formula is:

[0130] z=ax 2 +by 2 +cxy+dx+ey+f,

[0131] Wherein, z is the height coordinate of any point on the window surface, a, b, c, d, e, f are the undetermined coefficients of the fitting equation, and are solved based on the least square method,

[0132] Define error function E,

[0133]

[0134] Solve:

[0135] Get the optimal coefficient;

[0136] The occluded area image M(x,y) collected by the CCD camera is mapped to the surface model It is expressed as:

[0137]

[0138] According to Snell's law, the light propagation path is corrected, assuming that the light incidence angle is θ i , the refraction angle is θ t , and the direction vector calculation formula of the refraction path is:

[0139]

[0140] Wherein, is the incident direction vector, is the surface normal vector, n1 and n2 are the refractive indexes of air and window material respectively;

[0141] The angle adjustment mode of the mechanical arm is:

[0142] Suppose the head mold angle controlled by the mechanical arm is α, β, γ, and the rotation matrix thereof is:

[0143]

[0144]

[0145] Wherein, α is the rotation angle of the head mold around the x axis, β is the rotation angle of the head mold around the y axis, and γ is the rotation angle of the head mold around the z axis.

[0146] Specifically, by laser scanning and surface fitting, a three-dimensional window model is constructed, and a ray tracing method is used to correct refraction deviation to accurately locate the occluded area.

[0147] Step S4, using a computer image processing module, an occlusion recognition algorithm assisted by deep learning is used to extract the occluded area, project it into the occluded area model, calibrate the range and position of the occlusion, and quantify the occlusion degree;

[0148] The occlusion recognition algorithm adjusts the occlusion recognition parameters for different interpupillary distances;

[0149] The step of using the occlusion recognition algorithm assisted by deep learning to extract the occluded area, project it into the occluded area model, calibrate the range and position of the occlusion, and quantify the occlusion degree is,

[0150] The input image is preprocessed, and the original image I(x, y) collected by the CCD camera is converted into a gray image I g (x, y), and Gaussian filtering is applied for denoising, and the denoising formula is converted as:

[0151]

[0152] Wherein, I(x, y) is the original image, each pixel position is represented by x, y, I g (x, y) is the denoised image after gray scale, σ is the standard deviation of the Gaussian filter, exp is the exponential function, indicating the kernel calculation in the Gaussian function, and * is the convolution operator;

[0153] The convolutional neural network CNN is used to extract the image features φ(I g ), and the occlusion area mask M(x, y) is output, and the extraction formula is:

[0154] M(x, y) = σ(W·φ(I g )+b),

[0155] Wherein, M(x, y) is a binary mask image of the occluded area, a value of 1 indicates occlusion, and a value of 0 indicates no occlusion, φ(I g) represents the feature representation extracted from the preprocessed image by the deep learning model, W is the weight matrix of the neural network used for linear transformation, b is the bias term, and σ is the activation function;

[0156] Mapping the occlusion mask M(x,y) to the 3D viewport model yields a 3D representation of the occlusion region. The mapping formula is as follows:

[0157] M occlusion ={(x,y,z)∣M(x,y)=1 and z=ax 2 +by 2 +cxy+dx+ey+f},

[0158] Among them, M occlusion Let z be the height of the 3D occlusion region model after mapping, and a, b, c, d, e, f be the coefficients of the quadratic surface fitting equation.

[0159] The deviation between the occluded area and the actual area is calculated using the following formula:

[0160]

[0161] By minimizing the deviation, the extent and location of the occlusion area are optimized, where (x i ,y i ) detected Let x be the coordinates of the i-th occlusion region point detected by deep learning, (x i ,y i ) actual Let be the true coordinates of the i-th point in the actual 3D model, and n be the total number of points in the occluded region.

[0162] Δ x,y This represents the error value for detecting the occluded area;

[0163] To quantify the degree of occlusion, the degree of occlusion is defined as the proportion of occluded pixels to the total number of pixels, and the formula is defined as follows:

[0164]

[0165] Where η represents the degree of occlusion, used to quantitatively describe the impact of occlusion on the field of view, ∑ x,y M(x,y) is the number of pixels in the occluded region, ∑ x,y 1 represents the total number of pixels;

[0166] Parameters are adjusted for different interpupillary distances. Based on the input interpupillary distance d, the convolution kernel size of the occlusion recognition model is adjusted to k. d The formula is adjusted as follows:

[0167]

[0168] wherein d is the current pupil distance, d0 is the standard pupil distance, k d is the adjusted kernel size according to the current pupil distance, and k0 is the initial kernel size under the standard pupil distance.

[0169] Specifically, step S4 extracts the occlusion area by deep learning technology, and projects and calibrates the occlusion area in combination with the three-dimensional curved surface model, so that the model adapts to the complex curved surface.

[0170] Step S5, according to the calibrated occlusion area, calculates the field of view preservation rate and horizontal and vertical angles of view, generates a detection report, and the detection report includes field of view integrity analysis and occlusion impact assessment; and provides specific data comparison for different pupil distances;

[0171] The step of calculating the field of view preservation rate and horizontal and vertical angles of view according to the calibrated occlusion area, generating a detection report, and the detection report including field of view integrity analysis and occlusion impact assessment; and providing specific data comparison for different pupil distances is,

[0172] The horizontal and vertical angles of view are defined, and the definition formula is:

[0173]

[0174] wherein θ h is the horizontal angle of view, θ v is the vertical angle of view, w is the window width, h is the window height, and d is the distance from the pupil to the window.

[0175] The field of view preservation rate is calculated, which is defined as the ratio of the area of the unoccluded area to the total window area, and the calculation formula is:

[0176]

[0177] wherein ρ is the field of view preservation rate, A unobstructed is the unoccluded area, and A total is the total window area.

[0178] The field of view preservation rates of different pupil distances d i are compared, and the influence is quantified, and the difference calculation formula is:

[0179] Δρ=ρ(d max )-ρ(d min ),

[0180] wherein Δρ is the difference in field of view preservation rate caused by the change in pupil distance, d max , d min , and the minimum pupil distance.

[0181] Finally, a detection report is generated, and the detection report includes:

[0182] Field of view integrity analysis: through the values of p, q h , q v , the field of view performance of the window is evaluated,

[0183] Shading impact evaluation: quantitative analysis is performed on the changes in shading under different pupil distances.

[0184] Specifically, step S5 comprehensively analyzes the impact of the calibrated shading area on the field of view preservation rate and the viewing angle.

[0185] The embodiment also provides a detection device for the field of view of a protective clothing window, which comprises a head model, a fiber optic light source, a light guide screen, a CCD camera, and a computer processing module,

[0186] The head model is used to simulate the position of the pupil of the human eye and the range of the field of view, and the position of the pupil on the head model is adjustable.

[0187] The fiber optic light source is used to emit light to form an initial projection image of the field of view area, and the light of the fiber optic light source is projected onto the light guide screen through the simulated pupil of the eye part of the head model to provide a reference for the ideal field of view when the window is not worn. The light intensity and angle of the fiber optic light source are adjustable and uniformly cover the multi-arc surface structure.

[0188] The light guide screen is used to receive the light projection of the fiber optic light source after passing through the window to form a field of view area image.

[0189] The CCD camera is used to capture the projection image on the light guide screen, record the changes in the field of view before and after the protective window in real time, and transmit the image to the computer processing module.

[0190] The computer processing module is used to receive and process the image data collected by the CCD camera, extract the shading area, calibrate the range and position, and quantify the shading degree through the built-in image processing algorithm. The module combines the deep learning algorithm and the ray tracing technology, simulates the actual use environment of the window, calculates the field of view preservation rate, the horizontal viewing angle and the vertical viewing angle, and generates a detection report.

[0191] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A method for detecting the field of view of a protective suit's viewing window, characterized in that: Comprising, Step S1, set up the detection device in the darkroom, including head model, optical fiber light source, light guide screen, CCD camera and computer processing module, the light emitted by the optical fiber light source forms the initial visual field area image of the non-wearing protective window on the light guide screen; Step S2, wear the multi-arc surface protective window on the head model, observe the projection change of the optical fiber light source on the light guide screen after passing through the window, and capture the projection image by the CCD camera; Step S3, use a laser scanner or a structured light device to perform three-dimensional geometric modeling on the multi-arc surface window, generate a window surface model, and superimpose the window surface model and the occlusion area image collected by the CCD to build a complete occlusion area model; In step S3, the multi-arc surface window surface data is collected by a laser scanner or a structured light device, a point cloud set is obtained, and is set as : ， in, For the first point cloud The three-dimensional coordinates of each sampling point This represents the total number of sampling points in the point cloud. Use the quadric surface equation to perform surface fitting on the point cloud data to obtain the window surface model, and the model formula is: ， wherein, is the height coordinate of an arbitrary point of the window surface, is the undetermined coefficient of the fitting equation, solved based on the least square method; In step S3, the occlusion region image captured by the CCD camera mapped to the curved surface model is expressed as: ， According to Snell's law, the light propagation path is corrected, assuming that the light incidence angle is , the refraction angle is , and the direction vector calculation formula of the refraction path is: ， ， ， wherein is the incident direction vector, is the surface normal vector, are the refractive indices of air and the window material, respectively; Step S4, use the computer image processing module to extract the occlusion area by using the deep learning assisted occlusion recognition algorithm, project it into the occlusion area model, calibrate the range and position of the occlusion, and quantify the occlusion degree; The occlusion recognition algorithm adjusts the occlusion recognition parameters for different interpupillary distances; In step S4, the input image is preprocessed, and the original image collected by the CCD camera is converted into a gray-scale image and denoised by applying a Gaussian filter to obtain a denoised gray-scale image converted into a gray-scale image, and a Gaussian filter is applied for denoising to obtain a denoised gray-scale image ; Extracting image features using convolutional neural networks, CNNs , outputting an occlusion region mask , the extraction formula is: ， wherein, is a binary mask image of the occlusion region, with value denotes occlusion, with value denotes non-occlusion, is a feature representation extracted by a deep learning model from the preprocessed image, is a weight matrix of the neural network, used for linear transformation, is a bias term, is an activation function; obtaining an occlusion mask to a three-dimensional window model to obtain a three-dimensional representation of the occlusion region, and a mapping formula is: ， wherein, is the mapped three-dimensional occlusion region model, is the height of the three-dimensional surface model, is the coefficient of the quadratic surface fitting equation; Calculate the deviation of the occlusion area and the actual area, and the calculation formula is: ， By minimizing the deviation, the extent and location of the occlusion area are optimized, where... For the first to pass the deep learning detection Coordinates of a point in the occluded area. For the actual 3D model The actual coordinates of each point This represents the total number of points in the occluded area. This represents the error value for detecting the occluded area; Quantify the occlusion degree, define the occlusion degree as the proportion of occluded pixels to total pixels, and the definition formula is: ， wherein, is the degree of occlusion for quantitatively describing the influence of occlusion on the field of view, is the number of pixels of the occlusion area, is the total number of pixels; Step S5, according to the calibrated occlusion area, calculate the field of view preservation rate and the horizontal and vertical visual angles, generate a detection report, the detection report includes field of view integrity analysis and occlusion impact evaluation; and provide specific data comparison for different interpupillary distances.

2. The method of claim 1, wherein: The head model is used to simulate the position of the human eye pupil, and the simulated pupil position on the head model is adjustable to adapt to different interpupillary distances; The optical fiber light source is used to emit light, and the light forms a visual field area projection image on the light guide screen, which is an ideal visual field without wearing a window.

3. The method for detecting the field of view of a protective suit window as described in claim 2, characterized in that: In step S2, dynamically calibrate the positions of the head model and the light source, and the light always uniformly covers the arc surface of the window; Meanwhile, the relative position change of the head model pupil and the light intersection point is monitored by the photoelectric encoder, and the simulated pupil position is automatically adjusted.

4. The method of claim 3, wherein: The step of monitoring the relative position change of the head model pupil and the light intersection point by the photoelectric encoder and automatically adjusting the simulated pupil position is, The path of the light emitted by the fiber-optic light source is defined as a straight line, and the position of the head mold pupil is defined as coordinate The projection point of the light on the light guide screen is The deviation of the intersection of the pupil and the light is monitored by the photoelectric encoder, and the deviation calculation formula is: ， ， Based on the deviation value The pupil position of the head module is updated, and the update formula is: ， ， wherein, is the horizontal position coordinate of the current head model pupil, is the vertical position coordinate of the current head model pupil, is the adjusted horizontal position coordinate of the head model pupil, is the adjusted vertical position coordinate of the head model pupil, respectively represent the step adjustment coefficients in horizontal and vertical directions, for controlling the magnitude of each position update, is the deviation value of the pupil intersection with the light rays; The position change rate at each time step is recorded by the optical encoder The rate is calculated by the formula: ， ， If the following conditions are met, it indicates that dynamic deviation is detected: or , wherein, is the pupil movement rate in the horizontal direction, is the pupil movement rate in the vertical direction, is the time step for calculating the rate, is the rate threshold in the horizontal and vertical directions, When the above conditions are met, the automatic calibration program is triggered, the head model position is corrected, and the light uniformly covers the arc surface of the window.

5. The protective clothing window visual field detection method of claim 4, wherein: In step S3, a mechanical arm is used to adjust the angle of the head model to simulate the change of the visual field under different wearing postures.

6. The method for detecting the field of view of a protective suit window as described in claim 5, characterized in that, In step S3, the error function is defined as : ， Solve: , to obtain the optimal coefficients.

7. The method for detecting the field of view of a protective suit window as described in claim 6, characterized in that: The angle adjustment mode of the mechanical arm is: Assume the head module angle controlled by the robot arm is Its rotation matrix is: ， wherein is the rotation angle of the head mold around the axis, is the rotation angle of the head mold around the axis, is the rotation angle of the head mold around the axis.

8. The method for detecting the field of view of a protective suit window as described in claim 7, characterized in that, In step S4, the denoising formula is converted to: ， wherein, is the original image, each pixel position is represented by , is the denoised image after graying, is the standard deviation of the Gaussian filter, is the exponential function, representing the kernel calculation in the Gaussian function, is the convolution operator; In step S4, the parameter is adjusted for different pupil distances, and the pupil distance is input The convolution kernel size of the occlusion identification model is adjusted as The adjustment formula is: ， wherein, is the current interpupillary distance, is the standard interpupillary distance, is the kernel size adjusted according to the current interpupillary distance, is the initial kernel size at the standard interpupillary distance.

9. The method of claim 8, wherein: The step of calculating the field of view preservation rate and the horizontal and vertical visual angles according to the calibrated occlusion area, generating a detection report, and providing specific data comparison for different interpupillary distances is, Define the horizontal and vertical visual angles, and the definition formula is: ， wherein, is the horizontal viewing angle, is the vertical viewing angle, is the window width, is the window height, is the pupil to window distance; Calculate the field of view preservation rate, which is defined as the ratio of the area of the non-occluded area to the total window area, and the calculation formula is: ， wherein, is the field of view preservation rate, is the unobstructed area, is the total area of the window; Comparison of field preservation for different pupil distances The impact was quantified and the difference was calculated as follows: ， wherein the difference in field of view preservation for pupil distance changes, and the minimum pupil distance; Finally, a detection report is generated, and the detection report includes: Field of integrity analysis: The field performance of the evaluation window is assessed by the numerical values of Occlusion impact evaluation: quantitative analysis of occlusion changes under different interpupillary distance conditions.

10. A device for detecting the field of view of a window of a protective garment, based on the method for detecting the field of view of a window of a protective garment according to any one of claims 1 to 9, characterized in that, Comprising, Head model, fiber-optic light source, light guide screen, CCD camera and computer processing module, Head model, used to simulate the position of the pupil and the range of the field of view, the position of the pupil on the head model can be adjusted; Fiber-optic light source, used to emit light and form the initial projection image of the field of view, the light of the fiber-optic light source is projected onto the light guide screen through the simulated pupil of the head model eye, providing a reference for the ideal field of view when the protective window is not worn, the light intensity and angle of the fiber-optic light source can be adjusted to uniformly cover the multi-arc surface structure; Light guide screen, used to receive the light projection of the fiber-optic light source after passing through the protective window, forming the image of the field of view; CCD camera, used to capture the projection image on the light guide screen, record the changes of the field of view before and after the protective window in real time, and transmit the image to the computer processing module; Computer processing module, used to receive and process the image data collected by the CCD camera, extract the blocked area, calibrate the range and position, and quantify the degree of obstruction through the built-in image processing algorithm, this module combines deep learning algorithm and ray tracing technology to simulate the actual use environment of the window, calculate the field of view preservation rate, horizontal and vertical viewing angles, and generate a detection report.

Images