LED light source energy density distribution measurement method, device, equipment and storage medium

Abstract

The present invention provides a method, device, equipment and storage medium for measuring the energy density distribution of an LED light source. The method comprises: performing front image acquisition and preprocessing on the light-emitting surface of the LED light source to obtain light-emitting characteristic data. According to these data, a CMOS image sensor array is arranged, and a spherical scanning measurement of the LED light source is performed through the array to obtain spatial light field data. Using a pre-trained deep learning light field reconstruction model, the light field data is compensated for the light-emitting area effect to obtain corrected spatial light field distribution data. Finally, energy density interpolation calculation is performed based on the corrected data to generate a three-dimensional energy density distribution diagram. The method realizes accurate measurement of the energy density distribution of the LED light source by combining an image sensor array, deep learning reconstruction technology and a three-dimensional space measurement model, overcomes the limitations of existing methods, improves measurement accuracy, and can fully reflect the three-dimensional energy density distribution of the LED light source.

Description

Technical Field

[0001] The present invention relates to the field of light source measurement, and in particular to a method, device, equipment and storage medium for measuring energy density distribution of an LED light source. Background Art

[0002] LED light sources are widely used in display technology, lighting systems, optical communications and other fields, and their energy density distribution directly affects the lighting effect and application performance of the light source. Traditional methods for measuring the energy density of LED light sources usually rely on simple light intensity tests or spectral analysis. These methods usually only provide energy distribution information at a single angle or plane, and it is difficult to fully reflect the three-dimensional spatial light field characteristics of LED light sources, especially in complex application environments. The measurement methods in the prior art often ignore the luminous area effect of LED light sources and the spatial distribution of the light field, resulting in the inability to accurately evaluate the actual energy density distribution of LED light sources, which in turn affects the design of optical systems and performance optimization. Summary of the invention

[0003] The main purpose of the present invention is to solve the technical problem that the existing measurement methods often ignore the light emitting area effect of the LED light source and the spatial distribution of the light field, resulting in the inability to accurately evaluate the actual energy density distribution of the LED light source;

[0004] A first aspect of the present invention provides a method for measuring energy density distribution of an LED light source, the method comprising:

[0005] Performing front image acquisition and preprocessing on the light-emitting surface of the LED light source to obtain light-emitting characteristic data of the LED light source;

[0006] Arranging a CMOS image sensor array of the LED light source according to the light-emitting characteristic data, and performing spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data;

[0007] Calling a pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, compensating the light field data for luminous area effect, and obtaining corrected spatial light field distribution data;

[0008] According to the spatial light field distribution data, an energy density interpolation calculation is performed on the LED light source, and a three-dimensional energy density distribution diagram of the LED light source is generated according to the calculation result.

[0009] Optionally, in a first implementation of the first aspect of the present invention, the performing front image acquisition and preprocessing on the light-emitting surface of the LED light source to obtain the light-emitting characteristic data of the LED light source includes:

[0010] Capturing multiple frames of high-resolution images of the light-emitting surface of the LED light source to obtain original image data of the light-emitting surface;

[0011] Performing threshold segmentation and edge detection on the original image data of the light-emitting surface to obtain geometric dimensions and boundary contour data of the light-emitting surface;

[0012] Performing grid division on the original image data of the light-emitting surface to obtain micro-region distribution data of the light-emitting surface;

[0013] Performing grayscale value analysis on the light-emitting surface according to the boundary contour data and the micro-region distribution data to obtain brightness distribution data of the light-emitting surface;

[0014] The geometric dimensions, the boundary contour data, the micro-area distribution data and the brightness distribution data are integrated to obtain the light emitting characteristic data of the LED light source.

[0015] Optionally, in a second implementation of the first aspect of the present invention, the CMOS image sensor array for arranging the LED light sources according to the light emission characteristic data includes:

[0016] Calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array;

[0017] Arranging the CMOS image sensors spatially according to the arrangement parameters, arranging a first CMOS image sensor array in a near-field area away from the light-emitting surface of the LED light source, and arranging a second CMOS image sensor array in a far-field area away from the light-emitting surface of the LED light source;

[0018] Calibrate pixel sensitivities of the first CMOS image sensor array and the second CMOS image sensor array to obtain sensor response calibration data;

[0019] The sampling angles and sampling intervals of the first CMOS image sensor array and the second CMOS image sensor array are calculated according to the light emission characteristic data and the sensor response calibration data, and spherical scanning measurement parameters are obtained to realize the arrangement of the CMOS image sensor array of the LED light source.

[0020] Optionally, in a third implementation of the first aspect of the present invention, calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array includes:

[0021] Normalizing the brightness distribution data in the light-emitting characteristic data, and calculating the relative brightness distribution ratio on the light-emitting surface according to the normalized brightness distribution data;

[0022] Performing a preliminary spatial range estimation based on the relative brightness distribution ratio and the geometric dimensions of the light-emitting surface of the LED light source to obtain a spatial measurement range of spherical scanning;

[0023] Dividing the spatial measurement range into regions to obtain a boundary position between a near-field measurement range and a far-field measurement range;

[0024] Arrangement parameters of the first CMOS image sensor array and the second CMOS image sensor array are determined according to the boundary position.

[0025] Optionally, in a fourth implementation of the first aspect of the present invention, calling a pre-trained deep learning light field reconstruction model according to the luminous feature data and the spatial light field data, compensating the light field data for luminous area effect, and obtaining the corrected spatial light field distribution data includes:

[0026] According to the micro-region distribution data of the light-emitting surface in the light-emitting characteristic data, the spatial light field data is spatially divided to obtain light field sub-data corresponding to each micro-region;

[0027] Performing spatial coordinate mapping on the light field sub-data according to the geometrical dimensions and boundary contour data of the light-emitting surface in the light-emitting characteristic data to obtain micro-region light field mapping data;

[0028] Inputting the micro-area light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the light-emitting area effect;

[0029] The spatial light field data is corrected and calculated according to the compensation coefficient to obtain corrected spatial light field distribution data.

[0030] Optionally, in a fifth implementation of the first aspect of the present invention, the inputting the micro-area light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the light-emitting area effect includes:

[0031] Inputting the micro-region light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model, wherein the deep learning light field reconstruction model includes a convolution layer, a feature fusion layer, a pooling layer, a fully connected layer, and a deconvolution layer;

[0032] Performing a convolution operation on the micro-region light field mapping data through the convolution layer, extracting spatial features through the first convolution layer, and obtaining a primary feature map;

[0033] Performing feature fusion according to the primary feature map and the brightness distribution data through the feature fusion layer, extracting light intensity features through the second convolution layer, and obtaining a light field-brightness feature map;

[0034] Performing a pooling operation and a fully connected layer processing on the light field-brightness feature map through the pooling layer and the fully connected layer to obtain a luminous area effect feature vector;

[0035] The deconvolution layer performs spatial restoration according to the luminous area effect feature vector to obtain a compensation coefficient of the luminous area effect.

[0036] Optionally, in a sixth implementation of the first aspect of the present invention, performing energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generating a three-dimensional energy density distribution diagram of the LED light source according to the calculation result includes:

[0037] Performing spatial transformation on the spatial light field distribution data according to a spherical coordinate system to obtain polar coordinate data with the light emitting center of the LED light source as the origin;

[0038] Performing a segmented interpolation operation on the near field area and the far field area of ​​the LED light source according to the polar coordinate data to obtain a continuous energy density distribution function;

[0039] Performing coordinate transformation from spherical coordinates to rectangular coordinates on the energy density distribution function to obtain an energy density value in three-dimensional space;

[0040] Equal energy density curves of different cross sections are generated according to the energy density values ​​of the three-dimensional space, and a three-dimensional energy density distribution diagram of the LED light source is generated according to the equal energy density curves.

[0041] A second aspect of the present invention provides a device for measuring energy density distribution of an LED light source, the device comprising:

[0042] A preprocessing model, used for collecting and preprocessing the front image of the light-emitting surface of the LED light source to obtain light-emitting characteristic data of the LED light source;

[0043] A scanning module, used for arranging a CMOS image sensor array of the LED light source according to the light-emitting characteristic data, and performing spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data;

[0044] A compensation module, used to call a pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, compensate the light field data for the luminous area effect, and obtain corrected spatial light field distribution data;

[0045] A distribution calculation module is used to perform energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generate a three-dimensional energy density distribution diagram of the LED light source according to the calculation result.

[0046] The third aspect of the present invention provides an LED light source energy density distribution measurement device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the LED light source energy density distribution measurement device performs the steps of the above-mentioned LED light source energy density distribution measurement method.

[0047] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned LED light source energy density distribution measurement method.

[0048] The above-mentioned LED light source energy density distribution measurement method, device, equipment and storage medium obtain luminous characteristic data by collecting and preprocessing the front image of the light-emitting surface of the LED light source. According to these data, a CMOS image sensor array is arranged, and the LED light source is spherically scanned and measured by the array to obtain spatial light field data. Using the pre-trained deep learning light field reconstruction model, the light field data is compensated for the luminous area effect to obtain the corrected spatial light field distribution data. Finally, the energy density interpolation calculation is performed based on the corrected data to generate a three-dimensional energy density distribution map. This method combines the image sensor array, deep learning reconstruction technology and three-dimensional space measurement model to achieve accurate measurement of the energy density distribution of the LED light source, overcome the limitations of the existing methods, improve the measurement accuracy, and can fully reflect the three-dimensional energy density distribution of the LED light source.

[0049] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of a first embodiment of a method for measuring energy density distribution of an LED light source in an embodiment of the present invention;

[0052] Figure 2A schematic diagram of an embodiment of a device for measuring energy density distribution of an LED light source in an embodiment of the present invention;

[0053] Figure 3 Schematic diagram of an embodiment of an LED light source energy density distribution measurement device in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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.

[0055] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0056] To facilitate understanding of this embodiment, a method for measuring the energy density distribution of an LED light source disclosed in an embodiment of the present invention is first introduced in detail. Figure 1 As shown, the method comprises the following steps:

[0057] 101. Capture and preprocess the front image of the light-emitting surface of the LED light source to obtain light-emitting characteristic data of the LED light source;

[0058] In one embodiment of the present invention, the front image acquisition and preprocessing of the light-emitting surface of the LED light source to obtain the light-emitting characteristic data of the LED light source includes: acquiring multiple frames of high-resolution images of the light-emitting surface of the LED light source to obtain original image data of the light-emitting surface; performing threshold segmentation and edge detection on the original image data of the light-emitting surface to obtain geometric dimensions and boundary contour data of the light-emitting surface; performing grid division on the original image data of the light-emitting surface to obtain micro-region distribution data of the light-emitting surface; performing grayscale value analysis on the light-emitting surface according to the boundary contour data and the micro-region distribution data to obtain brightness distribution data of the light-emitting surface; and integrating the geometric dimensions, the boundary contour data, the micro-region distribution data and the brightness distribution data to obtain the light-emitting characteristic data of the LED light source.

[0059] Specifically, first, when capturing multiple frames of high-resolution images of the light-emitting surface of the LED light source, a 4096×3072 pixel CMOS image sensor with a pixel size of 3.45μm×3.45μm is used. The LED light source is fixed on a six-degree-of-freedom precision displacement platform, and the displacement platform is adjusted to keep the light-emitting surface parallel to the CMOS sensor. The acquisition system uses an apochromatic microscope objective with a working distance of 20mm and a numerical aperture of 0.4. When the LED operating current is stable and the junction temperature is controlled at 25±0.1℃, 100 frames of images are collected at each pixel, and the influence of random noise is eliminated by weighted superposition. The weighted method needs to take into account the response characteristics of the sensor and the influence of ambient temperature, and assign a weight coefficient to each frame of the image. ,in is the temperature when the i-th frame is collected, is the calibration temperature, is the temperature characteristic constant. The image data after weighted superposition is normalized to obtain the original image data of the luminous surface.

[0060] Specifically, when performing threshold segmentation and edge detection on the original image data of the light-emitting surface, the first task is to extract the geometric dimensions and boundary contour of the LED light-emitting surface. The specific operation is to establish a two-dimensional coordinate system for the original image data of the light-emitting surface, with the light-emitting center as the origin. The boundary contour extraction uses polar coordinate representation to express the boundary curve of the light-emitting surface as:

[0061] ;

[0062] in is the base radius, and is the Fourier series coefficient, obtained by least squares fitting. Based on the boundary curve function, the geometric characteristic parameters of the luminous surface are calculated: effective luminous area , equivalent diameter , shape factor , (L is the perimeter). These geometric parameters constitute the basic size characteristics of the luminous surface. At the same time, the boundary point sequence is recorded ,Each boundary point has sub-pixel accuracy, and these points are collected to form boundary contour data, which describes the precise boundary shape of the luminous surface.

[0063] Specifically, when meshing the original image data of the luminous surface, the light intensity distribution characteristics of the luminous surface need to be considered. The division adopts a non-uniform grid strategy, and the grid size is inversely proportional to the local light intensity gradient. First, calculate the light intensity gradient field G(x,y) of the entire luminous surface:

[0064] ;

[0065] in Image search point Based on the gradient field, construct a grid size function = ,in and are the minimum and maximum grid sizes, respectively, and k is the adjustment coefficient. According to this function, the luminous surface is adaptively meshed, with dense grids in areas where the light intensity changes dramatically and sparse grids in areas where the light intensity is gentle. Each grid records its position coordinates, size, and local light intensity statistical characteristics to form micro-region distribution data. When performing grayscale value analysis on the luminous surface based on the obtained boundary contour data and micro-region distribution data, it is necessary to establish an accurate light intensity distribution function. The grayscale values ​​in each micro-region are counted to calculate the average light intensity of the region and standard deviation Taking into account the physical characteristics of LED light emission, an improved Gaussian distribution model is used to describe the local light intensity distribution:

[0066] ;

[0067] in is the central light intensity, is the distribution width, is the correction coefficient, and r is the distance to the center of the area. The light intensity data of each micro-area is fitted by the least squares method to obtain the light intensity distribution function of the entire luminous surface. These data constitute the brightness distribution data, which fully describes the spatial distribution characteristics of the light intensity of the luminous surface.

[0068] Finally, when the geometric dimensions, boundary contour data, micro-area distribution data and brightness distribution data are integrated, a multi-layer data structure is used. In the spatial coordinate system, these data are organized according to physical meanings and spatial relationships. First, a reference coordinate system is established, with the luminous center as the origin and the normal direction of the luminous surface as the z-axis. The geometric dimension data is stored as a global parameter, including the effective area, equivalent diameter and shape factor. The boundary contour data is stored in the form of a parametric equation to facilitate subsequent spatial transformation and calculation. The micro-area distribution data is stored in a quadtree structure, and each node contains the location, size and topological relationship information of the region. The brightness distribution data is stored in the form of a piecewise continuous function to ensure the spatial continuity of the data. An index relationship is established between various types of data, and the spatial coordinates are used as the association key to facilitate subsequent light field reconstruction and energy density calculation. The integrated data set is the complete luminous characteristic data of the LED light source, which systematically describes the geometric and optical characteristics of the LED light-emitting surface.

[0069] 102. Arrange a CMOS image sensor array of the LED light source according to the light-emitting characteristic data, and perform spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data;

[0070] In one embodiment of the present invention, when arranging the CMOS image sensor array according to the LED light-emitting characteristic data, it is necessary to determine the arrangement position and orientation of the sensor in combination with the geometric size and brightness distribution characteristics of the light-emitting surface. The first CMOS sensor array is arranged in the near-field area 0.1-10 times the diameter of the light-emitting surface from the LED light-emitting surface, and the second CMOS sensor array is arranged in the far-field area 10-100 times the diameter of the light-emitting surface. Each sensor array is composed of a 4096×3072 pixel CMOS image sensor with a pixel size of 3.45μm×3.45μm. The spatial arrangement of the sensor adopts a spherical array structure, with the LED light-emitting center as the sphere center, and two measurement spheres are constructed in the near field and far field respectively. After completing the sensor arrangement, the LED light source is fixed on a precision spherical turntable for spherical scanning measurement. The scanning process is divided into two degrees of freedom: the polar angle θ is scanned from 0° to 180°, and the azimuth angle φ is scanned from 0° to 360°. The scanning step is determined according to the rate of change of the light intensity distribution on the luminous surface. A smaller angle step is used in areas where the light intensity changes dramatically, with a typical value of 0.5°, and a larger angle step is used in areas where the light intensity changes gently, with a typical value of 2°. Multiple frames of images are collected for each scanning position and superimposed and averaged to reduce the impact of random noise. At the same time, the precise spatial coordinates, LED operating parameters, and environmental parameters of each measurement point are recorded. In this way, complete spatial light field data including spatial position information, light intensity values, and measurement conditions are obtained.

[0071] Furthermore, the arranging the CMOS image sensor array of the LED light source according to the luminous characteristic data includes: calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array; spatially arranging the CMOS image sensors according to the arrangement parameters, arranging the first CMOS image sensor array in a near-field area away from the luminous surface of the LED light source, and arranging the second CMOS image sensor array in a far-field area away from the luminous surface of the LED light source; calibrating the pixel sensitivities of the first CMOS image sensor array and the second CMOS image sensor array to obtain sensor response calibration data; calculating the sampling angles and sampling intervals of the first CMOS image sensor array and the second CMOS image sensor array according to the luminous characteristic data and the sensor response calibration data to obtain spherical scanning measurement parameters, thereby realizing the arrangement of the CMOS image sensor array of the LED light source.

[0072] Specifically, when calculating the luminous intensity distribution range based on the luminous characteristic data of the LED light source, it is necessary to use the geometric dimensions, boundary contour data and brightness distribution data of the luminous surface to establish a preliminary spatial light intensity distribution estimation model. By spatially integrating the brightness distribution data of the luminous surface, the light intensity distribution function within different solid angle ranges is obtained. , where L(x,y) is the brightness value at point (x,y) on the luminous surface, is the observation angle, and the integration range is the entire luminous surface area. Based on this function, the attenuation characteristics of the luminous intensity in space are calculated to determine the spatial range that needs to be measured. According to the equivalent diameter D of the luminous surface, the measurement space is divided into the near field area (0.1D-10D) and the far field area (10D-100D). At the same time, the dynamic range of light intensity in each area is calculated to provide a basis for determining the layout parameters of the CMOS image sensor. These layout parameters include the spatial position of the sensor, the orientation angle, and the required dynamic range requirements.

[0073] Specifically, the spatial arrangement of CMOS image sensors is performed according to the obtained arrangement parameters. The first CMOS image sensor array is arranged in the near field area, using a 4×4 matrix arrangement, and the interval angle between each sensor is determined according to the near-field light intensity gradient. These sensors are fixed by precision brackets, and the positioning accuracy of the brackets is better than 0.1mm, and the angle accuracy is better than 0.1°. The second CMOS image sensor array is arranged in the far field area, using a 6×6 matrix arrangement, and the sensor interval angle is relatively uniform. Both sets of sensor arrays use high-resolution CMOS sensors with 4096×3072 pixels and a pixel size of 3.45μm×3.45μm. The optical axis of each sensor points to the LED light emission center, and precise alignment is achieved through a five-degree-of-freedom precision adjustment mechanism. The near-field sensor is equipped with a light intensity attenuation system to adapt to a higher light intensity range.

[0074] Specifically, the pixel sensitivity of two groups of CMOS image sensor arrays was calibrated. The calibration used an integrating sphere standard light source, whose light intensity uniformity was better than ±1% and light intensity stability was better than ±0.1%. The calibration process was carried out at different light intensity levels, and the light intensity range covered the actual working range of the LED. The pixel response curve of each sensor was fitted, and the response function used an improved S-shaped curve model:

[0075] ;

[0076] in is the pixel response value, is the incident light intensity, is the saturation response value, is the response coefficient, is the nonlinear correction coefficient, is the reference light intensity. In this way, the response characteristic curve of each pixel is obtained to form complete sensor response calibration data.

[0077] Specifically, the sampling angle and sampling spacing are calculated based on the luminous characteristic data and the sensor response calibration data. The sampling angle is determined based on the light intensity distribution characteristics of the luminous surface, and denser sampling points are used in areas where the light intensity changes dramatically. The sampling spacing is determined by the spatial resolution requirements and the measurement uncertainty requirements. The angle step range in the near field area is 0.5°-2°, and the angle step range in the far field area is 2°-5°. The scanning path optimization adopts the minimum path algorithm to improve the measurement efficiency. Through these parameter settings, a complete spherical scanning measurement parameter set is formed, which contains information such as the spatial coordinates, sensor selection, and exposure parameters of each measurement point. These parameters are directly used for subsequent automated measurement process control.

[0078] Furthermore, the step of calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array includes: normalizing the brightness distribution data in the luminous characteristic data, and calculating the relative brightness distribution ratio on the luminous surface according to the normalized brightness distribution data; performing a preliminary spatial range estimation according to the relative brightness distribution ratio and the geometric dimensions of the luminous surface of the LED light source to obtain the spatial measurement range of spherical scanning; dividing the spatial measurement range into regions to obtain the boundary position of the near-field measurement range and the far-field measurement range; and determining the arrangement parameters of the first CMOS image sensor array and the second CMOS image sensor array according to the boundary position.

[0079] Specifically, when normalizing the brightness distribution data in the luminous characteristic data, it is necessary to consider the actual light intensity value of each micro-region on the LED light-emitting surface. The normalization process uses a weighted method for each region to process the brightness value of each micro-region. The brightness normalization function is defined as:

[0080] ;

[0081] in For micro-region The original completion value at and are the maximum and minimum brightness values ​​of the entire luminous surface, is the position weight function. Weight function Used to compensate for edge effects. The weight function W(x,y) is determined by the geometric characteristics of the luminous surface:

[0082] ;

[0083] Where R is the equivalent radius of the luminous surface, and α is the attenuation coefficient. After normalization, the relative brightness distribution ratio on the luminous surface is obtained, which reflects the relative luminous intensity of each area, eliminates the influence of the absolute light intensity value, and facilitates the subsequent estimation of the spatial range.

[0084] Specifically, when estimating the spatial range based on the relative brightness distribution ratio and the geometric dimensions of the LED light-emitting surface, an improved light field spatial distribution model is used. This model regards the light-emitting surface as composed of multiple tiny light-emitting units, and the contribution of each unit is determined by its relative brightness value and spatial position. The relative light intensity at a point P(r,θ,φ) in space is given by the integral expression:

[0085] ;

[0086] , where r is the distance to the center of the luminous surface, θ is the observation angle, D is the diameter of the luminous surface, and k is the spatial attenuation coefficient. By calculating the relative light intensity distribution at different spatial positions, the effective measurement space range is determined. This range needs to cover the entire spatial area where the light intensity value decays from the peak value to 1%, forming a complete spherical scanning space measurement range.

[0087] Specifically, the obtained spatial measurement range is divided into regions, the main purpose of which is to determine the boundary position between the near field and the far field. The boundary position is determined based on the spatial coherence theory of the light field and the spatial coherence length is calculated: , where λ is the main wavelength of the LED, R is the observation distance, and θd is the opening angle of the light-emitting surface. When the observation distance is greater than 10 times the diameter of the light-emitting surface, the light field approximately meets the far-field condition. According to this theory, combined with the actual geometric dimensions of the light-emitting surface, 0.1-10 times the diameter of the light-emitting surface is defined as the near-field measurement range, and 10-100 times the diameter of the light-emitting surface is defined as the far-field measurement range. The precise determination of the boundary position also needs to consider the spatial gradient change of the light intensity. The light intensity gradient at the boundary position should meet the continuity condition.

[0088] Specifically, the layout parameters of the two CMOS image sensor arrays are calculated based on the determined boundary position. The layout parameters include the spatial position coordinates, orientation angle, measurement solid angle range, etc. of the sensor. For the near-field CMOS sensor array, it is arranged on a spherical surface of 0.1-10 times the diameter of the light-emitting surface, and the spacing between the sensors must meet the following requirements: , where d is the effective detection area of ​​the sensor and R is the measurement radius. The far-field CMOS sensor array is arranged on a spherical surface with a diameter of 10-100 times the luminous surface, and the sensor spacing can be appropriately increased. At the same time, the dynamic range matching of the sensor needs to be considered. The sensor in the near-field area needs to be equipped with a light intensity attenuation device, and the attenuation ratio is determined according to the relative brightness distribution. These layout parameters constitute a complete sensor array spatial layout plan.

[0089] 103. Calling a pre-trained deep learning light field reconstruction model according to the luminous feature data and the spatial light field data, compensating the luminous area effect of the light field data, and obtaining corrected spatial light field distribution data;

[0090] In one embodiment of the present invention, calling a pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, compensating the light field data for the luminous area effect, and obtaining corrected spatial light field distribution data includes: spatially dividing the spatial light field data according to the micro-area distribution data of the luminous surface in the luminous characteristic data to obtain light field sub-data corresponding to each micro-area; performing spatial coordinate mapping on the light field sub-data according to the geometric dimensions and boundary contour data of the luminous surface in the luminous characteristic data to obtain micro-area light field mapping data; inputting the micro-area light field mapping data and the brightness distribution data of the luminous surface into a pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the luminous area effect; and performing correction calculation on the spatial light field data according to the compensation coefficient to obtain corrected spatial light field distribution data.

[0091] Specifically, when spatially dividing the spatial light field data according to the micro-region distribution data in the luminous characteristic data, it is necessary to consider the contribution of each micro-region into which the original LED luminous surface is divided to the final light field distribution. The specific division method is based on the spatial mapping relationship of the micro-regions, and divides the entire measurement space into subspaces corresponding to the micro-regions of the luminous surface. Since the LED luminous surface is divided into m×n micro-regions, the measurement space is also divided into m×n subspaces accordingly. The light field data in each subspace is mapped to the corresponding micro-region through spatial transformation, and the transformation relationship is determined by the transformation matrix between the spherical coordinate system and the luminous surface coordinate system. After the spatial transformation, a series of light field sub-data are obtained, and each sub-data set contains the light field information of the corresponding micro-region at different angles and distances.

[0092] Specifically, the light field sub-data is spatially mapped according to the geometric dimensions and boundary contour data of the light-emitting surface. The spatial coordinate mapping uses a projection transformation method to map the light field data in the spherical coordinate system to a local coordinate system based on the light-emitting surface. The converted data maintains the spatial relationship of the original light field and establishes an accurate correspondence with the micro-region of the light-emitting surface. The light field data of each micro-region is independently mapped to obtain a set of micro-region light field mapping data, which reflects the independent contribution of each micro-region to the spatial light field distribution.

[0093] Specifically, the micro-area light field mapping data and the brightness distribution data of the luminous surface are input into a pre-trained deep learning light field reconstruction model. This model is a deep neural network designed specifically for the luminous area effect of LED light sources, and has been pre-trained with a large amount of standard LED light source data. The model input layer receives the micro-area light field mapping data and brightness distribution data, extracts spatial features through a multi-layer convolutional network, and finally outputs the compensation coefficient of the luminous area effect. The output compensation coefficient is a spatial distribution function that describes the degree of light field distortion caused by the luminous area effect.

[0094] Specifically, the spatial light field data is corrected and calculated according to the compensation coefficient. The correction process adopts the spatial compensation method to correct the light field data of each measurement point. The correction calculation takes into account the different characteristics of the near field and the far field, and adopts different compensation strategies for different areas. The corrected data eliminates the influence of the luminous area effect and obtains more accurate spatial light field distribution data, which can better reflect the true luminous characteristics of the LED light source.

[0095] Furthermore, the step of inputting the micro-area light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the light-emitting area effect includes: inputting the micro-area light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model, wherein the deep learning light field reconstruction model includes a convolution layer, a feature fusion layer, a pooling layer, a fully connected layer, and a deconvolution layer; performing a convolution operation on the micro-area light field mapping data through the convolution layer, extracting spatial features through the first convolution layer, and obtaining a primary feature map; performing feature fusion according to the primary feature map and the brightness distribution data through the feature fusion layer, extracting light intensity features through the second convolution layer, and obtaining a light field-brightness feature map; performing pooling operation and fully connected layer processing on the light field-brightness feature map through the pooling layer and the fully connected layer to obtain a light-emitting area effect feature vector; performing spatial restoration according to the light-emitting area effect feature vector through the deconvolution layer to obtain a compensation coefficient for the light-emitting area effect.

[0096] Specifically, when the micro-region light field mapping data and the brightness distribution data of the luminous surface are input into the pre-trained deep learning light field reconstruction model, the data needs to be formatted. The input layer of the model contains two branches: the first branch receives the m×n-dimensional micro-region light field mapping data, where m and n represent the number of rows and columns of the luminous surface, respectively, and the light field data of each micro-region contains spatial position and light intensity information; the second branch receives the brightness distribution data, which reflects the relative brightness value of each micro-region on the luminous surface. The input data needs to be normalized so that the data range is scaled to the [0,1] interval. The model structure adopts a multi-level design, including convolutional layer, feature fusion layer, pooling layer, fully connected layer and deconvolution layer. The convolutional layer is used to extract spatial features, the feature fusion layer realizes the fusion of light field data and brightness data, the pooling layer reduces the data dimension, the fully connected layer extracts global features, and the deconvolution layer restores the spatial dimension. All parameters of the model are pre-trained through a large number of standard LED light source measurement data, and the training process uses cross-validation to ensure the generalization performance of the model. The pre-trained model can automatically identify the luminous characteristics of LED light sources and generate corresponding compensation strategies for different types of luminous area effects.

[0097] Specifically, when performing spatial feature extraction on micro-region light field mapping data through the first convolution layer, an improved multi-scale convolution kernel design is adopted. The size of the convolution kernel ranges from 3×3 to 7×7. Small-sized convolution kernels are used to capture local detail features, and large-sized convolution kernels are used to extract large-scale spatial correlation features. Each convolution kernel is equipped with independent weight and bias parameters, and the weight parameters are determined by pre-training. The convolution operation adopts a sliding window method with a step size of 1 to ensure the continuity of feature extraction. The mathematical expression of the convolution layer is: ,in is the weight matrix of the kth convolution kernel, I is the input light field data, is the corresponding bias term, and ReLU is the rectified linear activation function. Through multi-layer convolution operations, the output feature map of each layer contains spatial feature information at different levels. The final primary feature map is a multi-channel feature representation, and each channel corresponds to a specific spatial feature pattern. These features include the spatial distribution features, edge features, gradient features, etc. of the light field, providing basic data for subsequent feature fusion.

[0098] Specifically, in the feature fusion layer, the primary feature map is fused with the brightness distribution data. The fusion process first upsamples the brightness distribution data so that its spatial resolution matches the primary feature map. Then the two types of data are combined in a channel cascade manner to form an augmented feature map. The augmented feature map is processed by the second convolution layer, which uses a 1×1 point convolution kernel to reduce the feature dimension and extract features related to light intensity. In this way, the spatial features and light intensity features are effectively fused to form a light field-brightness feature map. This feature map not only retains the spatial distribution information of the light field, but also includes the influence of brightness distribution, which can better reflect the characteristics of the luminous area effect. A residual connection structure is used in the feature fusion process to ensure that the original feature information is not lost during the fusion process.

[0099] Specifically, the light field-brightness feature map is dimensionally compressed and feature extracted through the pooling layer and the fully connected layer. First, the maximum pooling operation is used to reduce the data dimension, with a pooling window size of 2×2 and a step size of 2. The pooling operation can reduce the spatial resolution of the feature map while retaining the main features. The pooled feature map passes through two fully connected layers in sequence, with the number of neurons in the first fully connected layer being 1024 and the second being 512. The Dropout technique is used between the two layers with a dropout rate of 0.5 to prevent overfitting. The role of the fully connected layer is to compress the high-dimensional feature map into a low-dimensional feature vector, which contains the key feature information of the luminous area effect. In this process, spatial information and light intensity information are further integrated to form a compact representation of the luminous area effect. The final feature vector is a highly abstract feature representation that contains all the key information needed to compensate for the luminous area effect.

[0100] Specifically, the deconvolution layer is used to perform spatial restoration operations on the luminous area effect feature vector. The deconvolution process uses a transposed convolution method to gradually restore the low-dimensional feature vector to three-dimensional data that matches the original spatial dimension through a multi-layer deconvolution network. In the restoration process, upsampling and deconvolution are used alternately, and batch normalization is used after each layer of deconvolution to stabilize the feature distribution. The last layer of deconvolution outputs compensation coefficient data with the same dimension as the measurement space. These compensation coefficients form a three-dimensional tensor, and each element corresponds to the compensation value of a point in space. The generation of the compensation coefficient takes into account the geometric and optical characteristics of the LED light-emitting surface, and can effectively compensate for the measurement error caused by the light-emitting area effect. The generated compensation coefficient data is directly used for subsequent light field data correction to obtain a more accurate spatial light field distribution.

[0101] 104. Perform energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generate a three-dimensional energy density distribution diagram of the LED light source according to the calculation result.

[0102] In one embodiment of the present invention, the energy density interpolation calculation is performed on the LED light source according to the spatial light field distribution data, and a three-dimensional energy density distribution diagram of the LED light source is generated according to the calculation result, including: spatially transforming the spatial light field distribution data according to a spherical coordinate system to obtain polar coordinate data with the light-emitting center of the LED light source as the origin; performing piecewise interpolation operations on the near-field area and the far-field area of ​​the LED light source according to the polar coordinate data to obtain a continuous energy density distribution function; performing coordinate transformation from spherical coordinates to a rectangular coordinate system on the energy density distribution function to obtain an energy density value in a three-dimensional space; generating equal energy density curves of different cross sections according to the energy density values ​​in the three-dimensional space, and generating a three-dimensional energy density distribution diagram of the LED light source according to the equal energy density curves.

[0103] Specifically, when spatially transforming the spatial light field distribution data according to the spherical coordinate system, it is necessary to establish a standard spherical coordinate system with the luminous center of the LED light source as the origin. The spatial transformation process first maps the data points in the measurement coordinate system to the standard spherical coordinate system, and the transformation process needs to consider the translation and rotation of the coordinate system. The coordinates of each measurement point are converted from (x, y, z) to (r, θ, φ), where r represents the distance to the origin, θ represents the angle with the z-axis (polar angle), and φ represents the angle between the projection on the xy plane and the x-axis (azimuth). The transformation process is performed through the coordinate transformation matrix: P(r, θ, φ) = T·P(x, y, z), where T is the transformation matrix, which is determined by the spatial arrangement of the measurement system. When performing coordinate transformation, the installation error and alignment error of the measurement system also need to be considered, and these errors are compensated by the correction terms in the transformation matrix. For each spatial point, in addition to the coordinate transformation, the light intensity value of the point needs to be converted accordingly to ensure that the corresponding relationship between the light intensity value and the spatial position remains unchanged during the transformation process. The converted data forms a polar coordinate data set with the luminous center as the origin, which completely retains the spatial relationship and light intensity information of the original light field distribution. This standardized data set provides a unified spatial reference system for subsequent interpolation calculations, and also facilitates the analysis of the distribution characteristics of the light field in different spatial directions.

[0104] Specifically, when performing segmented interpolation operations on the near-field and far-field regions based on polar coordinate data, it is necessary to select interpolation methods based on the characteristics of different regions. In the near-field region (0.1-10 times the diameter of the luminous surface), due to the dramatic changes in light field distribution and the significant influence of the luminous area effect, the cubic spline interpolation method is used for data processing. The interpolation function needs to meet the conditions of equal values ​​at the nodes and continuous first-order and second-order derivatives to ensure the smoothness of the interpolation results. For each interpolation interval , construct a cubic polynomial function:

[0105] ;

[0106] coefficient , , and Determined by boundary conditions and smoothing conditions. In the far field area (10-100 times the diameter of the luminous surface), since the light field changes relatively slowly, radial basis function is used for interpolation, and the interpolation function form is: , where φ is the radial basis function, is the weight coefficient. In the transition area between the near field and the far field, a weighted average method is used to achieve a smooth transition. The weight function changes gradually with the distance to ensure the continuity of the energy density distribution function in the entire space. This piecewise interpolation method can accurately reflect the light field distribution characteristics of different regions, and the obtained continuous energy density distribution function provides basic data for subsequent coordinate transformation.

[0107] Specifically, when the energy density distribution function is transformed from spherical coordinates to rectangular coordinates, the mapping relationship between different coordinate systems needs to be considered. The transformation process first divides the space into grids and establishes uniform grid points in the rectangular coordinate system. For each grid point, the corresponding spherical coordinate position (r, θ, φ) is calculated according to its rectangular coordinate position (x, y, z), and then the energy density value of the point is calculated using the energy density distribution function in the spherical coordinate system. The transformation relationship of the metric element needs to be considered during the coordinate transformation process to ensure that the physical meaning of the energy density remains consistent in different coordinate systems. During the transformation, an appropriate Jacobian matrix should be used to transform the metric element to ensure that the corresponding relationship of the spatial volume element is correct during the transformation process. At the same time, it is also necessary to deal with the singular point problem in the coordinate transformation, especially in the area near the coordinate axis and the origin, and to ensure that the energy density values ​​of these special positions are accurately calculated by interpolation or extrapolation. The three-dimensional spatial energy density numerical data obtained after the transformation covers the entire measurement space, and each spatial point corresponds to a specific energy density value. These data directly reflect the energy distribution of the LED light source at each point in the space.

[0108] Specifically, when generating iso-energy density curves of different sections according to the energy density values ​​in three-dimensional space and further constructing a three-dimensional energy density distribution map, it is necessary to first determine the typical observation section. Multiple characteristic sections are selected on the xy plane, xz plane and yz plane respectively, and the positions of these sections are determined according to the spatial distribution characteristics of energy density. On each two-dimensional section, an iso-energy density curve is generated by using an iso-value algorithm, and the density and interval of the iso-value lines are determined according to the gradient distribution of energy density. In areas where the energy density changes drastically, a denser iso-value distribution is used to highlight the local change characteristics; in areas where the energy density changes gently, a sparser iso-value distribution is used to avoid information redundancy. The generation of iso-value lines uses an improved contour tracing algorithm, which can accurately capture energy density iso-value points and generate smooth iso-value lines. Then, the iso-energy density curves of these two-dimensional sections are combined through a three-dimensional reconstruction algorithm, and a three-dimensional interpolation method is used to generate transition surfaces between sections to form a complete three-dimensional energy density distribution map. This expression method intuitively shows the spatial energy distribution characteristics of LED light sources, which is convenient for analyzing the luminous characteristics and application design of light sources.

[0109] In this embodiment, the luminous characteristic data is obtained by collecting and preprocessing the front image of the luminous surface of the LED light source. According to these data, a CMOS image sensor array is arranged, and the LED light source is spherically scanned and measured by the array to obtain spatial light field data. Using the pre-trained deep learning light field reconstruction model, the light field data is compensated for the luminous area effect to obtain the corrected spatial light field distribution data. Finally, the energy density interpolation calculation is performed based on the corrected data to generate a three-dimensional energy density distribution diagram. This method combines the image sensor array, deep learning reconstruction technology and three-dimensional space measurement model to achieve accurate measurement of the energy density distribution of the LED light source, overcome the limitations of the existing methods, improve the measurement accuracy, and can fully reflect the three-dimensional energy density distribution of the LED light source.

[0110] The above describes the method for measuring the energy density distribution of an LED light source in an embodiment of the present invention. The following describes the device for measuring the energy density distribution of an LED light source in an embodiment of the present invention. Figure 2 In one embodiment of the present invention, an LED light source energy density distribution measuring device comprises:

[0111] The preprocessing model 201 is used to collect and preprocess the front image of the light-emitting surface of the LED light source to obtain the light-emitting characteristic data of the LED light source;

[0112] A scanning module 202 is used to arrange a CMOS image sensor array of the LED light source according to the light-emitting characteristic data, and perform spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data;

[0113] A compensation module 203 is used to call a pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, and perform luminous area effect compensation on the light field data to obtain corrected spatial light field distribution data;

[0114] The distribution calculation module 204 is used to perform energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generate a three-dimensional energy density distribution diagram of the LED light source according to the calculation result.

[0115] In an embodiment of the present invention, the LED light source energy density distribution measuring device runs the above-mentioned LED light source energy density distribution measuring method, and the LED light source energy density distribution measuring device obtains luminous characteristic data by performing front image acquisition and preprocessing on the light-emitting surface of the LED light source. According to these data, a CMOS image sensor array is arranged, and the LED light source is spherically scanned and measured by the array to obtain spatial light field data. Using the pre-trained deep learning light field reconstruction model, the light field data is compensated for the luminous area effect to obtain the corrected spatial light field distribution data. Finally, energy density interpolation calculation is performed based on the corrected data to generate a three-dimensional energy density distribution diagram. This method realizes accurate measurement of the energy density distribution of the LED light source by combining the image sensor array, deep learning reconstruction technology and three-dimensional space measurement model, overcomes the limitations of the existing methods, improves the measurement accuracy, and can fully reflect the three-dimensional energy density distribution of the LED light source.

[0116] above Figure 2 The LED light source energy density distribution measuring device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The LED light source energy density distribution measuring device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0117] Figure 3 3 is a schematic diagram of the structure of an LED light source energy density distribution measuring device provided by an embodiment of the present invention. The LED light source energy density distribution measuring device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the LED light source energy density distribution measuring device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the LED light source energy density distribution measuring device 300 to implement the steps of the above-mentioned LED light source energy density distribution measuring method.

[0118] The LED light source energy density distribution measuring device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the LED light source energy density distribution measuring device shown does not constitute a limitation on the LED light source energy density distribution measuring device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0119] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the LED light source energy density distribution measurement method.

[0120] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0121] If the integrated unit is implemented in the form of 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 the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0122] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the energy density distribution of an LED light source, characterized in that: The LED light source energy density distribution measurement includes: Performing front image acquisition and preprocessing on the light-emitting surface of the LED light source to obtain light-emitting characteristic data of the LED light source; Arranging the CMOS image sensor array of the LED light source according to the luminous characteristic data, and performing spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data; the arranging the CMOS image sensor array of the LED light source according to the luminous characteristic data includes: calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array; spatially arranging the CMOS image sensors according to the arrangement parameters, arranging the first CMOS image sensor array in a near-field area away from the luminous surface of the LED light source, and arranging the second CMOS image sensor array in a far-field area away from the luminous surface of the LED light source; calibrating the pixel sensitivities of the first CMOS image sensor array and the second CMOS image sensor array to obtain sensor response calibration data; calculating the sampling angles and sampling intervals of the first CMOS image sensor array and the second CMOS image sensor array according to the luminous characteristic data and the sensor response calibration data to obtain spherical scanning measurement parameters, and realizing the arrangement of the CMOS image sensor array of the LED light source; According to the luminous characteristic data and the spatial light field data, a pre-trained deep learning light field reconstruction model is called to compensate for the luminous area effect of the light field data to obtain the corrected spatial light field distribution data; the calling of the pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data to compensate for the luminous area effect of the light field data to obtain the corrected spatial light field distribution data includes: spatially dividing the spatial light field data according to the micro-region distribution data of the luminous surface in the luminous characteristic data to obtain light field sub-data corresponding to each micro-region; spatially coordinate mapping the light field sub-data according to the geometric size and boundary contour data of the luminous surface in the luminous characteristic data to obtain micro-region light field mapping data; inputting the micro-region light field mapping data and the brightness distribution data of the luminous surface into the pre-trained deep learning light field reconstruction model to obtain the compensation coefficient of the luminous area effect; performing correction calculation on the spatial light field data according to the compensation coefficient to obtain the corrected spatial light field distribution data; According to the spatial light field distribution data, an energy density interpolation calculation is performed on the LED light source, and a three-dimensional energy density distribution diagram of the LED light source is generated according to the calculation result.

2. The method for measuring energy density distribution of an LED light source according to claim 1, characterized in that: The front image acquisition and preprocessing of the light-emitting surface of the LED light source to obtain the light-emitting characteristic data of the LED light source includes: Capturing multiple frames of high-resolution images of the light-emitting surface of the LED light source to obtain original image data of the light-emitting surface; Performing threshold segmentation and edge detection on the original image data of the light-emitting surface to obtain geometric dimensions and boundary contour data of the light-emitting surface; Performing grid division on the original image data of the light-emitting surface to obtain micro-region distribution data of the light-emitting surface; Performing grayscale value analysis on the light-emitting surface according to the boundary contour data and the micro-region distribution data to obtain brightness distribution data of the light-emitting surface; The geometric dimensions, the boundary contour data, the micro-area distribution data and the brightness distribution data are integrated to obtain the light emitting characteristic data of the LED light source.

3. The method for measuring energy density distribution of an LED light source according to claim 1, characterized in that: The step of calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array comprises: Normalizing the brightness distribution data in the light-emitting characteristic data, and calculating the relative brightness distribution ratio on the light-emitting surface according to the normalized brightness distribution data; Performing a preliminary spatial range estimation based on the relative brightness distribution ratio and the geometric dimensions of the light-emitting surface of the LED light source to obtain a spatial measurement range of spherical scanning; Dividing the spatial measurement range into regions to obtain a boundary position between a near-field measurement range and a far-field measurement range; Arrangement parameters of the first CMOS image sensor array and the second CMOS image sensor array are determined according to the boundary position.

4. The method for measuring energy density distribution of an LED light source according to claim 1, characterized in that: The step of inputting the micro-area light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the light-emitting area effect includes: Inputting the micro-region light field mapping data and the brightness distribution data of the light-emitting surface into a pre-trained deep learning light field reconstruction model, wherein the deep learning light field reconstruction model includes a convolution layer, a feature fusion layer, a pooling layer, a fully connected layer, and a deconvolution layer; Performing a convolution operation on the micro-region light field mapping data through the convolution layer, extracting spatial features through the first convolution layer, and obtaining a primary feature map; Performing feature fusion according to the primary feature map and the brightness distribution data through the feature fusion layer, extracting light intensity features through the second convolution layer, and obtaining a light field-brightness feature map; Performing a pooling operation and a fully connected layer processing on the light field-brightness feature map through the pooling layer and the fully connected layer to obtain a luminous area effect feature vector; The deconvolution layer performs spatial restoration according to the luminous area effect feature vector to obtain a compensation coefficient of the luminous area effect.

5. The method for measuring energy density distribution of an LED light source according to claim 1, characterized in that: The step of performing energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generating a three-dimensional energy density distribution diagram of the LED light source according to the calculation result comprises: Performing spatial transformation on the spatial light field distribution data according to a spherical coordinate system to obtain polar coordinate data with the light emitting center of the LED light source as the origin; Performing a segmented interpolation operation on the near field area and the far field area of ​​the LED light source according to the polar coordinate data to obtain a continuous energy density distribution function; Performing coordinate transformation from spherical coordinates to rectangular coordinates on the energy density distribution function to obtain an energy density value in three-dimensional space; Equal energy density curves of different cross sections are generated according to the energy density values ​​of the three-dimensional space, and a three-dimensional energy density distribution diagram of the LED light source is generated according to the equal energy density curves.

6. A device for measuring energy density distribution of LED light source, characterized in that: The LED light source energy density distribution measuring device comprises: A preprocessing model, used for collecting and preprocessing the front image of the light-emitting surface of the LED light source to obtain light-emitting characteristic data of the LED light source; A scanning module, used for arranging the CMOS image sensor array of the LED light source according to the luminous characteristic data, and performing spherical scanning measurement on the LED light source according to the CMOS image sensor array to obtain spatial light field data; the arranging the CMOS image sensor array of the LED light source according to the luminous characteristic data comprises: calculating the luminous intensity distribution range of the LED light source according to the luminous characteristic data of the LED light source to obtain the arrangement parameters of the CMOS image sensor array; spatially arranging the CMOS image sensors according to the arrangement parameters, arranging the first CMOS image sensor array in a near field area away from the luminous surface of the LED light source, and arranging the second CMOS image sensor array in a far field area away from the luminous surface of the LED light source; calibrating the pixel sensitivities of the first CMOS image sensor array and the second CMOS image sensor array to obtain sensor response calibration data; calculating the sampling angles and sampling intervals of the first CMOS image sensor array and the second CMOS image sensor array according to the luminous characteristic data and the sensor response calibration data to obtain spherical scanning measurement parameters, and realizing the arrangement of the CMOS image sensor array of the LED light source; A compensation module, used to call a pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, compensate the light field data for the luminous area effect, and obtain corrected spatial light field distribution data; the calling of the pre-trained deep learning light field reconstruction model according to the luminous characteristic data and the spatial light field data, compensate the light field data for the luminous area effect, and obtain corrected spatial light field distribution data includes: spatially dividing the spatial light field data according to the micro-region distribution data of the luminous surface in the luminous characteristic data to obtain light field sub-data corresponding to each micro-region; spatially coordinate mapping the light field sub-data according to the geometric size and boundary contour data of the luminous surface in the luminous characteristic data to obtain micro-region light field mapping data; inputting the micro-region light field mapping data and the brightness distribution data of the luminous surface into the pre-trained deep learning light field reconstruction model to obtain a compensation coefficient for the luminous area effect; performing correction calculation on the spatial light field data according to the compensation coefficient to obtain corrected spatial light field distribution data; A distribution calculation module is used to perform energy density interpolation calculation on the LED light source according to the spatial light field distribution data, and generate a three-dimensional energy density distribution diagram of the LED light source according to the calculation result.

7. A device for measuring energy density distribution of LED light source, characterized in that: The LED light source energy density distribution measuring device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the LED light source energy density distribution measurement device to perform the steps of the LED light source energy density distribution measurement method as described in any one of claims 1-5.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the LED light source energy density distribution measurement method as described in any one of claims 1-5 are implemented.

Images