Part surface reflection light suppression method based on high dynamic range imaging and photometric stereo vision

Abstract

The application discloses a part surface reflection light suppression method based on high dynamic range imaging and photometric stereo vision, and has the characteristics that: first, a visual detection system is established, the visual detection system comprising a visual system framework, an industrial camera and a lens, a diffuse reflection light source and a strip light source, a polarizer and a light source controller; second, the industrial camera and the lens are assembled, the polarizer is arranged on the lens, the camera is fixed on the upper middle of the visual system framework, and the part to be detected is placed below the camera; third, the light source is arranged; meanwhile, the strip light source is arranged at multiple directions of the part to be detected, and the light source is connected with the light source controller; fourth, the detection is completed. The application solves the interference of the part surface high reflection light to the visual detection in the image preprocessing process, effectively suppresses the part high reflection light interference, and introduces the visual detection method into the industrial field, thereby expanding the applicability of the visual detection.

Description

Technical Field

[0001] This invention relates to a method for suppressing surface reflection of stamped industrial parts, particularly for stamped parts with high surface reflectivity. By suppressing reflection, the method improves the success rate of surface feature recognition. Specifically, it is a method for suppressing surface reflection of parts based on high dynamic range imaging and photometric stereo vision. It can be used to solve the visual inspection problem caused by high surface reflectivity of stamped parts, especially the problem of surface defect recognition. Background Technology

[0002] This invention primarily targets stamped parts, where defects mainly originate from the stamping process. Stamping is a processing method that uses a die to apply pressure or tension to sheet metal, causing it to plastically shape, and sometimes shearing force to separate the sheet metal, thereby obtaining a specific size, shape, and performance. The process from the punch contacting the sheet metal to the sheet metal separating is completed instantaneously. When the clearance between the punch and die is normal, the stamping process can be roughly divided into three stages: elastic deformation, plastic deformation, and fracture separation. During this stamping process, various types of forming defects can occur. These defects severely affect the dimensional accuracy, surface quality, and mechanical properties of the stamped parts. In summary, the main defects are wrinkling, cracking, and springback.

[0003] Wrinkling and cracking are relatively easy to cause defects on the smooth surface of parts. Wrinkling is the main manifestation of compressive instability in sheet metal forming. During sheet metal stamping, the sheet metal generates a complex stress state under the external force applied by the die. When the compressive stress in the sheet metal reaches a certain value, the thickness dimension is the smallest, making it most prone to compressive instability and wrinkling. At the same time, shear force, uneven tensile force, and in-plane bending force can also cause wrinkling. Cracking is the main manifestation of tensile instability in sheet metal stamping. As deformation develops during the sheet metal forming process, the load-bearing area of ​​the material continuously shrinks, and the strain hardening effect continuously increases. When the increase in strain hardening effect cannot compensate for the reduction in load-bearing area and exceeds the critical state, the sheet metal eventually cracks.

[0004] Besides surface defects caused by the stamping process, scratches can also occur on parts due to collisions with other parts and machining tools during production. Large-scale stacking of parts can also cause crushing damage. Inadequate cleaning can lead to rust due to residual dirt.

[0005] In summary, during the production of these parts, due to various factors such as raw materials, production processes, and operator skills, the surface of metal parts may exhibit various defects such as scratches, dents, rust, wrinkles, and dimensional misalignments. These surface defects not only pose significant safety hazards but also reduce the corrosion resistance, wear resistance, and fatigue resistance of the components, affecting the safe use of the product and shortening its lifespan.

[0006] The technology for detecting surface defects in metal parts has made significant progress over the past few decades, evolving from initial manual visual inspection to today's automated testing. Commonly used non-destructive testing methods include infrared non-destructive testing (NDT), magnetic flux leakage NDT, eddy current NDT, and machine vision NDT. With advancements in computers and hardware, machine vision has become increasingly prevalent in the field of inspection. Machine vision methods utilize industrial cameras to capture images of the surface of a part to be inspected. A series of visual processing algorithms then eliminate surface interference, ultimately identifying the shape of defects and classifying them based on their characteristics. Machine vision methods are not only fast and efficient but also offer significantly higher stability than manual methods, making them ideal for defect detection in large-scale parts production in factories.

[0007] Many parts currently require a high surface finish and often necessitate surface coatings, which often enhances the reflectivity of the surface. Due to the high reflectivity of stamped industrial parts, which are highly reflective, the resulting localized specular reflections can cause localized oversaturation in the images captured by industrial cameras, creating blind spots and obscuring surface quality information within the complex surrounding environment. This significantly reduces the success rate of visual defect detection. Therefore, solving the defect detection problem of highly reflective surfaces has become a significant challenge in the industrial field. Consequently, it is necessary to find a method to reduce reflective interference from highly reflective surfaces, thereby improving the success rate of visual defect detection. Summary of the Invention

[0008] The purpose of this invention is to address the problem that the brightness of images captured by industrial cameras during the surface image inspection of existing parts is prone to local oversaturation, creating blind spots and causing the quality information of the measured surface to be hidden in the surrounding complex environment, which greatly reduces the success rate of visual inspection of surface defects. The invention proposes a method for suppressing surface reflection of parts based on high dynamic range imaging and photometric stereo vision. By suppressing the reflection interference of highly reflective parts, the success rate of machine vision in detecting surface features of parts is greatly improved.

[0009] The technical solution of the present invention:

[0010] A method for suppressing surface reflection of parts based on high dynamic range imaging and photometric stereo vision, characterized by comprising the following steps:

[0011] First, establish a visual inspection system, which includes:

[0012] A vision system framework for supporting fixed vision inspection components;

[0013] Industrial cameras and lenses used to capture images of part surfaces;

[0014] Diffuse light sources and strip light sources are used to illuminate the surface of parts;

[0015] Polarizing filters are used to reduce the reflectivity of highly reflective surfaces.

[0016] A light source controller is used to control the intensity of light irradiation.

[0017] Next, assemble the industrial camera and lens, install the polarizer on the lens, fix the camera as a whole above the center of the vision system frame, place the part to be tested directly below the camera, and adjust the lens focus until the image is clear.

[0018] Third, set up the light sources. Place the diffuse light source within the frame; the position is arbitrary as long as it does not obstruct the shooting. At the same time, place the strip light sources at multiple locations on the part under test, and connect all the light sources to the light source controller.

[0019] Fourth, the testing steps are as follows:

[0020] (7) Keep the diffuse light source on and illuminate all the bar light sources and the part under test at a fixed angle θ, but turn off all the bar light sources;

[0021] (8) Turn on the strip light source 1 at the illumination angle 1 position, close the light sources in other directions, control the brightness of the light source to a fixed intensity using the light source controller, take pictures of multiple exposure times, select a few suitable exposure times according to the quality measure, and use the high dynamic range imaging algorithm to synthesize a high dynamic image 1 from several pictures of suitable exposure times.

[0022] (9) Turn on the bar light source 2 at the position of illumination angle 2 and close the bar light sources in other directions. Similarly, take pictures of multiple exposure times and synthesize high dynamic range image 2.

[0023] (10) Similarly, the images from the other directions are combined to form high dynamic range image 3 and high dynamic range image 4 respectively;

[0024] (11) Combine the high dynamic range images 1-4 with the photometric stereo algorithm to synthesize a final effect image;

[0025] (12) Visually process the final image and use the sub-pixel edge algorithm to extract the surface defects of the parts.

[0026] The beneficial effects of this invention are:

[0027] This invention addresses the application of visual defect detection on highly reflective surfaces of stamped parts, resolving the interference caused by high surface reflectivity during image preprocessing. By combining high dynamic range imaging and photometric stereo vision, this invention effectively suppresses interference from high reflectivity, while simultaneously introducing visual inspection methods into the industrial field, expanding their applicability. This invention also offers the following significant advantages:

[0028] (1) This invention includes two types of light source illumination: diffuse reflection light source and strip light source. Compared with the common single light source, it not only reduces the high reflectivity interference caused by specular reflection, but also enhances the detail information of the image, which is beneficial for subsequent defect detection.

[0029] (2) Currently, common visual inspection methods simply use the multi-exposure fusion method, while this invention adopts two methods: high dynamic range imaging technology and photometric stereo vision technology, which combine the advantages of both, thereby suppressing the interference of surface reflection of the parts to the greatest extent.

[0030] (3) The method of multi-exposure fusion requires the selection of a suitable exposure time. This invention adopts a method of selecting the exposure time based on quality measurement, which not only reduces the workload but also has good experimental results and can well meet the requirements.

[0031] (4) Photometric stereo vision technology is a research field that estimates the surface normal of an object by observing images under different lighting directions in a static scene. It can use multi-illuminance information to eliminate the interference of invalid information and obtain more accurate surface detail information, which is beneficial for defect identification. However, this method currently mostly uses the final generated surface gradient map for application in the field of three-dimensional stereo imaging. This invention applies the photometric stereo vision method to suppress reflection and only uses its reflectance map. When processing the results, it not only requires less computation but also meets the requirements of visual inspection. Attached Figure Description

[0032] Figure 1 This is the overall flowchart of the present invention.

[0033] Figure 2 This is a framework diagram of the visual inspection platform of the present invention.

[0034] Figure 3 This is a schematic diagram of the diffuse reflection principle of the present invention.

[0035] Figure 4 This is a basic flowchart of the high dynamic range imaging of the present invention.

[0036] Figure 5 This is an overall diagram illustrating the principle of the photometric stereo method of this invention.

[0037] Figure 6 This is a top view of the principle of the photometric stereo method of this invention. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more explicit definition of the scope of protection of the present invention.

[0039] like Figure 1-6 As shown.

[0040] A method for identifying surface defects of reflective parts based on high dynamic range imaging and photometric stereo vision, the overall process of which is described in [link to documentation]. Figure 1 This method includes a visual inspection platform and a defect recognition method based on high dynamic range imaging and photometric stereo vision, wherein:

[0041] The visual inspection platform consists of a vision system frame, industrial cameras and lenses, diffuse and strip light sources, polarizers, and a light source controller. During operation, the entire vision system operates in a closed space, preventing any natural light from entering. A schematic diagram of the overall visual inspection platform can be found here. Figure 2 .

[0042] The method for suppressing surface reflection of parts based on high dynamic range imaging and photometric stereo vision includes the following steps:

[0043] The first step addresses the issue of glare caused by direct illumination of a part's surface using traditional light sources. Specular reflection leads to significant glare interference. Therefore, this invention replaces the primary light source with diffuse reflection. Diffuse reflection is the phenomenon where light scatters uniformly in all directions after striking a rough surface. This scattering occurs because light encounters the surface's microscopic irregularities, causing reflections in various directions. Diffuse reflection results in a more uniform distribution of light across the object's surface, unlike specular reflection which exhibits a clear directionality, thus suppressing glare to some extent. The principle of diffuse reflection is explained in [link to article]. Figure 3 .

[0044] The second step involves installing a polarizer on the camera lens. Polarizers weaken the intensity of strongly reflected light beams. A polarizer is a special optical filter that can selectively filter out the direction of light wave vibration, thereby reducing or eliminating specific vibration directions in the reflected light. When light encounters a surface, it often reflects. Using a polarizer allows light from only one direction to pass through, while light from the other direction is absorbed or blocked, thus reducing or eliminating specific vibration directions in the reflected light and mitigating reflection interference to some extent.

[0045] The third step involves using high dynamic range (HDR) imaging technology to obtain composite images of the part from multiple directions. The principle of HDR imaging technology is to expose the same object multiple times, obtaining images at different exposure times, and then fusing these images to obtain a high dynamic range image with uniform intensity and rich detail.

[0046] (3.1) Firstly, the more exposure times selected, the more images can be used for fusion, and the richer the details will be. However, the processing efficiency will be greatly reduced. Therefore, it is necessary to select an appropriate exposure time. This invention uses quality measures to select the exposure time, which include image contrast, saturation, and good exposure.

[0047] The specific implementation steps of this method are as follows:

[0048] (3.1.1) Calculate contrast ratio. Contrast ratio is a measure of the difference between a pixel and its neighboring pixels. The formula for calculating contrast ratio is:

[0049]

[0050] Where (i,j) are the pixel coordinates, C(i,j) is the contrast ratio, and T(i,j) is the image grayscale value of the pixel at (i,j). H is the convolution operator, and H is the Laplace operator, where H = [(1,1,1),(1,-6,1),(1,1,1)];

[0051] (3.1.2) Calculate saturation. Higher saturation results in more vibrant colors in the image. Saturation measures how closely the value of each individual color channel approximates the average of the RGB channels. The formula for calculating saturation is:

[0052]

[0053] Where R, G, and B are the normalized pixel values ​​of the red, green, and blue channels of the image, respectively, and μ is the average value of the normalized pixel values ​​of the three channels.

[0054] (3.1.3) Calculate Good Exposure. Good exposure describes the degree of exposure of an image; an image that is not overexposed but still displays most of the details is a well-exposed image. First, set the pixel value range [a, b] for good exposure based on the shooting environment and image requirements. Both a and b range from 0 to 255, and a ≤ b. Classify all pixels. Pixel values ​​less than a are underexposed and assigned a value of 0; values ​​greater than b are overexposed and assigned a value of 1; values ​​within [a, b] are well-exposed and assigned a value of 0.5. The formula for calculating good exposure is:

[0055]

[0056] Where E is the optimal exposure, and x is a constant of 0.2;

[0057] (3.1.4) Calculate the quality measurement index. The quality measurement index is obtained by combining the above three indicators. The calculation formula is:

[0058] Z w (i,j)=C w (i,j) α S w (i,j)βE w (i,j) γ (4)

[0059] Among them, C w (i,j), S w (i,j), E w (i,j) represent the contrast, saturation, and exposure quality of the w-th image at pixel (i,j), and α, β, and γ are the corresponding adjustment indices.

[0060] (3.1.5) Obtain a suitable exposure time. In the example, the image quality measurement index is calculated based on the specific conditions such as the shooting environment. Images that meet certain standards are selected and used as a suitable exposure sequence.

[0061] (3.2) After selecting a suitable exposure sequence, the camera's response function is calculated using parameters such as the brightness value and exposure time of the exposure sequence images, thereby generating a high dynamic range image mapping map. Finally, the mapping map is compressed in dynamic range based on a tone mapping algorithm so that it can be displayed on the monitor. See the detailed flowchart below. Figure 4 .

[0062] The steps of this method in an example are as follows:

[0063] (3.2.1) Take multiple images with a suitable exposure sequence. When taking images, the intensity of the diffuse light source remains constant, and the intensity of the bar light source is set to a fixed value. After taking images using the above method, an exposure sequence suitable for the current shooting environment is obtained, consisting of four exposure times: 0.08s, 0.17s, 0.32s, and 0.45s. Four images are then taken at each of these four exposure times.

[0064] (3.2.2) Image alignment. The four images must be aligned, meaning there cannot be excessive displacement, otherwise there will be severe artifacts. Therefore, the camera and lens must be fixed during shooting, and the positions of the light source and parts must not be moved.

[0065] (3.2.3) Extracting the camera response function. The problem of the camera response function is essentially an optimal solution problem, which can be solved by reducing the dimensionality to a linear least squares problem or using singular value decomposition.

[0066] (3.2.4) Merging Images. Once the camera response function evaluation is complete, the four images are merged into a single high dynamic range image using the MergeDebevec algorithm.

[0067] (3.2.5) Tone Mapping. Tone mapping is used to convert high dynamic range images into 8-bit single-channel images. This technique adjusts the hue and color saturation of an image to achieve different visual effects. In the example, by comparing several commonly used tone mapping algorithms, the Reinhard tone mapping algorithm is found to be the most suitable for the experimental results of this invention.

[0068] The fourth step involves importing the high dynamic range image obtained in the third step into the photometric stereo algorithm to obtain a synthesized reflectance image. Photometric stereo is a 3D reconstruction method based on brightness and darkness information. The relative positions of the camera and the object under test remain fixed. The same object is illuminated by light sources from at least three different directions. The normal vector map, gradient map, and reflectance map of the object's surface are calculated. The resulting reflectance map is then used for defect identification. See the schematic diagram below. Figure 5 and Figure 6 .

[0069] A set of images of the object under test are captured by a camera under illumination from light sources at different spatial angles. According to the Lambertian ideal reflection model, the formula for the intensity of diffuse reflected light is:

[0070] I=ρLN (5)

[0071] Where I is the image sensor brightness, ρ is the surface reflectivity of the object being measured, its value is between 0 and 1, reflecting the surface characteristics of the object, and L = (L x, L y ,L z ) is the unit direction vector of the light source, which is the illumination direction of the light source, N = (N x N y N z ) T Let be the unit normal vector of a point on the surface of the object to be measured.

[0072] ρN corresponding to each pixel position is a three-dimensional vector, so a single light source cannot solve equation (5). At least three light sources at different positions are needed to solve for three unknowns. The classic photometric stereo algorithm uses three light source positions, but in this example, the photometric stereo method with four light source positions can more accurately describe the surface gradient and reflectivity information of the object under test.

[0073] In four images illuminated by light sources at four different angles, the brightness of a pixel is I1, I2, I3, and I4, respectively. The unit direction vectors of the four light sources are L1, L2, L3, and L4, respectively. The reflected light brightness of the pixel is as follows:

[0074]

[0075] N and L are unit vectors. Using the least squares method to solve equation (6), we obtain the surface reflectance of the object to be measured as:

[0076] ρ=‖(L T L) -1 (L T L)‖ (7)

[0077] The fifth step involves using a sub-pixel edge detection algorithm to extract the defect contours and locate the defect positions from the obtained reflectance map. Currently, numerous sub-pixel contour extraction algorithms exist. This invention compares several relatively effective algorithms, including mean filtering combined with local thresholding, Gaussian filtering combined with image convolution, grayscale thresholding, and Canny edge detection. Ultimately, the grayscale thresholding sub-pixel contour extraction algorithm was found to be the most effective. Based on a set threshold, the entire image is divided into foreground and background. Pixels with values ​​greater than or equal to the set threshold are considered foreground pixels, and pixels with values ​​less than the threshold are considered background pixels. Utilizing sub-pixel contours allows for more accurate acquisition of defect contours, improving recognition accuracy.

[0078] The parts not covered in this invention are the same as or can be implemented using existing technologies.

Claims

1. A method for suppressing surface reflection of parts based on high dynamic range imaging and photometric stereo vision, characterized by: It includes the following steps: First, establish a visual inspection system, which includes: A vision system framework for supporting fixed vision inspection components; Industrial cameras and lenses used to capture images of part surfaces; Diffuse light sources and strip light sources are used to illuminate the surface of parts; Polarizing filters are used to reduce the reflectivity of highly reflective surfaces. A light source controller is used to control the intensity of light irradiation. Next, assemble the industrial camera and lens, install the polarizer on the lens, fix the camera as a whole above the center of the vision system frame, place the part to be tested directly below the camera, and adjust the lens focus until the image is clear. Third, set up the light source; place the diffuse light source within the frame, the position is arbitrary, as long as it does not obstruct the shooting; at the same time, place the strip light source in multiple positions on the part to be tested, and connect all the light sources to the light source controller; Fourth, complete the testing; The detection process includes the following steps: (1) Keep the diffuse light source on and illuminate all the bar light sources and the part under test at a fixed angle θ, but turn off all the bar light sources; (2) Turn on the strip light source 1 at the illumination angle 1 position, close the light sources in other directions, control the brightness of the light source to a fixed intensity using the light source controller, take pictures of multiple exposure times, select several exposure times according to the quality measure, and use the high dynamic range imaging algorithm to synthesize several pictures of suitable exposure times into a high dynamic image 1. (3) Turn on the bar light source 2 at the position of illumination angle 2 and close the bar light sources in other directions. Similarly, take pictures of multiple exposure times and synthesize the high dynamic range image 2. (4) Similarly, the images from the other directions are combined to form high dynamic range image 3 and high dynamic range image 4 respectively; (5) Combine the high dynamic range images 1-4 with the photometric stereo algorithm to create a final image; (6) Visually process the final image and use the sub-pixel edge algorithm to extract the surface defects of the parts. (7) The selection of exposure time based on quality measurement includes the following steps: First, calculate the contrast, saturation, good exposure, quality metrics, and obtain the exposure time; (1) Calculate contrast ratio; contrast ratio is an indicator that measures the difference between a pixel and its neighboring pixels; the formula for calculating contrast ratio is: Where (i,j) are the pixel coordinates, C(i,j) is the contrast ratio, and T(i,j) is the image grayscale value of the pixel at (i,j). H is the convolution operator, and H is the Laplace operator, where H = [(1,1,1),(1,-6,1),(1,1,1)]; (2) Calculate saturation; the higher the saturation, the more vibrant the colors in the image; saturation measures how close the value of each individual color channel is to the average value of the RGB channels; the formula for calculating saturation is: Where R, G, and B are the normalized pixel values ​​of the red, green, and blue channels of the image, respectively, and μ is the average value of the normalized pixel values ​​of the three channels. (3) Calculate good exposure; good exposure describes the degree of exposure of an image. An image that is not overexposed but can display most details is a high-quality exposure image. First, set the pixel value range [a, b] for good exposure according to the shooting environment and image requirements. The range of a and b is 0-255, and a≤b. Classify all pixels: Pixels with a value less than a are underexposed and assigned a value of 0; pixels with a value greater than b are overexposed and assigned a value of 1; pixels within [a, b] are well exposed and assigned a value of 0.5; The formula for calculating good exposure is: Where E is the optimal exposure, and x is a constant of 0.2; (4) Calculate the quality measurement index; the quality measurement index is obtained by combining the above three indicators. The calculation formula is as follows: Z w (i,j)=C w (i,j) α S w (i,j) β E w (i,j) γ (4) Wherein, C w (i,j), S w (i,j), E w (i,j) represent the contrast, saturation, and exposure quality of the w-th image at pixel (i,j), and α, β, and γ are the corresponding adjustment indices. (5) Obtain the exposure time; calculate the image quality measurement index based on the shooting environment, select images that meet the set standard, and select them as the appropriate exposure sequence; Secondly, after selecting a suitable exposure sequence, the camera's response function is calculated using parameters such as the brightness value and exposure time of the exposure sequence images, thereby generating a high dynamic range image mapping map. Finally, the mapping image is compressed in dynamic range based on the tone mapping algorithm so that it can be displayed on the monitor.

2. The method according to claim 1, characterized in that: The process of selecting a suitable exposure sequence and then using parameters such as the brightness value and exposure duration of the exposure sequence images to calculate the camera's response function, thereby generating a high dynamic range image mapping map, includes the following steps: (1) Take multiple images with a suitable exposure sequence; when taking images, the intensity of the diffuse light source remains unchanged, and the intensity of the strip light source is set to a fixed value; after taking images using the above method, the exposure sequence that conforms to the current shooting environment is obtained as 4 exposure times, namely 0.08s, 0.17s, 0.32s and 0.45s, and take four images at these 4 exposure times; (2) Align the images; the four images must be aligned, that is, there cannot be too much displacement, otherwise there will be serious artifacts. Therefore, the camera and lens must be fixed during shooting, and the positions of the light source and parts cannot be moved. (3) Extract the camera response function; the problem of the camera response function is essentially an optimal solution problem, which can be solved by reducing the dimension to a linear least squares problem or using singular value decomposition. (4) Merge images; Once the camera response function evaluation is complete, the MergeDebevec algorithm is used to merge the four images into a single high dynamic range image; (5) Tone mapping; tone mapping is used to convert high dynamic range images into 8-bit single-channel images and adjust the hue and color saturation of the image to achieve different visual effects.

3. The method according to claim 1, characterized in that: The method described above is a three-dimensional reconstruction method based on light and dark information, which imports the obtained high dynamic range image into the photometric stereo algorithm to obtain a synthesized reflectance image. The relative positions of the camera and the object under test are fixed. The same object under test is illuminated by light sources from at least three different directions. The normal vector map, gradient map and reflectance map of the object surface are calculated. The reflectance map is then used for defect identification. A set of images of the object under test are captured by a camera under illumination from light sources at different spatial angles. According to the Lambertian ideal reflection model, the formula for the intensity of diffuse reflected light is: I=ρLN (5) Where I is the image sensor brightness, ρ is the surface reflectivity of the object being measured, its value is between 0 and 1, reflecting the surface characteristics of the object, and L = (L x, L y ,L z ) is the unit direction vector of the light source, which is the illumination direction of the light source, N = (N x N y N z ) T Let be the unit normal vector at a point on the surface of the object to be measured; ρN corresponding to each pixel position is a three-dimensional vector, so a single light source cannot solve equation (5). At least three light sources at different positions are needed to solve for three unknowns. The photometric stereo algorithm uses a photometric stereo method with four light source positions to accurately describe the surface gradient and reflectivity information of the object under test. In four images illuminated by light sources at four different angles, the brightness of a pixel is I1, I2, I3, and I4, respectively. The unit direction vectors of the four light sources are L1, L2, L3, and L4, respectively. The reflected light brightness of the pixel is as follows: N and L are unit vectors. Solving equation (6) using the least squares method, we obtain the surface reflectance of the object to be measured as: ρ=‖(L T L) -1 (L T L)‖ (7)。 4. The method according to claim 3, characterized in that: The tone mapping algorithm described uses the Reinhard method.

Images