A three-dimensional topography reconstruction method based on inter-sequence sparse aggregation

Abstract

The present application relates to a kind of three-dimensional topography reconstruction method based on inter-sequence sparse aggregation.It includes: step 1, obtaining the image sequence of different focusing level of the scene to be measured;Step 2, using Laplacian operator and image sequence are convolved to obtain focusing evaluation sparse coefficient;Step 3, by constructing three-dimensional pixel space, use the dense coefficient that the sparse coefficient is aggregated into by the three-dimensional space, the aggregation process takes into account the local focusing information of image itself and the structure information between image sequence;Step 4, the position of the maximum value of the absolute value of dense coefficient in image sequence dimension is obtained to obtain the final three-dimensional topography reconstruction result of the scene to be measured.The method of the present application can effectively retain the three-dimensional structure information in the scene to be measured, and obtain more accurate three-dimensional topography reconstruction result.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of three-dimensional reconstruction, and particularly relates to a three-dimensional topography reconstruction method based on inter-sequence sparse aggregation. BACKGROUND

[0002] The three-dimensional topography reconstruction method based on two-dimensional image sequence is widely concerned by the industry due to its advantages such as various data collection forms, easy use, less constraints on reconstruction algorithm conditions, consideration of high reconstruction efficiency and high precision, and additional texture information of depth image, and the related technology has been successfully applied to medical and cosmetic examination and virtual 3D modeling of customs goods.

[0003] At present, the three-dimensional topography reconstruction method based on two-dimensional image sequence is mainly divided into image sequence decomposition and depth representation information aggregation according to the reconstruction steps. The image sequence decomposition mainly performs convolution operation on each image in the image sequence through different types of operators, and then obtains the overall evaluation of the focusing level of all images. The overall focusing evaluation result is usually represented in the form of sparse coefficients. Typical operators include Laplace operators, deformable operators, gradient operators, etc. Different types of operators are suitable for different application scenarios. Therefore, how to eliminate the influence of different operators on the three-dimensional topography reconstruction result is a key task of depth representation information aggregation. After the image sequence is decomposed, the sparse coefficients representing the depth information can be obtained, and then the conversion from sparse coefficients to dense coefficients can be realized through the design of a specific sparse aggregation method. Finally, the final three-dimensional topography reconstruction result of the scene to be measured is obtained through the projection of the dense coefficients in the image sequence dimension. The entire depth representation information aggregation process realizes the mapping between the sparse coefficients of the image sequence and the depth result. Representative sparse aggregation methods include guided filtering, weighted median filtering, and mean filtering.

[0004] Through understanding the research status, it is considered that the methods in this field mainly face the following challenges: in the existing three-dimensional topography reconstruction method, the depth representation information aggregation process usually only considers the sparse aggregation problem of a single image itself, and does not consider the information between the image sequence. However, the image sequence will imply a lot of scene three-dimensional structure information, and the lack of such information will lead to the inability to maintain the integrity of the scene structure in the three-dimensional topography reconstruction result. Therefore, how to propose a sparse aggregation method that takes into account the structure information between the image sequence and the local focusing information of the image is the key to improving the accuracy of the three-dimensional topography reconstruction result.

[0005] As can be seen from the above, the present application introduces the correlation between the image sequences to solve the problem of lack of structure information in the sparse aggregation result in the existing three-dimensional topography reconstruction process. In the reconstruction process, the local focusing information of the image itself and the structure information between the image sequences are considered, and a three-dimensional topography reconstruction result with higher precision can be obtained. SUMMARY

[0006] In order to overcome the problems in the above-mentioned technologies, the present application aims to provide a three-dimensional topography reconstruction method based on inter-sequence sparse aggregation.

[0007] The technical solution adopted by the present application is: a three-dimensional topography reconstruction method based on inter-sequence sparse aggregation, comprising the following steps:

[0008] Step 1: obtaining an image sequence of different focus levels of a scene to be measured by moving the distance between the camera and the scene , wherein is the subscript of the image sequence and its value range is , is the total number of image sequences, is the pixel coordinate position and its value is ;

[0009] Step 2: performing convolution operation on the image sequence in step 1 according to formula (1) and Laplace operator (LAPA) to obtain the focus evaluation sparse coefficient of all images in the image sequence , ,

[0010] (1)

[0011] wherein represents the number of images and its value range is , is the pixel coordinate position and its value is , indicates the convolution operator, LAPA is the Laplace operator and its expression is , indicates the partial derivative symbol;

[0012] Step 3: performing sparse aggregation on the sparse coefficient obtained in step 2 according to formula (2) to obtain the dense coefficient ,

[0013] (2)

[0014] wherein indicates the pixel coordinate and the weight between the pixel coordinate , the expression is shown in formula (3),

[0015] (3)

[0016] wherein the pixel coordinate , is the proportion coefficient, exp is the exponential function identifier, and are respectively​ Gradient of position, expression as shown in equation (4),

[0017] (4)

[0018] Wherein represents the pixel value of the position in the image, represents the pixel value of the position in the image, represents the pixel value of the position in the image, represents the image number and its value range is ,

[0019] In equation (2) represents a three-dimensional pixel space with pixel coordinates as the center, length and width, height, Expression as shown in equation (5);

[0020] (5)

[0021] Step 4, according to equation (6), calculate the absolute value of the dense coefficient in the image sequence dimension The position of the maximum value of the dense coefficient is obtained to obtain the final three-dimensional topography reconstruction result of the scene ,

[0022] (6)

[0023] Wherein represents the absolute value function, represents the function of solving the dense coefficient Subscript .

[0024] Compared with the prior art, the method of the present application can effectively preserve the three-dimensional structure information in the scene to be measured, and obtain a three-dimensional topography reconstruction result with higher precision. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 It is a flow chart of a three-dimensional topography reconstruction method based on inter-sequence sparse aggregation;

[0026] Figure 2 It is a schematic diagram of a three-dimensional topography reconstruction method based on inter-sequence sparse aggregation. DETAILED DESCRIPTION

[0027] As Figure 1 , Figure 2 ​As shown, a three-dimensional shape reconstruction method based on inter-sequence sparse aggregation includes the following steps:

[0028] Step 1, obtain an image sequence of different focus levels of a scene to be measured by moving the distance between the camera and the scene wherein is the subscript of the image sequence and its value range is , is the total number of image sequences, is the pixel coordinate position and its value is ;

[0029] Step 2, according to formula (1) and Laplace operator (LAPA), the image sequence in step 1 is convoluted to obtain the focus evaluation sparse coefficient of all images in the image sequence ,

[0030] (1)

[0031] wherein represents the number of images and its value range is , is the pixel coordinate position and its value is , represents the convolution operator, LAPA is the Laplace operator and its expression is , represents the partial derivative symbol;

[0032] Step 3, according to formula (2), the sparse coefficient obtained in step 2 is sparsely aggregated to obtain a dense coefficient ,

[0033] (2)

[0034] wherein represents the pixel coordinate and the weight between the pixel coordinate , the expression is shown in formula (3),

[0035] (3)

[0036] wherein the pixel coordinate , is the proportion coefficient, exp is the exponential function identifier, and are the gradients of the positions and ,

[0037] (4) ​

[0038] in Indicates the first The location in the image is pixel values, Indicates the first The location in the image is pixel values, Represents the number of images and its value range is ,

[0039] In formula (2) Represented in pixel coordinates Centered on, For length and width, For a high three-dimensional pixel space, The expression is shown in equation (5);

[0040] (5)

[0041] Step 4, use the density coefficient obtained in Step 3. Calculate its dimension in the image sequence according to equation (6). The location of the maximum absolute value of the density coefficient yields the final 3D topography reconstruction result of the scene. ,

[0042] (6)

[0043] in Represents the absolute value function. This indicates the solution for the density coefficient. Subscript The function.

Claims

1. A three-dimensional topography reconstruction method based on sparse aggregation between sequences, characterized by the following steps: Step 1: Obtain image sequences of the scene under test at different focus levels by moving the camera away from the scene. ,in The subscript for the image sequence is [index] and its value range is [range]. , The total number of image sequences. The pixel coordinates are the values ​​of the pixels. ; Step 2, for the image sequence in Step 1 By performing a convolution operation with the Laplacian operator (LAPA) according to equation (1), the focus evaluation sparsity coefficients of all images in the image sequence are obtained. , （1） in Represents the number of images and its value range is , The pixel coordinate position is set to a value , LAPA represents the convolution operator, where it is the Laplace operator and its expression is: , Indicates the sign of a partial derivative; Step 3, convert the sparsity coefficients obtained in Step 2 into... The density coefficient is obtained by sparse polymerization according to equation (2). , （2） in Represents pixel coordinates With pixel coordinates The weights between them are expressed as shown in equation (3). （3） where pixel coordinates , Here, exp is the proportionality constant, and exp is the identifier for the exponential function. They are respectively and The gradient of the position is expressed as shown in equation (4). （4） in Indicates the first The location in the image is pixel values, Indicates the first The location in the image is pixel values, Represents the number of images and its value range is , In formula (2) Represented in pixel coordinates Centered on, For length and width, For a high three-dimensional pixel space, The expression is shown in equation (5); （5） Step 4, use the density coefficient obtained in Step 3. Calculate its dimension in the image sequence according to equation (6). The location of the absolute maximum value in the graph yields the final 3D topography reconstruction result of the scene. , （6） in Represents the absolute value function. This indicates the solution for the density coefficient. Subscript The function.

Images