A deep water measuring method and system based on binocular distance measurement

Abstract

The application discloses a kind of deep water measurement method and system based on binocular ranging, method includes the following steps: S1, in air or ideal environment, the color calibration board image of known RGB value is collected by binocular camera, as color reference image;S2, color calibration board is placed at underwater target scene, and the first underwater image of color calibration board is collected by RGBW light supplement lamp cooperates binocular camera;S3, contrast first underwater image and color reference image, calculate the high intensity configuration parameter of RGBW light supplement lamp;S4, RGBW light supplement lamp is adjusted to high intensity configuration parameter, cooperates binocular camera and collects underwater target scene image and the second underwater image of color calibration board;S5, underwater target scene image is preprocessed, and underwater target scene actual image is obtained.The application solves the problem that the measurement accuracy is not enough in the application of existing measurement technology under deep water condition, and can realize the efficient operation of accurate identification and reliable analysis of underwater target.

Description

Technical Field

[0001] This invention relates to the field of marine engineering technology. More specifically, this invention relates to a deep-water measurement method and system based on binocular ranging. Background Technology

[0002] Currently, underwater measurement and positioning technologies mainly include acoustic positioning systems, photogrammetry, inertial navigation systems (INS), laser scanning, and underwater image measurement. Among these, acoustic positioning systems have become the mainstream method for deep-water measurement due to their excellent propagation characteristics in underwater environments. Long baseline (LBL) acoustic positioning systems, because their positioning accuracy is independent of water depth, have been widely used in the installation and measurement of key structures such as crossover pipes in deep-water oil and gas fields. In addition, integrated measurement systems combining inclinometers, GPS, and underwater cameras have also achieved good results in channel improvement projects in water depths of nearly 10 meters, enabling the installation and positioning of components with centimeter-level accuracy.

[0003] However, using acoustic positioning systems for underwater measurement and positioning still faces many challenges:

[0004] 1. Although acoustic measurements have good penetration, their accuracy fluctuates due to factors such as changes in sound velocity profile and signal reflection interference.

[0005] 2. Existing measurement systems are mostly combinations of multiple devices, which have problems such as poor synchronization, accumulation of system errors, and complex operation, making it difficult to meet the real-time and stability requirements of high-precision deep-water installation.

[0006] Therefore, there is an urgent need to develop an integrated, high-precision measurement and installation technology solution that is suitable for complex deep-water environments in order to improve my country's deep-water engineering construction capabilities. Summary of the Invention

[0007] Another objective of this invention is to provide a deep-water measurement method and system based on binocular ranging that improves the accuracy and reliability of underwater measurements and integrates high precision.

[0008] To achieve these objectives and other advantages according to the present invention, a deep-water measurement method based on binocular ranging is provided, comprising the following steps:

[0009] S1. Calibrate the binocular camera; in air or an ideal environment, acquire a color calibration plate image with known RGB values ​​using the binocular camera, and use it as a color reference image.

[0010] S2. Place the color calibration board at the underwater target scene and use an RGBW fill light in conjunction with a binocular camera to acquire the first underwater image of the color calibration board;

[0011] S3. Compare the color difference between the first underwater image of the color calibration board and the color reference image, and calculate the high-intensity configuration parameters of the RGBW supplementary light;

[0012] S4. Adjust the configuration parameters of the RGBW fill light to the high intensity configuration parameters, and coordinate with the binocular camera to acquire underwater target scene images and the second underwater image of the color calibration board;

[0013] S5. Based on the second underwater image and color reference image of the color calibration plate, the underwater target scene image is preprocessed to obtain the actual underwater target scene image.

[0014] Preferably, step S5 is followed by:

[0015] Step S6: Optically enhance the actual image of the underwater target scene using the multi-frame fusion underwater optical image enhancement method to obtain a true image of the underwater target scene.

[0016] Preferably, step S3 specifically includes:

[0017] S31. Define the color vector of the i-th color patch in the color reference image as:

[0018] ;

[0019] In the formula, i = 1, 2, 3, ..., n, where n is the total number of color blocks on the color calibration plate; These are the red, green, and blue components of the i-th color patch in the color reference image, respectively.

[0020] The color vector corresponding to the i-th color patch in the first underwater image is defined as:

[0021] ;

[0022] In the formula, P represents the control parameters of the RGBW fill light. ; These are the red, green, and blue components of the i-th color patch in the first underwater image, respectively;

[0023] S32. Based on the color of the i-th color patch in the color reference image and the color of the corresponding color patch in the first underwater image, calculate the color difference value according to the color difference objective function, which is:

[0024] ;

[0025] S33. Using the current configuration parameters of the RGBW fill light as the initial value of the control parameter P, iteratively update the control parameter P using the gradient descent formula until the value of E(P) is less than a preset threshold or the number of iterations reaches a preset number, thus obtaining the high-intensity configuration parameters. The gradient descent formula is:

[0026] ;

[0027] In the formula, α is the learning rate, and 0.01 ≤ α ≤ 0.1;

[0028] in, , k≤ ≤k+1.

[0029] Preferably, step S5 specifically includes calculating a color correction matrix based on the second underwater image and the color reference image of the color calibration plate, performing color restoration on the underwater target scene image based on the color correction matrix, and performing descattering processing on the color-restored underwater target scene image to obtain the actual underwater target scene image.

[0030] Preferably, the method for calculating a color correction matrix based on the second underwater image and the color reference image using a color calibration plate, and then restoring the color of the underwater target scene image based on the color correction matrix, is as follows:

[0031] F1. Define the color vector of the i-th color patch in the second underwater image as:

[0032] ;

[0033] In the formula, These are the red, green, and blue components of the i-th color patch in the second underwater image, respectively.

[0034] F2. Construct color matrix A and color matrix B based on the color vector of the second underwater image and the color vector of the color reference graphic, respectively;

[0035] ;

[0036] F3. Based on color matrix A and color matrix B, construct color correction matrix M using the least squares method;

[0037] ;

[0038] ;

[0039] In the formula, X is a candidate solution to the correction matrix M;

[0040] F4. Based on the color correction matrix M, the RGB color values ​​of the target scene image are corrected using the color correction formula to restore the color of the target scene image. The expression of the color correction formula is:

[0041] ;

[0042] In the formula, , which is the RGB color value of the original target scene image at pixel coordinates (x,y); , which is the RGB color value of the target scene image at pixel coordinates (x,y) after color restoration.

[0043] Preferably, the method for obtaining the actual image of the underwater target scene by descattering the color-restored underwater target scene image is as follows:

[0044] A. Correct the RGB color values ​​of the second underwater image using a color correction formula to obtain the RGB color values ​​of the restored second underwater image. Extract the water body region excluding the color calibration plate from the second underwater image, calculate the RGB mean value of this region, and obtain the background scattered light intensity C; based on the background scattered light intensity C, and the RGB color values ​​of the color reference map Calculate the transmittance of the color calibration plate. The calculation formula is as follows:

[0045] ;

[0046] B. Measure the distance di of each color patch on the color calibration board from the binocular camera using a binocular camera. Based on the background scattered light intensity C and di, calculate the attenuation coefficient β using the following formula:

[0047] ;

[0048] C. Measure the distance d(x) of each color block in the underwater target scene image from the binocular camera. Based on d(x) and the attenuation coefficient β, calculate the transmittance t(x) of the underwater target scene using the following formula:

[0049] t(x)=exp(- d(x));

[0050] D. Based on the transmittance t(x) and the background scattered light intensity C, and using the simplified underwater Jaffe-McGlamery model, establish the degradation relationship of the target scene image at pixel coordinate x:

[0051] ;

[0052] In the formula, J(x) is the descattered underwater target scene image; This represents the RGB color value at pixel coordinate x in the color-restored target scene image.

[0053] Among them, the descattered underwater target scene image is the actual underwater target scene image.

[0054] Preferably, the actual underwater target scene image includes a sequence of multiple frames of the actual underwater target scene image; step S6 specifically includes:

[0055] S61. Preprocess the multi-frame underwater target scene image sequence; select one frame from the image sequence as the reference image, extract multiple key feature points on the reference image, and obtain the pixel coordinates of each key feature point. and feature descriptors;

[0056] S62. Based on the feature descriptor, sequentially obtain the corresponding points of multiple key feature points of the reference image in the remaining frame image sequence; based on the key feature points and corresponding points, calculate the displacement between the remaining frame image sequence and the reference image. ,based on , The remaining frame image sequence is then registered with the reference image using a registration constraint formula, which is:

[0057] ;

[0058] In the formula, A sequence of adjacent frame images. underwater noise;

[0059] S63. Obtain the reflected light of the underwater target scene region contained in the actual image sequence of the underwater target scene after registration, and obtain the scattered light of the water region in the image sequence. The scattered light includes forward scattered light and back scattered light. Based on the physical difference between reflected light and scattered light, establish an underwater optical imaging physical model. Estimate the forward scattering component and back scattered component of each frame image sequence through the underwater optical imaging physical model.

[0060] S64. Separate the forward scattering component and the backscattering component from the multi-frame underwater target scene actual image sequence to obtain the multi-frame underwater target scene actual image sequence with scattering correction.

[0061] S65. Based on the feature information of the actual image sequence of underwater target scene with multi-frame scattering correction, construct an adaptive weight map. Based on the adaptive weight map, fuse the registered multi-frame descattered image sequence by weighted average to obtain a preliminarily enhanced actual image of underwater target scene.

[0062] S66. Based on the underwater light attenuation characteristics, post-process the preliminarily enhanced actual image of the underwater target scene to obtain a true image of the underwater target scene.

[0063] A deep-water measurement system based on binocular ranging includes:

[0064] The integrated device body is hinged to one end of a robotic arm, and the other end of the robotic arm is hinged to a calibration board mainboard. One side of the calibration board mainboard is a checkerboard pattern, and the other side is a color calibration board. The integrated device body also integrates a binocular camera, an RGBW fill light, and a white light fill light.

[0065] The control and processing module is electrically connected to the binocular camera, robotic arm, and RGBW fill light, and is used to control the operation of the binocular camera, robotic arm, RGBW fill light, and white light fill light.

[0066] The present invention has at least the following beneficial effects:

[0067] 1. The underwater binocular measurement system proposed in this invention solves the problems of insufficient measurement accuracy caused by the inability to calibrate and actively compensate for weak light / color shift in the application of existing binocular measurement technology in deep water conditions by using a robotic arm and RGBW full-color supplementary light. Thus, it can clearly determine the true image of underwater targets, eliminate underwater environmental interference, obtain accurate information about targets, and ultimately achieve efficient operation of accurate identification and reliable analysis of underwater targets.

[0068] 2. This invention compensates for the selective absorption of water directly from the light source by adjusting the RGBW light source, improving the quality of the original image data and providing better input for subsequent processing. It can adapt to optical environments with different waters, depths, and turbidities. It combines supplementary light correction and software descattering processing, and solves the two major problems of color distortion and contrast reduction, thus comprehensively improving the underwater imaging quality.

[0069] 3. By using multi-frame fusion technology, random noise is suppressed in principle, the dynamic range is effectively expanded and color attenuation is compensated, and high-quality images with realistic colors and good preservation of details in both light and dark areas are output. This provides a more accurate and reliable data source for subsequent applications such as underwater photogrammetry, structured light 3D reconstruction, target recognition and classification, and significantly improves the accuracy and reliability of underwater measurements.

[0070] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0071] Figure 1 This is a flowchart illustrating the preprocessing of underwater target scene images in the deep-water measurement method based on binocular ranging according to the present invention.

[0072] Figure 2 This is a flowchart illustrating the optical enhancement of an actual underwater target scene image using the binocular ranging-based deep-water measurement method of the present invention.

[0073] Figure 3 This is a flowchart illustrating the calibration of a binocular camera in this invention;

[0074] Figure 4 This is a schematic diagram of the deep-water measurement system based on binocular ranging according to the present invention;

[0075] Explanation of reference numerals on the accompanying drawings:

[0076] 1. Left camera; 2. Right camera; 3. White light fill light; 4. RGBW fill light; 5. Robotic arm; 6. Calibration board motherboard. Detailed Implementation

[0077] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0078] It should be noted that in the description of this invention, the terms "lateral", "longitudinal", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0079] like Figure 1-3 As shown, this invention provides a deep-water measurement method based on binocular ranging, suitable for precision measurement scenarios such as the installation and inspection of underwater structures in deep water conditions. Specifically, it includes the following steps:

[0080] S1. Calibrate the binocular camera using a checkerboard pattern; in air or an ideal environment, acquire a color calibration board image with known RGB values ​​using the binocular camera as a color reference image; the ideal environment is one without interference and with stable optical characteristics.

[0081] S2. Place the color calibration board at the depth of the underwater target scene. In the underwater working environment, control the RGBW supplement light to illuminate with the initial configuration parameters to cooperate with the binocular camera to acquire the first underwater image of the color calibration board.

[0082] S3. Compare the color difference between the first underwater image and the color reference image on the color calibration board, analyze the attenuation of light in the R, G, and B channels by the water body, and calculate the high-intensity configuration parameters of the RGBW supplementary light with the goal of minimizing the color difference between the first underwater image and the color reference image. The calculation adopts an iterative optimization algorithm, and by fine-tuning the RGBW supplementary light configuration parameters multiple times and evaluating the color difference of the captured images, the high-intensity configuration parameters that minimize the color difference are found. Specifically, this includes the following steps:

[0083] S31. Define the color vector of the i-th color patch in the RGB space as:

[0084] ;

[0085] In the formula, i = 1, 2, 3, ..., n, where n is the total number of color blocks on the color calibration plate; These are the red, green, and blue components of the i-th color patch in the color reference image, respectively.

[0086] The color vector corresponding to the i-th color patch in the first underwater image is defined as:

[0087] ;

[0088] In the formula, P is the normalized control parameter of the RGBW fill light intensity. ; These are the red, green, and blue components of the i-th color patch in the first underwater image, respectively;

[0089] S32. Based on the color of the i-th color patch in the color reference image and the color of the corresponding color patch in the first underwater image, calculate the color difference value according to the color difference objective function. The expression of the color difference objective function is:

[0090] ;

[0091] S33. Using the current configuration parameters of the RGBW fill light as the initial value of the control parameter P, iteratively update the control parameter P using the gradient descent formula until the value of E(P) is less than a preset threshold or the number of iterations reaches a preset number. Then, use the current optimal control parameter P as the high-intensity configuration parameter. The expression for the gradient descent formula is:

[0092] ;

[0093] In the formula, α is the learning rate, used to control the update step size, 0.01≤α≤0.1; ,

[0094] k≤ ≤k+1;

[0095] S4. Adjust the configuration parameters of the RGBW fill light to the high intensity configuration parameters, control the RGBW fill light to illuminate with the high intensity configuration parameters, and once again coordinate with the binocular camera to acquire underwater target scene images and the second underwater image of the color calibration board respectively.

[0096] S5. Based on the second underwater image and the color reference image from the color calibration plate, calculate the color correction matrix. Based on the color correction matrix, perform color restoration on the underwater target scene image. Further, perform descattering processing on the color-restored underwater target scene image to improve its contrast and clarity, obtaining the actual underwater target scene image. This specifically includes the following steps:

[0097] S51. Define the color vector of the i-th color patch in the second underwater image as:

[0098] ;

[0099] In the formula, These are the red, green, and blue components of the i-th color patch in the second underwater image, respectively.

[0100] S52, Color vector based on the second underwater image and color vector of color reference base graphic Construct color matrix A and color matrix B respectively;

[0101] ;

[0102] S53. Based on color matrix A and color matrix B, the color correction matrix M is obtained using the least squares method. The calculation formula is as follows:

[0103] ;

[0104] ;

[0105] In the formula, X is a candidate solution to the color correction matrix M;

[0106] S54. Based on the color correction matrix M, the RGB color values ​​of the target scene image are corrected using the color correction formula to restore the color of the target scene image. The expression of the color correction formula is:

[0107] ;

[0108] In the formula, It is the RGB color value of the original target scene image at pixel coordinates (x, y). , These are the red, green, and blue components of the pixel (x, y), respectively. , which is the RGB color value of the target scene image at pixel coordinates (x,y) after color restoration;

[0109] S55. Correct the RGB color values ​​of the second underwater image using a color correction formula to obtain the RGB color values ​​of the color-restored second underwater image. Extract the water body region excluding the color calibration plate from the second underwater image, calculate the RGB mean value of this region, and obtain the background scattered light intensity C; based on the background scattered light intensity C, and the RGB color values ​​of the color reference map Calculate the transmittance of the color calibration plate. The calculation formula is as follows:

[0110] ;

[0111] S56. The distance di of each color block on the color calibration plate from the binocular camera is measured using a binocular camera. Based on the background scattered light intensity C and di, the attenuation coefficient β is calculated using the least squares method. The calculation formula is as follows:

[0112] ;

[0113] S57. Using a binocular camera, measure the distance d(x) of each color block in the underwater target scene image from the binocular camera. Based on d(x) and the attenuation coefficient β, calculate the transmittance t(x) of the underwater target scene. The calculation formula is as follows:

[0114] t(x)=exp(- d(x));

[0115] S58. Based on the transmittance t(x) and the background scattered light intensity C, and using the simplified underwater Jaffe-McGlamery model, establish the degradation relationship of the target scene image at pixel coordinate x. The degradation relationship expression is as follows:

[0116] ;

[0117] In the formula, J(x) is the descattered underwater target scene image; This represents the RGB color value at pixel coordinate x in the color-restored target scene image.

[0118] Among them, the descattered underwater target scene image is the actual underwater target scene image.

[0119] In the above technical solution, this invention, by incorporating a robotic arm and an RGBW supplementary light, solves the problems of insufficient measurement accuracy caused by the inability to calibrate and actively compensate for weak light / color shift in existing technologies during deep-water applications. This allows for clear determination of the true image of underwater targets, elimination of underwater environmental interference, and acquisition of accurate target information, ultimately achieving efficient operation for precise identification and reliable analysis of underwater targets. By adjusting the RGBW light source, selective absorption by the water is directly compensated from the light source end, improving the quality of the original image data and providing better input for subsequent processing. It can adapt to optical environments of different waters, depths, and turbidities. Simultaneously, by combining supplementary light correction and descattering processing, it solves the two major problems of color distortion and contrast reduction, comprehensively improving underwater imaging quality. It provides an accurate and reliable data source for applications such as underwater photogrammetry, structured light 3D reconstruction, target recognition, and classification, significantly improving the accuracy and reliability of underwater measurements.

[0120] In another technical solution, the actual underwater target scene image includes a sequence of multiple frames of actual underwater target scene images. The image sequence includes, but is not limited to, multiple frames of images under the same exposure parameters used for subsequent noise suppression and detail enhancement, as well as image sequences under different exposure times used for synthesizing high dynamic range images.

[0121] Step S5 is followed by step S6: A registration algorithm combining feature points and optical flow is used to overcome inter-frame misalignment caused by water flow disturbance and carrier motion, precisely aligning all images to the same coordinate system. Multi-frame fusion underwater optical image enhancement is then used to optically enhance the actual underwater target scene image, resulting in a true image of the underwater target scene. This specifically includes the following steps:

[0122] S61. Preprocess the multi-frame underwater target scene image sequence, including dark current correction and optical lens distortion correction, to eliminate the influence of noise and lens optical errors on subsequent registration; select one frame from the image sequence as the reference image, and use the SIFT algorithm to detect and extract multiple key feature points on the reference image, obtaining the pixel coordinates of each key feature point. and feature descriptors;

[0123] S62. Based on the feature descriptor, sequentially obtain the corresponding points of multiple key feature points of the reference image in the remaining frame image sequence; based on the key feature points and corresponding points, use optical flow method for accurate tracking, and calculate the displacement of the key feature points of the remaining frame image sequence relative to the reference image. ,based on , By using the registration constraint formula and minimizing the error of this equation, the global motion parameters of the remaining frame image sequences relative to the reference image sequence are solved sequentially. Using these solved global motion parameters, geometric transformations are performed on the remaining frame image sequences sequentially, registering them to the coordinate system of the reference image sequence. This achieves registration between the remaining frame image sequences and the reference image. The registration constraint formula is as follows:

[0124] ;

[0125] In the formula, A sequence of adjacent frame images. underwater noise;

[0126] S63. Obtain the reflected light of the underwater target scene region contained in the actual image sequence of the underwater target scene after registration, and obtain the scattered light of the water region in the image sequence. The scattered light includes forward scattered light and back scattered light. Based on the physical difference between reflected light and scattered light, establish an underwater optical imaging physical model. Estimate the forward scattering component and back scattered component of each frame image sequence through the underwater optical imaging physical model.

[0127] S64. Separate the forward scattering component and the backscattering component from the multi-frame underwater target scene actual image sequence, and then remove the scattering noise corresponding to the forward scattering component and the backscattering component from the mixed signal of the image sequence to obtain the direct reflection signal carrying the target information, and obtain the multi-frame scattering corrected underwater target scene actual image sequence.

[0128] S65. Based on the actual image sequence of the underwater target scene with multi-frame scattering correction, sharpness measurement, exposure evaluation, and scattering evaluation are performed on each frame of the image to obtain the corresponding sharpness evaluation map, exposure evaluation map, and scattering evaluation map. Sharpness measurement uses local gradient and information entropy of the image as evaluation indicators, and assigns higher weight to areas with rich details. Exposure evaluation is based on pixel brightness value, and assigns higher weight to areas with moderate exposure to avoid overexposed or underexposed areas from participating in fusion. Scattering evaluation is based on the scattering map estimated in step S63, and assigns higher weight to areas with weak scattering effects.

[0129] After normalizing the obtained sharpness evaluation map, exposure evaluation map and scattering evaluation map, an adaptive weight map is constructed. Based on the adaptive weight map, the registered multi-frame descattered image sequence is fused with the corresponding weight map by weighted average to obtain a preliminarily enhanced actual image of the underwater target scene.

[0130] S66. Based on the underwater light attenuation characteristics, post-processing is performed on the preliminarily enhanced actual image of the underwater target scene to obtain a true image of the underwater target scene. The post-processing includes using algorithms such as Rayleigh-Lambert law or gray world hypothesis to compensate and correct severely distorted color channels (especially the red channel) to restore the true color of the object, and applying an adaptive contrast stretching algorithm to further optimize the visual effect of the image.

[0131] In the above technical solution, multi-frame fusion technology suppresses random noise in principle, effectively expands the dynamic range, and compensates for color attenuation, outputting high-quality images with realistic colors and well-preserved details in both light and dark areas. This provides a more accurate and reliable data source for subsequent applications such as underwater photogrammetry, structured light 3D reconstruction, target recognition, and classification, significantly improving the accuracy and reliability of underwater measurements. The purpose of using multi-frame fusion underwater optical image enhancement to optically enhance actual underwater target scene images is to address the problem of drastic changes in measurement accuracy and their impact on measurement results caused by water waves, flow, or changes in light and shadow, thereby improving measurement accuracy and stability. When water turbulence is significant, when using a binocular camera to capture underwater target scene images, multiple frames of underwater target scene images can be acquired. Color correction and descattering are then performed on the underwater target scene images before further optical enhancement. If the water flow conditions are relatively good, with only scattering issues, only one underwater target scene image needs to be acquired, and only color correction and descattering processing are performed on the underwater target scene image.

[0132] The deep-water measurement method based on binocular ranging of this invention utilizes the "adaptive color restoration-supplementary lighting coordination" method to dynamically optimize the RGBW four-channel spectral power distribution based on the real-time water color index, achieving color fidelity and clear details in underwater imaging with strong environmental adaptability. It employs the "multi-frame fusion underwater optical image enhancement" method, which significantly improves the image signal-to-noise ratio and detail preservation through temporal and spatial joint descattering, denoising, and sharpening. Through deep hardware-algorithm coupling, it jointly addresses complex deep-water environments, achieving sub-pixel-level corner point extraction and millimeter-level 3D reconstruction, providing reliable, efficient, and one-click technical support for high-precision deep-water construction and installation.

[0133] A deep-water measurement system based on binocular ranging includes:

[0134] The integrated device body is hinged to one end of a robotic arm, and the other end of the robotic arm is hinged to a calibration board mainboard. One side of the calibration board mainboard is a checkerboard pattern, and the other side is a color calibration board. The integrated device body also integrates a binocular camera, an RGBW fill light, and a white light fill light.

[0135] The control and processing module is electrically connected to the binocular camera, robotic arm, and RGBW fill light, and is used to control the operation of the binocular camera, robotic arm, RGBW fill light, and white light fill light.

[0136] In the above technical solutions, such as Figure 4 As shown, the binocular camera includes a left camera and a right camera. The cameras are global exposure, distortion-free, high-definition color industrial cameras to ensure image quality. A pair of RGBW fill lights are provided, each composed of four high-brightness LEDs (R, G, B, W channels). The brightness of each channel can be independently controlled via a signal, with a brightness control resolution of at least 8 bits. The RGBW fill lights are triggered synchronously with the camera or are always on. A pair of white light fill lights are provided, primarily to provide basic illumination when underwater light is insufficient; any missing colors are supplemented by the RGBW fill lights. The integrated device is housed in a high-density, pressure-resistant housing, encapsulating the binocular camera, white light fill lights, and RGBW fill lights within this housing. The calibration board's main board has a checkerboard pattern on the front, achieving an accuracy of 0.01mm, allowing for binocular camera calibration via corner detection. The reverse side features the X-RiteColorChecker Classic color calibration board. The 24-color chart has known sRGB values ​​for all its color blocks; the robotic arm is a multi-segment, dual-arm structure, with one end hinged to the main body of the integrated device and the other end hinged to the calibration board motherboard. It is used to drive the calibration board motherboard to perform forward, backward, left, right, pitch, yaw, and roll movements along a predetermined trajectory within the binocular field of view under the control of the control and processing module. The control and processing module is connected to the binocular camera, robotic arm, RGBW supplemental light, and white light supplemental light via an umbilical cable. The umbilical cable integrates communication and power supply functions. Its underwater end is connected to the underwater equipment via a watertight connector, and its surface end is connected to the surface processor. The control and processing module is configured to perform tasks such as image color difference comparison, calculation of high-intensity configuration parameters for the RGBW supplemental light, image preprocessing of the underwater target scene, and optical enhancement of the actual image of the underwater target scene in the deep-water measurement method based on binocular ranging. It also coordinates and controls the synchronous operation of the RGBW supplemental light and the binocular camera, as well as the operation of the white light supplemental light. The deep-water measurement system of this invention has an integrated structure, which can realize fully automatic calibration in deep-water environments. It has good integration, stability, synchronization, and simple operation.

[0137] The process of calibrating a stereo camera is as follows:

[0138] 1. Based on the parameters of robotic arm length, calibration plate size, maximum camera angle, and image resolution, the robotic arm's range of motion is pre-set on the shore to ensure that the calibration plate always occupies 1 / 3 to 1 / 2 of the left and right camera images during the calibration process, thus ensuring calibration accuracy.

[0139] Based on the defined range of motion of the robotic arm, the control and processing module is used to set the automatic motion program for the robotic arm, including the forward, backward, left, right, pitch, and rotation movements of the calibration plate. At each distance point, the joints of the robotic arm are controlled to rotate the calibration plate around the X, Y, and Z axes (pitch, yaw, roll), supplemented by translational movements, producing at least 10-15 different orientations. In particular, some large tilt attitudes (such as ±30°, ±45°) should be included to better constrain the lens distortion model.

[0140] 2. Lower the entire system to the target working depth, wait for the system pressure to equalize and the environment to stabilize, and then start the automatic calibration program.

[0141] 3. The control and processing module controls the robotic arm through algorithms, enabling the calibration plate to move in a binocular field of view that fully covers different postures and distances. During the movement, the camera synchronously acquires effective image pairs.

[0142] The entire process should involve acquiring no fewer than 40 valid image pairs. These images should correspond to different distances and viewing angles covering the calibration board, and the checkerboard pattern in the images should be clear and unblurred.

[0143] 4. For all acquired left and right images, automatically locate the checkerboard corner points (internal corner points, i.e., the points where black and white squares intersect). Use a sub-pixel-level optimization algorithm to refine the corner point pixel coordinates, providing high-precision input for subsequent calibration. Automatically remove image pairs that failed extraction or have extremely poor quality. Based on Zhang Zhengyou's calibration method, calculate the camera's intrinsic and extrinsic parameters. After verifying the results, save the final optimized left and right camera intrinsic parameter matrices, distortion coefficient vectors, and the rotation matrix R and translation vector t between the left and right cameras as a configuration file for subsequent underwater visual ranging, 3D reconstruction, and other tasks.

[0144] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A deep-water measurement method based on binocular ranging, characterized in that, Includes the following steps: S1. Calibrate the binocular camera; in air or an ideal environment, acquire a color calibration plate image with known RGB values ​​using the binocular camera, and use it as a color reference image. S2. Place the color calibration board at the underwater target scene and use an RGBW fill light in conjunction with a binocular camera to acquire the first underwater image of the color calibration board; S3. Compare the color difference between the first underwater image on the color calibration board and the color reference image, and calculate the high-intensity configuration parameters of the RGBW supplementary light, which specifically include: S31. Define the color vector of the i-th color patch in the color reference image as: ； In the formula, i = 1, 2, 3, ..., n, where n is the total number of color blocks on the color calibration plate; These are the red, green, and blue components of the i-th color patch in the color reference image, respectively. The color vector corresponding to the i-th color patch in the first underwater image is defined as: ； In the formula, P represents the control parameters of the RGBW fill light. ; These are the red, green, and blue components of the i-th color patch in the first underwater image, respectively; S32. Based on the color of the i-th color patch in the color reference image and the color of the corresponding color patch in the first underwater image, calculate the color difference value according to the color difference objective function, which is: ； S33. Using the current configuration parameters of the RGBW fill light as the initial value of the control parameter P, iteratively update the control parameter P using the gradient descent formula until the value of E(P) is less than a preset threshold or the number of iterations reaches a preset number, thus obtaining the high-intensity configuration parameters. The gradient descent formula is: ； In the formula, α is the learning rate, and 0.01 ≤ α ≤ 0.1; in, , k≤ ≤k+1; S4. Adjust the configuration parameters of the RGBW fill light to the high intensity configuration parameters, and coordinate with the binocular camera to acquire underwater target scene images and the second underwater image of the color calibration board; S5. Based on the second underwater image and the color reference image from the color calibration plate, preprocess the underwater target scene image to obtain the actual underwater target scene image, which specifically includes: Based on the second underwater image and the color reference image of the color calibration plate, a color correction matrix is ​​calculated. Based on the color correction matrix, the underwater target scene image is color restored. The underwater target scene image after color restoration is descattered to obtain the actual underwater target scene image.

2. The deep-water measurement method based on binocular ranging as described in claim 1, characterized in that, Step S5 is followed by: Step S6: Optically enhance the actual image of the underwater target scene using the multi-frame fusion underwater optical image enhancement method to obtain a true image of the underwater target scene.

3. The deep-water measurement method based on binocular ranging as described in claim 1, characterized in that, Based on the second underwater image and the color reference image using a color calibration plate, a color correction matrix is ​​calculated. The method for color restoration of the underwater target scene image based on this color correction matrix is ​​as follows: F1. Define the color vector of the i-th color patch in the second underwater image as: ； In the formula, These are the red, green, and blue components of the i-th color patch in the second underwater image, respectively. F2. Construct color matrix A and color matrix B based on the color vector of the second underwater image and the color vector of the color reference graphic, respectively; ； F3. Based on color matrix A and color matrix B, construct color correction matrix M using the least squares method; ； ； In the formula, X is a candidate solution to the correction matrix M; F4. Based on the color correction matrix M, the RGB color values ​​of the target scene image are corrected using the color correction formula to restore the color of the target scene image. The expression of the color correction formula is: ； In the formula, , which is the RGB color value of the original target scene image at pixel coordinates (x,y); , which is the RGB color value of the target scene image at pixel coordinates (x,y) after color restoration.

4. The deep-water measurement method based on binocular ranging as described in claim 3, characterized in that, The method for descattering the color-restored underwater target scene image to obtain the actual image of the underwater target scene is as follows: A. Correct the RGB color values ​​of the second underwater image using a color correction formula to obtain the RGB color values ​​of the restored second underwater image. ; Extract the water body region excluding the color calibration plate from the second underwater image, calculate the RGB mean value of the region, and obtain the background scattered light intensity C; Based on background scattered light intensity C, and the RGB color values ​​of the color reference map Calculate the transmittance of the color calibration plate. The calculation formula is as follows: ； B. Measure the distance di of each color patch on the color calibration board from the binocular camera using a binocular camera. Based on the background scattered light intensity C and di, calculate the attenuation coefficient β using the following formula: ； C. Measure the distance d(x) of each color block in the underwater target scene image from the binocular camera. Based on d(x) and the attenuation coefficient β, calculate the transmittance t(x) of the underwater target scene using the following formula: t(x)=exp(- d(x)); D. Based on the transmittance t(x) and the background scattered light intensity C, and using the simplified underwater Jaffe-McGlamery model, establish the degradation relationship of the target scene image at pixel coordinate x: ； In the formula, J(x) is the descattered underwater target scene image; This represents the RGB color value at pixel coordinate x in the color-restored target scene image. Among them, the descattered underwater target scene image is the actual underwater target scene image.

5. The deep-water measurement method based on binocular ranging as described in claim 2, characterized in that, The actual underwater target scene image includes a sequence of multiple frames of actual underwater target scene images; step S6 specifically includes: S61. Preprocess the multi-frame underwater target scene image sequence; select one frame from the image sequence as the reference image, extract multiple key feature points on the reference image, and obtain the pixel coordinates of each key feature point. and feature descriptors; S62. Based on the feature descriptor, sequentially obtain the corresponding points of multiple key feature points of the reference image in the remaining frame image sequence; based on the key feature points and corresponding points, calculate the displacement between the remaining frame image sequence and the reference image. ,based on , The remaining frame image sequence is then registered with the reference image using a registration constraint formula, which is: ； In the formula, , A sequence of adjacent frame images. underwater noise; S63. Obtain the reflected light of the underwater target scene region contained in the actual image sequence of the underwater target scene after registration, and obtain the scattered light of the water region in the image sequence. The scattered light includes forward scattered light and back scattered light. Based on the physical difference between reflected light and scattered light, establish an underwater optical imaging physical model. Estimate the forward scattering component and back scattered component of each frame image sequence through the underwater optical imaging physical model. S64. Separate the forward scattering component and the backscattering component from the multi-frame underwater target scene actual image sequence to obtain the multi-frame underwater target scene actual image sequence with scattering correction. S65. Based on the feature information of the actual image sequence of underwater target scene with multi-frame scattering correction, construct an adaptive weight map. Based on the adaptive weight map, fuse the registered multi-frame descattered image sequence by weighted average to obtain a preliminarily enhanced actual image of underwater target scene. S66. Based on the underwater light attenuation characteristics, post-process the preliminarily enhanced actual image of the underwater target scene to obtain a true image of the underwater target scene.

6. A deep-water measurement system based on binocular ranging as described in any one of claims 1-5, characterized in that, include: The integrated device body is hinged to one end of a robotic arm, and the other end of the robotic arm is hinged to a calibration board mainboard. One side of the calibration board mainboard is a checkerboard pattern, and the other side is a color calibration board. The integrated device body also integrates a binocular camera, an RGBW fill light, and a white light fill light. The control and processing module is electrically connected to the binocular camera, robotic arm, and RGBW fill light, and is used to control the operation of the binocular camera, robotic arm, RGBW fill light, and white light fill light.

Images