A short-wave infrared dual-camera calibration device and calibration method

Abstract

The present invention discloses a short-wave infrared dual-camera calibration device, which uses a halogen lamp with relatively stable light intensity as a short-wave infrared light source, designs an elliptical cylindrical reflector, and places the lamp at the focus of the ellipse to achieve relatively uniform line illumination at the far end. The elliptical cylindrical reflector is made of aluminum, and a heat sink and a fan are designed on the back to solve the problem of large amounts of heat generated by the infrared light source. In addition, in order to reduce the temperature rise of the target after heat absorption, white alumina is used as the pattern background of the target to reduce heat absorption, and the target uses an aluminum substrate to facilitate heat conduction. The present invention can achieve precise focusing and sub-pixel alignment of dual cameras. It effectively solves the problem of high-brightness, high-uniformity strip lighting light sources faced by infrared line array camera calibration, and takes into account the heat dissipation problem therein; designs a sub-pixel quantitative alignment method for line array cameras based on customized infrared targets, thereby solving the problems of high-precision correction and quantitative adjustment.

Description

Technical Field

[0001] The present invention belongs to the technical field of camera calibration and registration, and particularly relates to a short-wave infrared dual-camera calibration method and device. Background Art

[0002] Line scan cameras image objects using a linear pattern, offering high scanning frequencies and relatively low cost, making them widely used in industrial production. Currently, a growing number of quality control processes in high-speed industrial production require dual-band or even multispectral imaging. This is achieved primarily through two approaches on the camera and sensor side: multi-line cameras, where the sensor has multiple rows of detectors or pixels sensitive to different spectra; and prism-based beam splitting, where the light is received by multiple single-line cameras. The former offers low cost and convenient assembly and adjustment, but the sensor's multiple rows of detectors simultaneously image different locations on the object. Image quality is affected by nonuniform motion speeds and variations in image quality across the optical system's field of view. These effects cannot be fully compensated for through software, making it unsuitable for applications requiring high sensitivity and high resolution. The latter, on the other hand, features multiple single-line cameras simultaneously imaging the same location on the object. Consequently, even with nonuniform motion speeds, the pixels of each sensor maintain a strict correspondence, benefiting both imaging and spectral measurement, making it the preferred solution for high-end online spectral measurement equipment. The most common prism-based beamsplitting solution is the dual-line array camera system. This system uses a dichroic prism to split the incoming light into two desired wavelengths, which are then received by two separate sensors. The goal of calibration is to ensure that the two cameras image the same area at the same time, effectively achieving sub-pixel alignment or registration of the two cameras. Furthermore, the increasing application of near-infrared, short-wave infrared, and even medium-wave infrared line array cameras in recent years has placed higher demands on the light sources, targets, and dynamic imaging systems required for dual-camera calibration. Existing camera calibration methods primarily fall into two categories: one involves establishing coordinate transformation matrices between multiple cameras and targets for visual measurement; the other involves multi-camera image registration based on image processing techniques, which can also achieve sub-pixel accuracy. These two calibration methods primarily target the visible light band and place low demands on the camera's own alignment accuracy. Even using image processing and coordinate calculation methods, they cannot compensate for image mismatches and resolution degradation caused by uneven motion speeds.

[0003] For high-precision, high-resolution applications such as material type detection, surface defect detection, and infrared temperature field measurement, alignment accuracy can even be required to be less than 0.1 pixel, making traditional methods inadequate. To meet such high-precision alignment requirements, developing targeted, high-precision calibration schemes and designing specialized calibration devices are key challenges. This is especially true for shortwave infrared (0.9-2.5μm) dual-camera calibration, as the design of the infrared light source and infrared target (used for calibration) also pose challenges.

[0004] Prior art related to the present invention

[0005] The technical difficulties in infrared dual-camera calibration mainly include three aspects: infrared light source and target design, sub-pixel alignment of linear array cameras, and dynamic imaging system design.

[0006] Existing calibration technologies are mainly aimed at visible light image sensors, and the light source is mainly visible light LED. Multiple light-emitting LEDs are arranged in an array, and after reflection or refraction, they form a long strip lighting area to achieve illumination of the imaging area of ​​the linear array camera. Infrared calibration requires a short-wave infrared light source with a large heat output, such as a high-temperature black body, a heating wire, an infrared LED, a halogen lamp, etc. Therefore, the lens-type focusing solution used in the visible light system is no longer applicable. Halogen lamps are commonly used short-wave infrared light sources. Due to their high heat output and poor energy uniformity, it is relatively difficult to achieve uniform long strip lighting. The present invention proposes a strip lighting method based on the principle of elliptical cylinder reflection. Analysis and testing show that the light source has good uniformity. The target is a pattern used for camera alignment. Existing targets are usually checkerboards. The black illuminated area of ​​the target is easily heated by the short-wave infrared light source, resulting in damage or deformation. Therefore, the traditional checkerboard target is no longer applicable.

[0007] Some existing sub-pixel alignment technologies for linear array cameras rely on static photography methods. These methods use a camera to photograph a static grid or point target, generating a one-dimensional grayscale distribution curve. Aligning the distribution curves of the two cameras achieves registration. However, this static registration technique has an alignment accuracy of approximately 0.5 pixels, which cannot meet the requirements of some high-precision, high-resolution applications. Other technologies, such as patent CN10623234A, employ a moving target for linear array CCD alignment. This technology can also be used for dual-camera alignment, but it fails to achieve quantitative and visual adjustment of the two-way focus accuracy, angle, and offset. This method is complex to operate, has limited alignment accuracy, and cannot meet the requirements of infrared dual-camera calibration.

[0008] In summary, traditional methods still have shortcomings in terms of the influence of infrared light sources, comprehensive infrared target design, and dynamic registration methods, and are unable to achieve high-precision and convenient dual-camera calibration. Summary of the Invention

[0009] The purpose of the present invention is to solve the above-mentioned main problems and provide an infrared dual-camera calibration method and device. The method adopts a customized short-wave infrared illumination light source, an infrared target and a linear array camera alignment method based on dynamic imaging. This method expands the traditional dual-camera calibration method to the short-wave infrared band, which can effectively solve the multi-degree-of-freedom dual-camera alignment problem. The designed dynamic imaging device can realize visualization, quantification, sub-pixel accuracy, efficient and convenient infrared dual-camera calibration.

[0010] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is:

[0011] A short-wave infrared dual-camera calibration device includes a passive vibration isolation optical platform, on which a system support frame is fixedly mounted. The system support frame comprises a bottom panel, a top panel, two side panels, and a back panel. The bottom panel is mounted with a load-carrying motion platform, on which an infrared target is provided. The lower surface of the top panel is mounted with an infrared light source system. Two short-wave infrared cameras are mounted on the top panel, each infrared camera being provided with a corresponding beam splitter prism and a lens.

[0012] in,

[0013] The infrared light source system includes a halogen lamp tube, which is located at one focus of an elliptical cylindrical reflector made of stainless steel, and the other focus of the elliptical cylindrical reflector is on the infrared target, so as to achieve uniform long strip lighting;

[0014] The beam splitter prism receives the infrared light from the infrared light source system reflected by the infrared target, and forms images on two short-wave infrared cameras after energy splitting;

[0015] It also includes an image acquisition module that realizes image cropping and registration operations. The image acquisition module includes image acquisition cards respectively set on two short-wave infrared cameras, and the image acquisition cards are connected to the controller of the object motion stage; the two image acquisition cards are connected to the two cameras and the main computer, receive image data from the cameras, and send images to the main computer at the same time.

[0016] The customized infrared target substrate is made of aluminum alloy and moves back and forth along the X direction driven by the object moving platform, wherein the uniform motion segment images the target.

[0017] The infrared target includes edge patterns, vertical grid lines, horizontal grid lines, dense vertical stripes, and dark blocks on the edges; wherein,

[0018] The white part of the infrared target substrate is a diffuse reflective aluminum oxide film, and the black part is a light-absorbing coating;

[0019] There are four blade edge patterns. After infrared illumination, black represents the dark area and white represents the bright area. After collecting and processing the images, the ESF of the center and edge fields of view can be obtained. The number of pixels occupied by the rising edge of the ESF curve from 10% intensity to 90% intensity is used as the ESF value, which is used to quantitatively evaluate the overall defocus amount and local defocus amount of the field of view. When the ESF of the center of the field of view meets the image quality requirements and the difference between the edge ESF and the center is small, it indicates that the entire field of view is in focus.

[0020] The vertical grid lines are unevenly distributed within the field of view, with the distribution of the vertical grid lines at the center being sparser than at the edges. The vertical grid lines are used to assess system distortion and are also used to align the two cameras in the pixel arrangement direction. Sub-pixel camera alignment is achieved through an edge recognition algorithm and a precision adjustment device.

[0021] The horizontal grid lines are evenly distributed and are used for aligning the two cameras perpendicular to the pixel arrangement direction, using the same method as the pixel arrangement direction alignment;

[0022] The dense vertical stripes are used for coarse focus adjustment of the linear array camera when it is static. When the stripes are dense, the grayscale peak-to-valley values ​​are low when the camera is out of focus, and high when the sensor is at the focal plane.

[0023] The edge dark block is used as a reference dark background. By adjusting the gain and bias of the two cameras, the brightness of the images of the black area and the brightness of the images of the white area are equal.

[0024] A further preferred technical solution provided by the present invention is that the image cropping of the image acquisition module is as follows: the dual cameras are connected to two image acquisition cards and a load carrier motion stage controller; when the load carrier motion stage moves to a certain position, a trigger signal is sent to the two cameras, and image acquisition starts at the same time, and is transmitted to the main computer through the image acquisition card; after sufficient target images are acquired, the acquisition of local images is stopped at the same time, i.e., image cropping.

[0025] A further preferred technical solution provided by the present invention is that the image registration operation of the image acquisition module is that the main computer performs image registration operation, calculates the ESF value, the Z-axis rotation value of the infrared target, and the XY-axis offset value of the infrared target, and outputs the value together with the image to the display screen.

[0026] A further preferred technical solution provided by the present invention is that each infrared camera detector window is installed with a narrow bandpass filter, with central wavelengths of 1.2 μm and 1.55 μm respectively and a bandwidth of 30 nm.

[0027] The present invention also provides a calibration method for a short-wave infrared dual-camera calibration device.

[0028] The following steps are involved:

[0029] Step S1, static coarse adjustment;

[0030] Step S10: Power on the camera, turn on the halogen lamp of the infrared light source system, and power on the object motion stage;

[0031] Step S11, controlling the object motion stage to move the infrared target to the dual-camera imaging area;

[0032] Step S12, controlling the camera to statically image the dense vertical stripes of the infrared target, and coarsely adjusting the camera until the grayscale peak-to-valley value in the center area is the maximum;

[0033] Step S13, image plane tilt adjustment;

[0034] Step S14: Check whether the two edges of the linear array single-row image are sharp. If not, return to step S13; if so, proceed to the next step S15;

[0035] Step S15, coarsely adjust the Z-axis rotation, X-axis offset, and Y-axis offset, and iterate the adjustments repeatedly;

[0036] Step S16, determining whether the alignment accuracy reaches the pixel level, if so, the coarse adjustment ends, if not, returning to step S15;

[0037] Step S2, dynamic fine-tuning;

[0038] Step S20, setting the motion parameters of the carrier motion stage, including motion range, speed, and acceleration;

[0039] Step S21, controlling the object-carrying motion stage to move back and forth, and collecting camera images in real time;

[0040] Step S22, cutting the image at each round trip time, displaying the target image on the screen, and calculating and outputting the offset and image angle difference between the two camera images in the X and Y directions in real time;

[0041] Step S23, calculating the center and edge ESF using the target edge image, and displaying the ESF values ​​on the screen;

[0042] Step S24, fine-tuning the Z-axis translation and X-axis rotation to make the image center and edges clear and the ESF difference small;

[0043] Step S25, fine-tuning the Z-axis rotation, X-axis offset, and Y-axis offset, and iterating the adjustments repeatedly;

[0044] Step S26, determining whether the alignment accuracy reaches the sub-pixel level, if so, the fine adjustment ends, if not, returning to step S25.

[0045] Preferably, in step S22, the specific process of calculating and outputting the offset and image angle difference of the two camera images in the X and Y directions in real time is: taking one of the cameras as a reference, calculating the offset and image angle difference between the two cameras in the X and Y directions through an image registration algorithm, and the calculation method is to use an edge recognition algorithm such as a Sobel edge detection operator to identify the corner points in the two camera images, and calculate the offset and angle difference based on the relative positions of multiple corner points in the two images.

[0046] Preferably, in step S23, the specific process of calculating the center and edge ESF using the target edge image is as follows:

[0047] There are four blade-edge patterns. Under infrared illumination, black represents the dark area and white represents the bright area. The ESF is obtained from the blade-edge image by selecting a transition edge between black and white, obtaining the grayscale distribution of the transition area, and plotting a curve with coordinate position as the horizontal axis and grayscale as the vertical axis. The mean of the white area is used as the baseline intensity. The ESF curve is processed, and the number of pixels occupied by the rising edge of the ESF curve from 10% to 90% intensity is used as the ESF value. The ESF values ​​of the four blade-edge patterns from left to right in the image represent the clarity of the edge, slightly left of center, slightly right of center, and edge, respectively.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] Infrared light sources emit shortwave infrared light within the infrared camera's response band. Traditional infrared light sources, such as LEDs, suffer from low power, insufficient stability, and difficulty achieving uniform line illumination through beam shaping, making them unsuitable for calibrating infrared dual cameras. This invention utilizes a customized infrared halogen lamp with stable power and high shortwave radiance, placed at one focal point of a stainless steel ellipsoidal cylindrical reflector. At the other focal point, the light energy is concentrated into a narrow rectangular area, achieving the strip illumination required by line array cameras.

[0050] This invention enables precise focusing and sub-pixel alignment of dual cameras. It effectively addresses the challenges of high-brightness, high-uniformity strip lighting sources faced by infrared line array camera calibration, while also taking into account heat dissipation issues. A sub-pixel quantitative alignment method for line array cameras based on a customized infrared target is designed, further addressing the challenges of high-precision calibration and quantitative adjustment. The device of this invention not only achieves infrared dual-camera alignment with an accuracy of up to 0.1 pixel, but also quantifies and visualizes the camera calibration process, bringing a certain degree of convenience. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the infrared target structure of the present invention;

[0052] Figure 2 Schematic diagram of the coordinate system of the short-wave infrared dual-camera calibration system of the present invention;

[0053] Figure 3 This is a flow chart of the dynamic coarse adjustment of the short-wave infrared dual camera of the present invention;

[0054] Figure 4 This is a flow chart of the dynamic fine-tuning of the short-wave infrared dual camera of the present invention;

[0055] Figure 5 This is a schematic diagram of the overall structure of the short-wave infrared dual-camera calibration device of the present invention. DETAILED DESCRIPTION

[0056] In order to more clearly explain the short-wave infrared dual-camera calibration device and calibration method provided by the present invention, they are now described in detail with reference to specific embodiments and the accompanying drawings.

[0057] like Figure 5 As shown, a short-wave infrared dual-camera calibration device includes a passive vibration isolation optical platform 101, on which a system support frame 102 is fixedly mounted. The system support frame 102 has a bottom panel, a top panel, two side panels, and a back panel. A load motion platform 107 is mounted on the bottom panel, and an infrared target 106 is provided on the load motion platform 107; an infrared light source system 103 is mounted on the lower surface of the top panel; two short-wave infrared cameras 105 are mounted on the top panel, and each infrared camera 105 is correspondingly provided with a beam splitter prism 104 and a lens; each infrared camera detector window is installed with a narrow bandpass filter, with center wavelengths of 1.2μm and 1.55μm, respectively, and a bandwidth of 30nm.

[0058] The infrared light source system 102 includes a halogen lamp tube, which is located at one focus of an elliptical cylindrical reflector made of stainless steel, and the other focus of the elliptical cylindrical reflector is on the infrared target 106, so as to achieve uniform long strip lighting;

[0059] Infrared light sources emit shortwave infrared light within the infrared camera's response band. Traditional infrared light sources, such as LEDs, suffer from low power, insufficient stability, and difficulty achieving uniform line illumination through beam shaping, making them unsuitable for calibrating infrared dual cameras. This invention utilizes a customized infrared halogen lamp with stable power and high shortwave radiance, placed at one focal point of a stainless steel ellipsoidal cylindrical reflector. At the other focal point, the light energy is concentrated into a narrow rectangular area, achieving the strip illumination required by line array cameras.

[0060] The infrared target 106 of the present invention is a customized infrared target, which moves back and forth along the X direction driven by the object moving platform, wherein the uniform speed motion segment images the target, such as Figure 1 As shown, the infrared target 106 of the present invention includes a blade edge pattern 1, vertical grid lines 2, horizontal grid lines 3, dense vertical stripes 4, and edge dark blocks 5. The characteristic of the infrared target 106 is that the base material is an aluminum alloy with good thermal conductivity. Figure 1 The white part in the middle is the diffuse reflective aluminum oxide film, and the black part is the light-absorbing coating. Figure 1Black represents dark areas, and white represents bright areas. Image acquisition and processing yield the ESF at the center and edges of the field of view. The number of pixels occupied by the rising edge of the ESF curve from 10% to 90% intensity is used as the ESF value, which is used to quantitatively assess overall and local defocus in the field of view. When the center ESF meets image quality requirements and the ESF at the edges differs little from the center, the entire field of view is in focus. Vertical grid lines 2 are unevenly distributed across the field of view (sparser at the center than at the edges), facilitating the assessment of system distortion. This grid is also used for dual-camera alignment in the pixel arrangement direction (corresponding to the non-moving object direction during dynamic imaging). Sub-pixel camera alignment is achieved through an edge recognition algorithm and precision adjustment. Horizontal grid lines 3 are evenly distributed and are used for dual-camera alignment perpendicular to the pixel arrangement direction, using the same method as for pixel alignment. Dense vertical stripes 4 are used for coarse focus adjustment when the line scan camera is static. When the stripes are dense, the grayscale peaks and valleys are low (unclear) if the image is out of focus. When the sensor is in focus (focused), the grayscale peaks and valleys are high (sharp and clear). This novel design offers the advantage of conveniently testing image quality in both the center and edge fields of view under static imaging conditions. The edge dark patches 5 serve as a reference dark background. By adjusting the gain and offset of the two cameras, the brightness of the images captured by both cameras is equal, as is the brightness of the images captured by both cameras for white areas. Using the edge dark patches 5 as a reference dark background when evaluating image quality using dense vertical stripes facilitates determining whether the line scan camera is out of focus.

[0061] The spectroscopic prism infrared target 104 receives the infrared light of the infrared light source system infrared target 103 reflected by the infrared target 106, and forms images on two short-wave infrared camera infrared targets 105 after energy splitting;

[0062] It also includes an image acquisition module that realizes image cropping and registration operations. The image acquisition module includes image acquisition cards respectively set on two short-wave infrared cameras, and the image acquisition cards are connected to the controller of the object motion stage; the two image acquisition cards are connected to the two cameras and the main computer, receive image data from the cameras, and send images to the main computer at the same time.

[0063] The short-wave infrared dual-camera calibration device of the present invention can achieve dynamic imaging. The purpose of dynamic imaging is to enable the line array camera to obtain a stable two-dimensional image of a fixed target area, simulating the imaging effect of an area array camera. Compared with the traditional static single-line alignment method of line array cameras, its advantage is that it can achieve sub-pixel precision and visual alignment of infrared dual cameras, and its convenience and accuracy are significantly improved.

[0064] The short-wave infrared dual-camera calibration device mainly realizes the acquisition and transmission of dynamic images as well as image processing and display, including cropping of two camera images, image offset and tilt angle calculation, real-time image display, etc. It also has a local zoom function to verify the camera alignment effect in real time. Figure 2 The coordinate system 6 shown in the figure has the center pixel 7 of one of the camera sensors as its origin, with the pixel arrangement direction being X and the optical axis direction being Z. The degrees of freedom required for dual-camera calibration include X, Y, and Z translation and X, Y, and Z rotation. Z translation is for focus adjustment, Y rotation is for image plane tilt, Z rotation represents the angle between the two camera sensors, and X / Y translation achieves pixel alignment for the two cameras.

[0065] The dynamic imaging method process for calibrating short-wave infrared dual cameras of the present invention includes two parts: static coarse adjustment and dynamic fine adjustment. The static coarse adjustment process is as follows: Figure 3 As shown, under static imaging conditions, each camera forms only one row of images, which is expressed as a line. Figure 1 The dense vertical stripes 4 are used for coarse focus adjustment, and the vertical and horizontal grid lines are used to adjust the Z rotation, X offset and Y offset at the pixel level to prepare for further fine adjustment.

[0066] The specific steps of static coarse adjustment are as follows:

[0067] Step S1, static coarse adjustment;

[0068] Step S10: Power on the camera, turn on the halogen lamp of the infrared light source system, and power on the object motion stage;

[0069] Step S11, controlling the object motion stage to move the infrared target to the dual-camera imaging area;

[0070] Step S12, controlling the camera to statically image the dense vertical stripes of the infrared target, and coarsely adjusting the camera until the grayscale peak-to-valley value in the center area is the maximum;

[0071] Step S13, image plane tilt adjustment;

[0072] Step S14: Check whether the two edges of the linear array single-row image are sharp. If not, return to step S13; if so, proceed to the next step S15;

[0073] Step S15, coarsely adjust the Z-axis rotation, X-axis offset, and Y-axis offset, and iterate the adjustments repeatedly;

[0074] Step S16, determining whether the alignment accuracy reaches the pixel level, if so, the coarse adjustment ends, if not, returning to step S15;

[0075] After rough adjustment, dynamic fine adjustment is performed. The process is as follows: Figure 4As shown in the figure, its characteristic is that a relatively complete target image is obtained, and the difference in imaging clarity and position of the dual cameras can be directly obtained through visual observation and image processing.

[0076] The specific steps of dynamic fine-tuning are as follows:

[0077] Step S2, dynamic fine-tuning;

[0078] Step S20, setting the motion parameters of the carrier motion stage, including motion range, speed, and acceleration;

[0079] Step S21, controlling the object-carrying motion stage to move back and forth, and collecting camera images in real time;

[0080] Step S22, cutting the image at each round trip time, displaying the target image on the screen, and calculating and outputting the offset and image angle difference between the two camera images in the X and Y directions in real time;

[0081] In step S22, the specific process of real-time calculation and output of the offset and image angle difference between the two camera images in the X and Y directions is as follows: using one of the cameras as a reference, the offset and image angle difference between the two cameras in the X and Y directions are calculated through an image registration algorithm. The calculation method is to use an edge recognition algorithm such as the Sobel edge detection operator to identify the corner points in the two camera images, and calculate the offset and angle difference based on the relative positions of multiple corner points in the two images.

[0082] Step S23, calculating the center and edge ESF using the target edge image, and displaying the ESF values ​​on the screen;

[0083] In step S23, the specific process for calculating the center and edge ESF using the target blade edge image is as follows: there are four blade edge patterns, with black representing dark areas and white representing bright areas after infrared illumination. The ESF is obtained from the blade edge image by selecting a black-to-white transition edge, obtaining the grayscale distribution of the transition area, and plotting a curve with coordinate position as the horizontal axis and grayscale as the vertical axis. The grayscale distribution curve (ESF curve) is then processed using the mean of the white area as the reference intensity. The number of pixels occupied by the rising edge of the ESF curve from 10% to 90% intensity is used as the ESF value. The ESF values ​​of the four blade edge patterns from left to right in the image represent the clarity of the edge, slightly left of the center, slightly right of the center, and edge areas, respectively.

[0084] Step S24, fine-tuning the Z-axis translation and X-axis rotation to make the image center and edges clear and the ESF difference small;

[0085] Step S25, fine-tuning the Z-axis rotation, X-axis offset, and Y-axis offset, and iterating the adjustments repeatedly;

[0086] Step S26, determining whether the alignment accuracy reaches the sub-pixel level, if so, the fine adjustment ends, if not, returning to step S25.

[0087] The key parts of dynamic fine-tuning are image segmentation and the calculation of various parameters: (1) The image obtained is a superposition of the two cameras and is displayed in RGB. The two images are red and green respectively. The angle and position differences between the two images can be directly observed from the image. Through image position recognition and segmentation, only the target image is displayed on the screen. During fine-tuning, the relative position movement of the image can be directly observed, realizing visual adjustment. (2) The calculation of XY offset parameters is mainly achieved through edge recognition and multi-feature statistics, and Z rotation is achieved through sub-pixel recognition of horizontal grid lines and straight line fitting. After fine-tuning, the X / Y offsets are reduced to the sub-pixel level, and the Z rotation angle is generally less than 1mrad.

[0088] The embodiments described above are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

Claims

1. A short-wave infrared dual-camera calibration device, comprising a passive vibration isolation optical platform, characterized in that: The passive vibration isolation optical platform is fixedly mounted with a system support frame, which includes a bottom panel, a top panel, two side panels, and a back panel. The bottom panel is mounted with a load-carrying motion platform, which is provided with an infrared target. The infrared target includes a knife-edge pattern, vertical grid lines, horizontal grid lines, dense vertical stripes, and dark blocks on the edge. The lower surface of the top panel is mounted with an infrared light source system. Two short-wave infrared cameras are mounted on the top panel, and each infrared camera is provided with a corresponding beam splitter prism and a lens. in, The infrared light source system includes a halogen lamp tube, which is located at one focus of an elliptical cylindrical reflector made of stainless steel, and the other focus of the elliptical cylindrical reflector is on the infrared target, so as to achieve uniform long strip lighting; The beam splitter prism receives the infrared light from the infrared light source system reflected by the infrared target, and forms images on two short-wave infrared cameras after energy splitting; The invention also includes an image acquisition module for performing image cropping and registration operations, wherein the image acquisition module includes image acquisition cards respectively provided on the two short-wave infrared cameras, and the image acquisition cards are connected to the controller of the object movement stage; the two image acquisition cards are connected to the two cameras and the host computer, receive image data from the cameras, and simultaneously send images to the host computer; The image registration operation of the image acquisition module is that the main computer performs image registration operation, calculates the ESF value, the Z rotation value of the infrared target, and the XY offset value of the infrared target, and outputs the value together with the image to the display screen.

2. A shortwave infrared dual camera calibration device according to claim 1, characterized in that: The customized infrared target substrate is made of aluminum alloy and moves back and forth along the X direction driven by the object moving platform, wherein the uniform motion segment images the target.

3. A shortwave infrared dual camera calibration device according to claim 2, characterized in that: The infrared target includes edge patterns, vertical grid lines, horizontal grid lines, dense vertical stripes, and dark blocks on the edges; wherein, The white part of the infrared target substrate is a diffuse reflective aluminum oxide film, and the black part is a light-absorbing coating; There are four blade edge patterns. After infrared illumination, black represents the dark area and white represents the bright area. After collecting and processing the images, the ESF of the center and edge fields of view can be obtained. The number of pixels occupied by the rising edge of the ESF curve from 10% intensity to 90% intensity is used as the ESF value, which is used to quantitatively evaluate the overall defocus amount and local defocus amount of the field of view. When the ESF of the center of the field of view meets the image quality requirements and the difference between the edge ESF and the center is small, it indicates that the entire field of view is in focus. The vertical grid lines are unevenly distributed within the field of view, with the distribution of the vertical grid lines at the center being sparser than at the edges. The vertical grid lines are used to assess system distortion and are also used to align the two cameras in the pixel arrangement direction. Sub-pixel camera alignment is achieved through an edge recognition algorithm and a precision adjustment device. The horizontal grid lines are evenly distributed and are used for aligning the two cameras perpendicular to the pixel arrangement direction, using the same method as the pixel arrangement direction alignment; The dense vertical stripes are used for coarse focus adjustment of the linear array camera when it is static. When the stripes are dense, the grayscale peak-to-valley values ​​are low when the camera is out of focus, and high when the sensor is at the focal plane. The edge dark block is used as a reference dark background. By adjusting the gain and bias of the two cameras, the brightness of the images of the black area and the brightness of the images of the white area are equal.

4. The shortwave infrared dual-camera calibration device according to claim 1, characterized in that: The image cropping of the image acquisition module is as follows: the dual cameras are connected to two image acquisition cards and a load-carrying motion stage controller. When the load-carrying motion stage moves to a certain position, a trigger signal is sent to the two cameras, which start to acquire images at the same time and transmit them to the main computer through the image acquisition card. After acquiring enough target images, the acquisition of local images is stopped at the same time, i.e., image cropping.

5. The shortwave infrared dual-camera calibration device according to claim 1, characterized in that: Each infrared camera detector window is installed with a narrow bandpass filter with a center wavelength of 1.2 μm and 1.55 μm and a bandwidth of 30 nm.

6. A calibration method for a short-wave infrared dual-camera calibration device according to any one of claims 1 to 5, It is characterized in that The following steps are involved: Step S1, static coarse adjustment; Step S10: Power on the camera, turn on the halogen lamp of the infrared light source system, and power on the object motion stage; Step S11, controlling the object motion stage to move the infrared target to the dual-camera imaging area; Step S12, controlling the camera to statically image the dense vertical stripes of the infrared target, and coarsely adjusting the camera until the grayscale peak-to-valley value in the center area is the maximum; Step S13, image plane tilt adjustment; Step S14: Check whether the two edges of the linear array single-row image are sharp. If not, return to step S13; if so, proceed to the next step S15; Step S15, coarsely adjust the Z-axis rotation, X-axis offset, and Y-axis offset, and iterate the adjustments repeatedly; Step S16, determining whether the alignment accuracy reaches the pixel level, if so, the coarse adjustment ends, if not, returning to step S15; Step S2, dynamic fine-tuning; Step S20, setting the motion parameters of the carrier motion stage, including motion range, speed, and acceleration; Step S21, controlling the object-carrying motion stage to move back and forth, and collecting camera images in real time; Step S22, cutting the image at each round trip time, displaying the target image on the screen, and calculating and outputting the offset and image angle difference between the two camera images in the X and Y directions in real time; Step S23, calculating the center and edge ESF using the target edge image, and displaying the ESF values ​​on the screen; Step S24, fine-tuning the Z-axis translation and X-axis rotation to make the image center and edges clear and the ESF difference small; Step S25, fine-tuning the Z-axis rotation, X-axis offset, and Y-axis offset, and iterating the adjustments repeatedly; Step S26, determining whether the alignment accuracy reaches the sub-pixel level, if so, the fine adjustment ends, if not, returning to step S25.

7. The calibration method of a short-wave infrared dual-camera calibration device according to claim 6, characterized in that: In step S22, the specific process of real-time calculation and output of the offset and image angle difference between the two camera images in the X and Y directions is as follows: using one of the cameras as a reference, the offset and image angle difference between the two cameras in the X and Y directions are calculated through an image registration algorithm. The calculation method is to use the edge recognition algorithm of the Sobel edge detection operator to identify the corner points in the two camera images, and calculate the offset and angle difference based on the relative positions of multiple corner points in the two images.

8. The calibration method of a short-wave infrared dual-camera calibration device according to claim 7, characterized in that: In step S23, the specific process of calculating the center and edge ESF using the target edge image is as follows: There are four blade edge patterns in total. After infrared illumination, black represents the dark area and white represents the bright area. The method for obtaining ESF from the blade edge image is to select a black-white transition edge, obtain the grayscale distribution of the transition area, draw a curve with the coordinate position as the horizontal axis and the grayscale as the vertical axis, use the mean of the white area as the reference intensity, process the ESF curve, and take the number of pixels occupied by the rising edge of the ESF curve from 10% intensity to 90% intensity as the ESF value. The ESF values ​​of the four blade edge patterns from left to right in the image represent the clarity of the edge, slightly left of the center, slightly right of the center, and edge area, respectively.

Images