Rock debris white light image and fluorescence image registration method
Through the color histogram matching of fluorescent image and SIFT feature extraction, combined with the non-reflective similarity transformation model, the problem of white light and fluorescent lithotripsy images is solved, and the efficiency and registration accuracy of feature extraction are achieved, which is suitable for changing photography. Lithotripsy image registration in the environment.
Patent Information
- Application Number
- CN202310600264.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-25
- Publication Date
- 2024-11-26
AI Technical Summary
The prior art is difficult to accurately register white light and fluorescent lithotripsy images under changing shooting environments, resulting in analysis errors, and conventional methods cannot be directly applicable to fluorescent images.
Through fluorescent image color histogram matching, SIFT feature extraction and matching, combined with the non-reflective similarity transformation model, the coordinate offset parameters were estimated, and the image registration was performed using a bilinear interpolation method.
It realizes registration of white light and fluorescence images in a changing shooting environment, improves feature extraction effectiveness and registration accuracy, can estimate angle changes and translation transformation, and is suitable for common lens changes in well recording production.
Smart Images

Figure CN119027465A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of computer vision and image processing, and is a method for registering a rock cuttings white light image and a fluorescent image. Background Art
[0002] The ultimate goal of drilling geology is to discover and study oil and gas layers. Therefore, it is very important to determine whether there is oil and gas display and the degree of display during the drilling process. In addition to focusing on the description and observation of sandstone and other reservoirs, field logging requires fluorescence analysis. Fluorescence analysis is a direct means of testing oil and gas displays and is an indispensable analytical data for implementing oil and gas layers in geological logging work. In order to accurately reflect the corresponding relationship between fluorescence and reservoirs, it is necessary to include the description of fluorescence in the naming of rock cuttings. Therefore, when describing rock cuttings, it is necessary to describe the name and fluorescence of the same rock cutting.
[0003] In the process of intelligent description of rock cuttings based on image recognition, it is necessary to unify the local subtle image features of rock cuttings in both white light and fluorescence shooting scenes. Therefore, the image physical coordinates of white light image and fluorescence image should be accurately aligned to avoid misalignment and analysis errors. However, in logging production practice, due to unstable shooting conditions, the physical coordinates of white light and fluorescence images often have a certain degree of offset. Therefore, it is a necessary processing step to use image registration technology to register white light and fluorescence images with coordinate offset.
[0004] Conventional computer vision image registration methods, such as registration methods based on optical flow and registration methods based on key point matching, are mainly applicable to continuous motion images under the same shooting environment (generally under natural light). Under the same shooting environment, the images to be registered often have very similar image features, which makes it easier to extract features, match, and estimate offset parameters. However, for the same object, the imaging effects of white light and fluorescent images are very different, and the same image features cannot be directly extracted, let alone directly matched. This poses a great challenge to the problem of white light fluorescent rock chip image registration.
[0005] A Chinese patent document with publication number CN110942475A discloses a UV and visible light image fusion system and a fast image registration method. This method can be used for image registration of UV imaging and visible light images. Its characteristics are that it is necessary to calibrate the camera lenses of UV and visible light imaging equipment in advance, estimate their distortion coefficients, and then directly use the distortion coefficients and shooting parameters to estimate the registration parameters during the registration process. Compared with this method, its defect is that it is often not feasible to accurately calibrate the camera for each lens in production, and the cost is relatively high. In addition, this method cannot be directly applied to fluorescent images.
[0006] Integrated Intensity and Point-Feature Nonrigid Registration (paper DOI: 10.1007 / 978-3-540-30135-6_93.) provides a typical image registration method based on feature point matching, which is applied to medical images. Compared with the more cluttered rock debris images with many targets and large feature variations, the background and targets of medical images are relatively simple and do not have a lot of image noise, so registration is easier. It is difficult to obtain image features with invariance in fluorescence images using this method directly, so it is difficult to apply it to white light and fluorescence image registration.
[0007] The Chinese patent document with publication number CN115588033A discloses a synthetic aperture radar and optical image registration system and method based on structure extraction. The main feature of this method is to use the structural characteristics of terrain and ground targets to construct a multi-layer structure map, and then use the similarity of the structure in the radar image and the optical image for registration. However, this method relies more on being able to extract the terrain structure from the image, but small-scale targets such as rock debris do not have an obvious multi-layer structure, so this method is difficult to apply to the problem of rock debris image registration.
[0008] The above-mentioned related technologies are all methods applied to the non-visible light image registration problem. Although there are many conventional visible light, natural scene image and video registration methods, they cannot be directly applied to white light and fluorescent rock debris image registration. Summary of the invention
[0009] The present invention provides a method for registering rock debris white light images and fluorescent images, which overcomes the shortcomings of the above-mentioned prior art and can effectively solve the problem in the prior art that the inconsistency of white light and fluorescent rock debris image features makes it difficult to perform image registration.
[0010] The technical solution of the present invention is achieved by the following measures: A method for registering a rock debris white light image and a fluorescent image is performed according to the following steps:
[0011] S1, fluorescence image color histogram matching, collect white light image and fluorescence image of rock cutting particles, calculate grayscale white light and grayscale fluorescence images, use grayscale white light image as reference, perform image color histogram matching on grayscale fluorescence image, and obtain the mapped grayscale fluorescence image;
[0012] S2, SIFT feature extraction and matching, extract SIFT features on the grayscale white light image and the mapped grayscale fluorescence image, and perform key point matching;
[0013] S3, coordinate shift parameter estimation, maps each pixel in the fluorescence image to the new coordinate system, and then fills the image using the bilinear interpolation method to obtain the registered fluorescence image.
[0014] The following are further optimizations and / or improvements to the above technical solutions:
[0015] In the above step S1, the specific operation of fluorescence image color histogram matching is:
[0016] S11, respectively calculating the single-channel grayscale images of the white light image and the fluorescence image, processing the white light image and the fluorescence image into RGB three-channel, 256-color images, and calculating the grayscale value of each pixel using an empirical formula based on luminosity;
[0017] S12, performing grayscale value clipping on the grayscale fluorescence image, taking the m% of pixels with the largest grayscale values in the grayscale fluorescence image, recording the minimum value of their grayscale values as τ, and setting the grayscale values of all pixels with grayscale values greater than τ in the grayscale fluorescence image to τ;
[0018] S13, respectively calculating the color distribution histograms of the grayscale white light image and the grayscale fluorescence image, and counting the distribution of the grayscale value of each pixel in the range of 0 to 255, and recording the color distribution histograms of the grayscale white light image and the grayscale fluorescence image as H bright and H flu , respectively 256-dimensional vectors, and then calculate H bright and H flu The cumulative density distribution is C bright and C flu , are also 256-dimensional vectors;
[0019] S14, using the grayscale white light image as a reference, performing color histogram matching on the grayscale fluorescence image after grayscale value clipping, and matching C bright and C flu Interpolation is performed, which is regarded as a reversible mapping from the real number interval [0, 256) to the interval [0, 1], and then each pixel in the grayscale fluorescence image is mapped:
[0020]
[0021] Where I is the original gray value of the grayscale fluorescence image pixel, and I′ is the gray value after mapping.
[0022] In the above step S2, the operation steps of extracting SIFT features on the grayscale white light image and the mapped grayscale fluorescence image are as follows:
[0023] S21, use Harris corner detection algorithm to calculate the key points in the image;
[0024] S22, calculate the edge orientation and strength of each neighboring window centered on each key point;
[0025] S23, calculate the SIFT feature descriptor for each key point.
[0026] In the above step S3, the image coordinate migration model used for estimating the coordinate migration parameters is a non-reflective similarity transformation, and the specific operation is as follows:
[0027] S31, the transformation matrix of the parameters to be estimated is:
[0028]
[0029] Among them, h1, h2, h3, and h4 are the sine angle coefficient, cosine angle coefficient, longitudinal translation coefficient, and lateral offset coefficient respectively, and -h2 is the opposite of the cosine angle coefficient;
[0030] S32, from the coordinates [x, y] of the white light image to the coordinates of the fluorescence image The mapping method is:
[0031]
[0032] Its inverse transformation matrix is:
[0033]
[0034] S33, coordinates from the fluorescence image The mapping to the coordinates [x, y] of the white light image is:
[0035]
[0036] S34, a set of paired key points from the white light image and the fluorescent image generated by the key point matching in step S2 is: By minimizing its minimum mean square error, the optimal coordinate offset parameters can be estimated. in, and [x i ,y i ] are respectively and The coordinates of .
[0037] The above step S3 further includes the following steps:
[0038] S35, after completing the coordinate shift parameter estimation, each pixel in the fluorescence image is mapped to the new coordinate system using H′, and then the image is filled using a bilinear interpolation method to obtain the registered fluorescence image.
[0039] In the above step S11, the empirical formula based on luminosity is:
[0040]
[0041] Wherein, R, G, and B are integers ranging from 0 to 255, and the calculated grayscale value I is the rounded-down value of the weighted sum.
[0042] In the above step S12, m% is 0.1%.
[0043] The present invention can effectively solve the image registration problem in a changing shooting environment that cannot be solved by general image registration methods. For white light and fluorescent images of rock cuttings, it has better feature extraction effectiveness, consistency and registration accuracy, and can estimate common lens changes in logging production cuttings shooting, such as angle changes and translation transformations. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Attached Figure 1 This is a comparison diagram of the image fusion results before and after the registration of the rock cuttings white light image and the fluorescence image in Example 2.
[0045] Attached Figure 2 1 and 2 are the color distribution histograms of the white light image and the fluorescence image of the rock debris in Example 2 and the color histogram of the fluorescence image after equalization.
[0046] Attached Figure 3 This is a schematic diagram of the key point detection results of the white light and fluorescence images of rock cuttings in Example 3.
[0047] Attached Figure 4 This is a comparison diagram of the image fusion results before and after the registration of the rock cuttings white light image and the fluorescence image in Example 8.
[0048] Attached Figure 5 1 and 2 are the color distribution histograms of the white light image and the fluorescence image of the rock debris in Example 8 and the color histogram of the fluorescence image after equalization.
[0049] Attached Figure 6 This is a schematic diagram of the key point detection results of the white light and fluorescence images of rock cuttings in Example 8. DETAILED DESCRIPTION
[0050] The present invention is not limited by the following embodiments, and specific implementation methods can be determined based on the technical solution of the present invention and actual conditions.
[0051] The present invention will be further described below in conjunction with embodiments:
[0052] Embodiment 1: The rock debris white light image and fluorescence image registration method is performed according to the following steps:
[0053] S1, fluorescence image color histogram matching, collect white light image and fluorescence image of rock cutting particles, calculate grayscale white light and grayscale fluorescence images, use grayscale white light image as reference, perform image color histogram matching on grayscale fluorescence image, and obtain the mapped grayscale fluorescence image;
[0054] S2, SIFT feature extraction and matching, extract SIFT features on the grayscale white light image and the mapped grayscale fluorescence image, and perform key point matching;
[0055] S3, coordinate shift parameter estimation, maps each pixel in the fluorescence image to the new coordinate system, and then fills the image using the bilinear interpolation method to obtain the registered fluorescence image.
[0056] Embodiment 2: As an optimization of the above embodiment, in step S1, the specific operation of fluorescence image color histogram matching is:
[0057] S11, respectively calculating the single-channel grayscale images of the white light image and the fluorescence image, processing the white light image and the fluorescence image into RGB three-channel, 256-color images, and using an empirical formula based on luminosity to calculate the grayscale value for each pixel.
[0058] S12, grayscale value clipping is performed on the grayscale fluorescence image, and the m% of pixels with the largest grayscale values in the grayscale fluorescence image are selected, and their minimum grayscale value is recorded as τ, and the grayscale values of all pixels with grayscale values greater than τ in the grayscale fluorescence image are set to τ. The purpose of this step is to reduce the interference of strong fluorescence on the image contrast balance.
[0059] S13, respectively calculating the color distribution histograms of the grayscale white light image and the grayscale fluorescence image, and counting the distribution of the grayscale value of each pixel in the range of 0 to 255, and recording the color distribution histograms of the grayscale white light image and the grayscale fluorescence image as H bright and H flu , respectively 256-dimensional vectors, and then calculate H bright and H flu The cumulative density distribution is C bright and C flu , are also 256-dimensional vectors respectively.
[0060] S14, using the grayscale white light image as a reference, performing color histogram matching on the grayscale fluorescence image after grayscale value clipping, and matching C bright and C flu Interpolation is performed, which is regarded as a reversible mapping from the real number interval [0,256) to the interval [0,1], and then the mapping is performed for each pixel in the grayscale fluorescence image:
[0061]
[0062] Where I is the original gray value of the grayscale fluorescence image pixel, and I′ is the gray value after mapping.
[0063] Since the imaging effects of white light and fluorescent images are very different, the same image features cannot be directly extracted, let alone matched. Figure 1 As shown in the figure, when directly fusing unregistered white light and fluorescence images, their coordinates can be clearly observed to be offset, and the lens has displacement traces. White light images are taken under sufficient lighting conditions, and generally have high contrast and contain relatively clear detail features. However, fluorescence images are taken under weak light conditions, and the image contrast is very low, and the detail features are difficult to show. Therefore, using the white light image as a reference, image color histogram matching is performed on the fluorescence image, and features similar to those of the white light image can be restored from the fluorescence image. Figure 1 As shown in FIG. 1 , after the registration using the method of the present invention, the coordinates of the fused image are basically consistent with the white light image. The color distribution histogram of the rock debris white light image and the fluorescence image and the color histogram of the fluorescence image after equalization are shown in FIG. Figure 2 .
[0064] Embodiment 3: As an optimization of the above embodiment, in step S2, the operation steps of extracting SIFT features on the grayscale white light image and the mapped grayscale fluorescence image are:
[0065] S21, use Harris corner detection algorithm to calculate the key points in the image;
[0066] S22, calculate the edge orientation and strength of each neighboring window centered on each key point;
[0067] S23, calculate the SIFT feature descriptor for each key point.
[0068] After SIFT feature extraction is completed, for each SIFT feature descriptor p in the grayscale white light image bright , search for the SIFT feature descriptor with the smallest Euclidean distance in the descriptors of the grayscale fluorescence image, denoted as p flu The schematic diagram of the key point detection results of the white light and fluorescence images of the rock cuttings in the present invention is shown in Figure 3 .
[0069] Embodiment 4: As an optimization of the above embodiment, in step S3, the image coordinate offset model used for estimating the coordinate offset parameter is a non-reflective similarity transformation (i.e., considering the coordinate offset caused by lens rotation, scale scaling, and translation), and the specific operation is as follows:
[0070] S31, the parameters (transformation matrix) to be estimated are:
[0071]
[0072] Among them, h1, h2, h3, and h4 are the sine angle coefficient, cosine angle coefficient, longitudinal translation coefficient, and lateral offset coefficient respectively, and -h2 is the opposite of the cosine angle coefficient;
[0073] S32, from the coordinates [x, y] of the white light image to the coordinates of the fluorescence image The mapping method is:
[0074]
[0075] Its inverse transformation matrix is:
[0076]
[0077] S33, coordinates from the fluorescence image The mapping to the coordinates [x, y] of the white light image is:
[0078]
[0079] S34, a set of paired key points from the white light image and the fluorescent image generated by the key point matching in step S2 is: By minimizing its minimum mean square error, the optimal coordinate offset parameters can be estimated. in, and [x i ,y i ] are respectively and The coordinates of .
[0080] Embodiment 5: As an optimization of the above embodiment, step S3 further includes the following steps:
[0081] S35, after completing the coordinate shift parameter estimation, each pixel in the fluorescence image is mapped to the new coordinate system using H′, and then the image is filled using a bilinear interpolation method to obtain the registered fluorescence image.
[0082] Embodiment 6: As an optimization of the above embodiment, in step S11, the empirical formula based on luminosity is:
[0083]
[0084] Wherein, R, G, and B are integers ranging from 0 to 255, and the calculated grayscale value I is the rounded-down value of the weighted sum.
[0085] Embodiment 7: As an optimization of the above embodiment, in step S12, m% is 0.1%.
[0086] Embodiment 8: The rock cuttings white light image and fluorescence image registration method registers a group of rock cuttings particles with white light images and fluorescence images according to the following steps:
[0087] S1, fluorescence image color histogram matching, collect white light image and fluorescence image of rock debris particles, calculate grayscale white light and grayscale fluorescence images, use grayscale white light image as reference, perform image color histogram matching on grayscale fluorescence image, and obtain the mapped grayscale fluorescence image; the image fusion results before and after registration are shown in Figure 4 In the fluorescence image color histogram matching step, the clipping ratio m% = 0.1% (i.e., the pixels with the largest 1% grayscale values in the grayscale fluorescence image are taken). The results after implementing the color histogram matching are as follows: Figure 5 shown.
[0088] S2, SIFT feature extraction and matching, extract SIFT features on the grayscale white light image and the mapped grayscale fluorescence image, and perform key point matching; the results of SIFT feature extraction and key point matching are shown in Figure 6 shown.
[0089] S3, coordinate shift parameter estimation, maps each pixel in the fluorescence image to the new coordinate system, and then fills the image using the bilinear interpolation method to obtain the registered fluorescence image.
[0090] Finally, the coordinate offset parameter estimation step estimates that the scale scaling ratio of the fluorescence image is about 101%, the rotation angle is about 3°, and the translation is about (-10.0, +10.0).
[0091] In summary, the present invention provides a method for aligning white light images and fluorescent images of rock cuttings, which can effectively solve the image alignment problem in a changing shooting environment that cannot be solved by general image alignment methods. For white light and fluorescent images of rock cuttings, it has better feature extraction effectiveness, consistency and alignment accuracy, and can estimate common lens changes in the shooting of rock cuttings in logging production, such as angle changes and translation transformations.
[0092] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Non-essential technical features can be added or reduced according to actual needs to meet the requirements of different situations.
Claims
1. A method for registering white light images and fluorescence images of rock debris, characterized in that Follow these steps: S1, fluorescence image color histogram matching, collect white light image and fluorescence image of rock cutting particles, calculate grayscale white light and grayscale fluorescence images, use grayscale white light image as reference, perform image color histogram matching on grayscale fluorescence image, and obtain the mapped grayscale fluorescence image; S2, SIFT feature extraction and matching, extract SIFT features on the grayscale white light image and the mapped grayscale fluorescence image, and perform key point matching; S3, coordinate shift parameter estimation, maps each pixel in the fluorescence image to the new coordinate system, and then fills the image using the bilinear interpolation method to obtain the registered fluorescence image.
2. The method for registering rock debris white light images and fluorescence images according to claim 1, characterized in that In step S1, the specific operation of fluorescence image color histogram matching is: S11, respectively calculating the single-channel grayscale images of the white light image and the fluorescence image, processing the white light image and the fluorescence image into RGB three-channel, 256-color images, and calculating the grayscale value of each pixel using an empirical formula based on luminosity; S12, performing grayscale value clipping on the grayscale fluorescence image, taking the m% of pixels with the largest grayscale values in the grayscale fluorescence image, recording the minimum value of their grayscale values as τ, and setting the grayscale values of all pixels with grayscale values greater than τ in the grayscale fluorescence image to τ; S13, respectively calculating the color distribution histograms of the grayscale white light image and the grayscale fluorescence image, and counting the distribution of the grayscale value of each pixel in the range of 0 to 255, and recording the color distribution histograms of the grayscale white light image and the grayscale fluorescence image as H bright and H flu , respectively 256-dimensional vectors, and then calculate H bright and H flu The cumulative density distribution is C bright and C flu , are also 256-dimensional vectors; S14, using the grayscale white light image as a reference, performing color histogram matching on the grayscale fluorescence image after grayscale value clipping, and matching C bright and C flu Interpolation is performed, which is regarded as a reversible mapping from the real number interval [0, 256) to the interval [0, 1], and then each pixel in the grayscale fluorescence image is mapped: Where I is the original gray value of the grayscale fluorescence image pixel, and I′ is the gray value after mapping.
3. The method for registering rock debris white light images and fluorescence images according to claim 1 or 2, characterized in that In step S2, the operation steps of extracting SIFT features on the grayscale white light image and the mapped grayscale fluorescence image are as follows: S21, use Harris corner detection algorithm to calculate the key points in the image; S22, calculate the edge orientation and strength of each neighboring window centered on each key point; S23, calculate the SIFT feature descriptor for each key point.
4. The method for registering rock debris white light images and fluorescence images according to claim 1, characterized in that In step S3, the image coordinate offset model used for estimating the coordinate offset parameters is a non-reflective similarity transformation, and the specific operation is as follows: S31, the transformation matrix of the parameters to be estimated is: Among them, h1, h2, h3, and h4 are the sine angle coefficient, cosine angle coefficient, longitudinal translation coefficient, and lateral offset coefficient respectively, and -h2 is the opposite of the cosine angle coefficient; S32, from the coordinates [x, y] of the white light image to the coordinates of the fluorescence image The mapping method is: Its inverse transformation matrix is: S33, coordinates from the fluorescence image The mapping to the coordinates [x, y] of the white light image is: S34, a set of paired key points from the white light image and the fluorescent image generated by the key point matching in step S2 is: By minimizing its minimum mean square error, the optimal coordinate offset parameters can be estimated. in, and [xi, yi] are and The coordinates of .
5. The method for registering rock debris white light images and fluorescence images according to claim 4, characterized in that Step S3 also includes the following steps: S35, after completing the coordinate shift parameter estimation, each pixel in the fluorescence image is mapped to the new coordinate system using H′, and then the image is filled using a bilinear interpolation method to obtain the registered fluorescence image.
6. The method for registering rock debris white light images and fluorescence images according to claim 2, characterized in that In step S11, the empirical formula based on luminosity is: Wherein, R, G, and B are integers ranging from 0 to 255, and the calculated grayscale value I is the rounded-down value of the weighted sum.
7. The method for registering rock debris white light images and fluorescence images according to claim 2, characterized in that In step S12, m% is 0.1%.
