Multi-scale target detection method, surgical robot and related products

Abstract

The application discloses a multiscale target detection method, a surgical robot and related products. The method comprises the following steps: acquiring a first three-dimensional CT image, the first three-dimensional CT image being obtained by scanning a first object by a three-dimensional CT scanning device with a first layer thickness. Based on the first layer thickness, a first convolution kernel and a second convolution kernel are determined. The first three-dimensional CT image is processed by using the first convolution kernel to obtain first feature information of pixels in the first three-dimensional CT image. The first three-dimensional CT image is processed by using the second convolution kernel to obtain second feature information of the pixels in the first three-dimensional CT image. Based on the first feature information and the second feature information, a first image region corresponding to the first object is determined from the first three-dimensional CT image. The corresponding products are also disclosed. The method provided by the application can improve the accuracy of determining the image region corresponding to the first object from the first three-dimensional CT image.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical imaging, and in particular to a multi-scale target detection method, a surgical robot and related products. BACKGROUND

[0002] Three-dimensional computed tomography (CT) images have been widely used in the medical field due to their advantage of clearly presenting the three-dimensional structure information of a target, and are the core means for obtaining spatial information of an object. Specifically, by performing target detection on a three-dimensional CT image, an image region corresponding to the object can be determined from the three-dimensional CT image, and then the spatial information of the object, such as the position information of the object, can be determined based on the image region.

[0003] However, based on the current technology, the accuracy of the determined image region corresponding to the object is low. SUMMARY

[0004] The present application provides a multi-scale target detection method, a surgical robot and related products to improve the accuracy of determining an image region corresponding to a first object from a three-dimensional CT image. The related products include a multi-scale target detection device and an electronic device.

[0005] In a first aspect, a multi-scale target detection method is provided, and the method includes:

[0006] obtaining a first three-dimensional CT image, the first three-dimensional CT image being obtained by a three-dimensional CT scanning device scanning a first object with a first layer thickness;

[0007] determining a first convolution kernel and a second convolution kernel based on the first layer thickness, the size of the first convolution kernel and the size of the second convolution kernel being positively correlated with the first layer thickness, and the size of the first convolution kernel being different from the size of the second convolution kernel;

[0008] processing the first three-dimensional CT image using the first convolution kernel to obtain first feature information of pixels in the first three-dimensional CT image;

[0009] processing the first three-dimensional CT image using the second convolution kernel to obtain second feature information of the pixels in the first three-dimensional CT image;

[0010] determining a first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information.

[0011] In combination with any of the embodiments of the present application, the processing of the first three-dimensional CT image by using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image comprises:

[0012] Based on the first three-dimensional CT image, first related information and / or second related information are obtained, the first related information indicating the correlation of the curvature of the pixels in the first three-dimensional CT image with the curvature of adjacent pixels, and the second related information indicating the correlation of the gray value of the pixels in the first three-dimensional CT image with the gray value of adjacent pixels.

[0013] Based on the first related information and / or the second related information, the first feature information is obtained.

[0014] In combination with any of the embodiments of the present application, the correlation of the curvature of the pixels in the first three-dimensional CT image with the curvature of adjacent pixels comprises at least one of the following: the correlation of the curvature of the pixels in the first three-dimensional CT image with the curvature of adjacent pixels in a transverse direction, the correlation of the curvature of the pixels in the first three-dimensional CT image with the curvature of adjacent pixels in a longitudinal direction, and the correlation of the curvature of the pixels in the first three-dimensional CT image with the curvature of adjacent pixels in a vertical direction, the transverse direction being the direction of the horizontal axis of the image coordinate system of the first three-dimensional CT image, the longitudinal direction being the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image, and the vertical direction being the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image.

[0015] In combination with any of the embodiments of the present application, the first object is a sphere.

[0016] The determination of the first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information comprises:

[0017] Based on the first feature information, the pixels in the first three-dimensional CT image corresponding to the center of the sphere of the first object are determined to obtain n first candidate pixels, the n being a positive integer.

[0018] Based on the first feature information, the radius of the first object in the first three-dimensional CT image is determined to obtain n first candidate radii, the first candidate radii corresponding to the first candidate pixels one by one.

[0019] Based on the second feature information, the pixels in the first three-dimensional CT image corresponding to the center of the sphere of the first object are determined to obtain m second candidate pixels, the m being a positive integer.

[0020] determine a radius of the first object in the first three-dimensional CT image based on the second feature information, to obtain m second candidate radii, the second candidate radii corresponding to the second candidate pixels one by one;

[0021] determine a sphere center pixel based on the n first candidate pixels and the m second candidate pixels, the sphere center pixel being a pixel in the first three-dimensional CT image corresponding to a sphere center of the first object;

[0022] determine a target radius based on the n first candidate radii and the m second candidate radii, the target radius being the radius of the first object in the first three-dimensional CT image;

[0023] determine the first image region from the first three-dimensional CT image based on the sphere center pixel and the target radius.

[0024] In any of the embodiments of the present application, the first object is attached to a surface of a second object, and the first three-dimensional CT image includes the second object;

[0025] The determination of the sphere center pixel based on the n first candidate pixels and the m second candidate pixels includes:

[0026] determine a second image region corresponding to the surface of the second object from the first three-dimensional CT image;

[0027] determine n first candidate spheres based on the n first candidate pixels and the n first candidate radii;

[0028] determine n third image regions from the second image region based on the n first candidate spheres, the first candidate spheres corresponding to the third image regions one by one, and the third image regions including image regions in contact with the first candidate spheres and image regions within the first candidate spheres;

[0029] determine r third candidate pixels from the n first candidate pixels based on the n third image regions, r being a positive integer less than or equal to n, the r third candidate pixels being sphere centers of r third candidate spheres in the n first candidate spheres, and a ratio of an area of the third image region corresponding to the third candidate sphere to a surface area of the third candidate sphere being less than a first threshold value;

[0030] determine the sphere center pixel based on the r third candidate pixels and the m second candidate pixels.

[0031] In any of the embodiments of the present application, the determination of the pixel in the first three-dimensional CT image corresponding to the sphere center of the first object based on the first feature information to obtain n first candidate pixels includes:

[0032] determine, based on the first feature information, t first surface pixel sets, the first surface pixel set being a set of pixels belonging to a surface of the first object, t being an integer greater than or equal to n;

[0033] fit pixels in the t first surface pixel sets respectively to obtain t fourth candidate spheres;

[0034] obtain the n first candidate pixels based on centers of the t fourth candidate spheres.

[0035] With any of the embodiments of the present application, the determination of the sphere center pixel based on the r third candidate pixels and the m second candidate pixels comprises:

[0036] determine, based on the r third candidate pixels, r two-dimensional CT images from the first three-dimensional CT image, the third candidate pixel corresponding to the two-dimensional CT image one by one, the two-dimensional CT image not including the third candidate pixel, the two-dimensional CT image intersecting with a second surface pixel set, the second surface pixel set being a set corresponding to the third candidate pixel in the t first surface pixel sets;

[0037] determine radii of circles in the r two-dimensional CT images respectively to obtain r circle radii;

[0038] determine distances between the r two-dimensional CT images and the corresponding third candidate pixels respectively to obtain r distances;

[0039] obtain r third candidate radii based on the r circle radii and the r distances, the third candidate radius being a radius of the first object in the first three-dimensional CT image, the third candidate radius corresponding to the third candidate pixel one by one;

[0040] determine, from the n first candidate radii, r fourth candidate radii corresponding to the r third candidate pixels;

[0041] determine, based on the r third candidate radii and the r fourth candidate radii, s fourth candidate pixels from the r third candidate pixels, s being a positive integer less than or equal to r, the absolute value of the difference between the third candidate radius corresponding to the fourth candidate pixel and the fourth candidate radius corresponding to the fourth candidate pixel being less than a second threshold value;

[0042] determine the sphere center pixel based on the s fourth candidate pixels and the m second candidate pixels.

[0043] In a second aspect, a multi-scale target detection device is provided, the multi-scale target detection device comprising:

[0044] an acquisition unit, configured to acquire a first three-dimensional CT image, the first three-dimensional CT image being obtained by a three-dimensional CT scanning device by scanning a first object with a first layer thickness;

[0045] a processing unit, configured to determine a first convolution kernel and a second convolution kernel based on the first layer thickness, sizes of the first convolution kernel and the second convolution kernel being positively correlated with the first layer thickness, and the size of the first convolution kernel being different from the size of the second convolution kernel;

[0046] the processing unit is further configured to process the first three-dimensional CT image by using the first convolution kernel to obtain first feature information of pixels in the first three-dimensional CT image;

[0047] the processing unit is further configured to process the first three-dimensional CT image by using the second convolution kernel to obtain second feature information of the pixels in the first three-dimensional CT image;

[0048] the processing unit is further configured to determine a first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information.

[0049] In combination with any one of the embodiments of the present application, the processing unit is further configured to:

[0050] obtain first related information and / or second related information based on the first three-dimensional CT image, the first related information indicating a correlation between curvatures of pixels in the first three-dimensional CT image and curvatures of adjacent pixels, and the second related information indicating a correlation between gray values of the pixels in the first three-dimensional CT image and gray values of adjacent pixels;

[0051] obtain the first feature information based on the first related information and / or the second related information.

[0052] In combination with any one of the embodiments of the present application, the correlation between the curvatures of the pixels in the first three-dimensional CT image and the curvatures of the adjacent pixels includes at least one of a correlation between the curvatures of the pixels in the first three-dimensional CT image and the curvatures of the adjacent pixels in a transverse direction, a correlation between the curvatures of the pixels in the first three-dimensional CT image and the curvatures of the adjacent pixels in a longitudinal direction, and a correlation between the curvatures of the pixels in the first three-dimensional CT image and the curvatures of the adjacent pixels in a vertical direction, the transverse direction being a direction of a horizontal axis of an image coordinate system of the first three-dimensional CT image, the longitudinal direction being a direction of a vertical axis of the image coordinate system of the first three-dimensional CT image, and the vertical direction being a direction of a vertical axis of the image coordinate system of the first three-dimensional CT image.

[0053] In any of the embodiments of the present application, the first object is a sphere; and the processing unit is further configured to:

[0054] determine, based on the first feature information, a pixel corresponding to a sphere center of the first object in the first three-dimensional CT image, to obtain n first candidate pixels, the n being a positive integer;

[0055] determine, based on the first feature information, a radius of the first object in the first three-dimensional CT image, to obtain n first candidate radii, the first candidate radii corresponding to the first candidate pixels one by one;

[0056] determine, based on the second feature information, a pixel corresponding to a sphere center of the first object in the first three-dimensional CT image, to obtain m second candidate pixels, the m being a positive integer;

[0057] determine, based on the second feature information, a radius of the first object in the first three-dimensional CT image, to obtain m second candidate radii, the second candidate radii corresponding to the second candidate pixels one by one;

[0058] determine, based on the n first candidate pixels and the m second candidate pixels, a sphere center pixel, the sphere center pixel being the pixel corresponding to the sphere center of the first object in the first three-dimensional CT image;

[0059] determine, based on the n first candidate radii and the m second candidate radii, a target radius, the target radius being the radius of the first object in the first three-dimensional CT image;

[0060] determine, based on the sphere center pixel and the target radius, the first image region from the first three-dimensional CT image.

[0061] In any of the embodiments of the present application, the first object is attached to a surface of a second object, and the first three-dimensional CT image includes the second object;

[0062] the processing unit is further configured to:

[0063] determine, from the first three-dimensional CT image, a region corresponding to the surface of the second object, to obtain a second image region;

[0064] determine, based on the n first candidate pixels and the n first candidate radii, n first candidate spheres;

[0065] determine, based on the n first candidate spheres, n third image regions from the second image region, the first candidate spheres corresponding to the third image regions one by one, and the third image regions including image regions in contact with the first candidate spheres and image regions within the first candidate spheres;

[0066] determining r third candidate pixels from the n first candidate pixels based on the n third image regions, the r being a positive integer less than or equal to the n, the r third candidate pixels being the centers of r third candidate spheres in the n first candidate spheres, a ratio of an area of the third image region corresponding to the third candidate sphere to a surface area of the third candidate sphere being less than a first threshold value;

[0067] determining the center pixel based on the r third candidate pixels and the m second candidate pixels.

[0068] Optionally, the first threshold value is determined based on an actual contact area of the first object and the second object. For example, the actual contact area of the first object and the second object is A1, a surface area of the first object is A2, wherein A1 / A2 = 0.3, and the first threshold value can be 0.3.

[0069] According to any one of the embodiments of the present application, the processing unit is further configured to:

[0070] determining t first surface pixel sets based on the first feature information, the first surface pixel set being a set of pixels belonging to a surface of the first object, the t being an integer greater than or equal to the n;

[0071] fitting the pixels in the t first surface pixel sets respectively to obtain t fourth candidate spheres;

[0072] obtaining the n first candidate pixels based on the centers of the t fourth candidate spheres.

[0073] According to any one of the embodiments of the present application, the processing unit is further configured to:

[0074] determining r two-dimensional CT images from the first three-dimensional CT image based on the r third candidate pixels, the third candidate pixel corresponding to the two-dimensional CT image one by one, the two-dimensional CT image not including the third candidate pixel, the two-dimensional CT image intersecting with a second surface pixel set, the second surface pixel set being a set corresponding to the third candidate pixel in the t first surface pixel sets;

[0075] determining radii of circles in the r two-dimensional CT images respectively to obtain r circle radii;

[0076] determining distances between the r two-dimensional CT images and the corresponding third candidate pixels respectively to obtain r distances;

[0077] Based on the r radii of the circles and the r distances, r third candidate radii are obtained, the third candidate radii being radii of the first object in the first three-dimensional CT image, the third candidate radii corresponding to the third candidate pixels one by one;

[0078] From the n first candidate radii, r fourth candidate radii corresponding to the r third candidate pixels are determined;

[0079] Based on the r third candidate radii and the r fourth candidate radii, s fourth candidate pixels are determined from the r third candidate pixels, the s being a positive integer less than or equal to the r, an absolute value of a difference between the third candidate radius corresponding to the fourth candidate pixel and the fourth candidate radius corresponding to the fourth candidate pixel being less than a second threshold value;

[0080] Based on the s fourth candidate pixels and the m second candidate pixels, the ball center pixel is determined.

[0081] In the third aspect, the surgical robot can perform the multi-scale target detection method through the multi-scale target detection device, and the following effects can be achieved: improving the accuracy of determining the image region corresponding to the first object from the three-dimensional CT image.

[0082] In the fourth aspect, an electronic device is provided, including a processor and a memory, the memory being configured to store computer program code, the computer program code including computer instructions, and the electronic device being configured to execute the method according to the first aspect and any possible implementation manner thereof when the processor executes the computer instructions.

[0083] In the fifth aspect, another electronic device is provided, including a processor, a sending device, an input device, an output device and a memory, the memory being configured to store computer program code, the computer program code including computer instructions, and the electronic device being configured to execute the method according to the first aspect and any possible implementation manner thereof when the processor executes the computer instructions.

[0084] In the sixth aspect, a computer readable storage medium is provided, the computer readable storage medium storing a computer program, the computer program including program instructions, and the program instructions causing a processor to execute the method according to the first aspect and any possible implementation manner thereof when the program instructions are executed by the processor.

[0085] In a seventh aspect, a computer program product is provided, which includes a computer program or instructions, which, when running on a computer, causes the computer to perform the method of the first aspect and any possible implementation thereof.

[0086] It should be understood that the general description above and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

[0087] In the embodiments of the present application, the first three-dimensional CT image is obtained by scanning the first object with a first three-dimensional CT scanning device at a first layer thickness. The detection device determines a first convolution kernel and a second convolution kernel based on the first layer thickness, wherein the size of the first convolution kernel and the size of the second convolution kernel are both positively correlated with the first layer thickness, and the size of the first convolution kernel is different from the size of the second convolution kernel. The first three-dimensional CT image is processed using the first convolution kernel to obtain first feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the first feature information. The first three-dimensional CT image is processed using the second convolution kernel to obtain second feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the second feature information. Thus, feature information of different scales can be obtained. Then, based on the first feature information and the second feature information, a first image region corresponding to the first object is determined from the first three-dimensional CT image, which can improve the accuracy of the first image region. BRIEF DESCRIPTION OF DRAWINGS

[0088] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be described below.

[0089] The drawings herein are incorporated into the specification and form part of the specification, which show embodiments consistent with the present application, and together with the specification, serve to explain the technical solutions of the present application.

[0090] Figure 1 A flowchart of a multi-scale target detection method provided by the embodiments of the present application;

[0091] Figure 2 A schematic diagram of a first three-dimensional CT image in the embodiments of the present application;

[0092] Figure 3 A schematic diagram of a cross section of a first image region provided by the embodiments of the present application;

[0093] Figure 4 A structural schematic diagram of a multi-scale target detection device provided by the embodiments of the present application;

[0094] Figure 5 A hardware structural schematic diagram of an electronic device provided by the embodiments of the present application. Detailed Implementation

[0095] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0096] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0097] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. It should be understood that in this application, "at least one" means one or more, "more" means two or more, and "at least two" means two or three or more.

[0098] The execution subject of this application embodiment is a multi-scale target detection device (hereinafter referred to as the detection device), wherein the detection device can be any electronic device capable of executing the technical solution disclosed in the method embodiment of this application. Optionally, the detection device can be one of the following: a computer, a server.

[0099] It should be understood that the method embodiments of this application can also be implemented by a processor executing computer program code. The embodiments of this application are described below with reference to the accompanying drawings. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating a multi-scale target detection method provided in an embodiment of this application.

[0100] 101. Acquire a first three-dimensional CT image, wherein the first three-dimensional CT image is obtained by scanning the first object with a first slice thickness using a three-dimensional CT scanning device.

[0101] In the embodiments of the present application, the first three-dimensional CT image is obtained by scanning the first object with a CT scanning device using a first layer thickness. In the process of scanning to obtain the first three-dimensional CT image, the CT scanning device scans a region with a thickness of the first layer thickness to obtain the first three-dimensional CT image, i.e., the first three-dimensional CT image includes information of the region with the thickness of the first layer thickness. The greater the first layer thickness, the lower the resolution of the first three-dimensional CT image, i.e., the first layer thickness is negatively correlated with the resolution of the first three-dimensional CT image.

[0102] The first object can be any target that needs to be detected from the three-dimensional CT image. In some schemes, the first object is an optical marker attached to the surface of the human body. For example, after an optical small ball is attached to the surface of the human body, the position of the optical small ball in the world coordinate system can be determined based on an optical tracking device. Scanning the human body with a CT scanning device can obtain a three-dimensional CT image including the optical small ball. Based on the position of the optical small ball in the world coordinate system and the position of the optical small ball in the image coordinate system of the three-dimensional CT image, the conversion relationship between the world coordinate system and the image coordinate system can be determined.

[0103] Optionally, the number of the first objects is greater than 1. For example, the first object is six optical small balls attached to the surface of the human body, and the number of the first object is six.

[0104] 102. Determine the first convolution kernel and the second convolution kernel based on the first layer thickness, wherein the size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the first layer thickness, and the size of the first convolution kernel is different from the size of the second convolution kernel.

[0105] In the embodiments of the present application, the convolution kernel (including the first convolution kernel and the second convolution kernel described above) is used to process the first three-dimensional CT image to obtain the feature information of the first three-dimensional CT image. Optionally, the convolution kernel is a Gaussian kernel. The Gaussian kernel is a three-dimensional matrix constructed based on a Gaussian function, and the Gaussian kernel can be used for convolution processing of the three-dimensional CT image.

[0106] The size of the convolution kernel indicates the scale of the convolution kernel, and correspondingly, the receptive field of the convolution kernel is large. The size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the first layer thickness, i.e., the scale of the first convolution kernel and the scale of the second convolution kernel are positively correlated with the first layer thickness. The size of the first convolution kernel is different from the size of the second convolution kernel, i.e., the scale of the first convolution kernel is different from the scale of the second convolution kernel.

[0107] 103. Process the first three-dimensional CT image using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image.

[0108] In one aspect, the lower the resolution of the first three-dimensional CT image, the smaller the pixel density of the first three-dimensional CT image, and correspondingly, the smaller the amount of information in the image region in the first three-dimensional CT image, i.e., the resolution of the first three-dimensional CT image is positively correlated with the amount of information in the image region in the first three-dimensional CT image. For example, the first three-dimensional CT image includes an image region r1. When the resolution of the first three-dimensional CT image is f1, the amount of information of the image region r1 is m1. When the resolution of the first three-dimensional CT image is f2, the amount of information of the image region r1 is m2. If f1 is greater than f2, then m1 is greater than m2. Since the first layer thickness is negatively correlated with the resolution of the first three-dimensional CT image, the first layer thickness is negatively correlated with the amount of information in the image region in the first three-dimensional CT image.

[0109] On the other hand, when the first three-dimensional CT image is processed using the convolution kernel, the image region processed by the convolution kernel is the image region corresponding to the receptive field of the convolution kernel, e.g., the size of the convolution kernel is 3x3x3, and the image region corresponding to the receptive field of the convolution kernel is the image region with a size of 3x3x3 in the first three-dimensional CT image. Since the first layer thickness is negatively correlated with the amount of information in the image region in the first three-dimensional CT image, if the size of the convolution kernel is unchanged, the volume of the image region processed by the convolution kernel is unchanged, and in this case, the first layer thickness is negatively correlated with the amount of information in the image region processed by the convolution kernel.

[0110] Since the more the amount of information in the image region processed by the convolution kernel, the higher the accuracy of the feature information extracted by the convolution kernel from the first three-dimensional CT image, the greater the first layer thickness, the greater the size of the first convolution kernel should be, so that the more the amount of information in the image region processed by the first convolution kernel, and thus the accuracy of the feature information extracted by the first convolution kernel from the first three-dimensional CT image can be improved.

[0111] Based on this, the detection apparatus processes the first three-dimensional CT image using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image, in a case where the first layer thickness is positively correlated with the size of the first convolution kernel, so that the first convolution kernel can obtain the first feature information based on more amount of information, thereby improving the accuracy of the first feature information.

[0112] 104. processing the first three-dimensional CT image using the second convolution kernel to obtain the second feature information of the pixels in the first three-dimensional CT image.

[0113] As described in step 103, the greater the first layer thickness, the greater the size of the second convolution kernel should be, so that the second convolution kernel processes more information in the image region, and in turn can improve the accuracy of the feature information extracted by the second convolution kernel from the first three-dimensional CT image. Therefore, the detection device uses the second convolution kernel to process the first three-dimensional CT image in the case that the first layer thickness is positively correlated with the size of the second convolution kernel, and obtains the second feature information of the pixels in the first three-dimensional CT image, which can enable the second convolution kernel to obtain the second feature information based on more information, thereby improving the accuracy of the second feature information.

[0114] 105. Determine, based on the first feature information and the second feature information, a first image region corresponding to the first object from the first three-dimensional CT image.

[0115] In the embodiment of the application, the first image region is an image region corresponding to the first object in the first three-dimensional CT image. Because the size of the first convolution kernel and the size of the second convolution kernel are different, the size of the first feature information obtained based on the first convolution kernel is different from the size of the second feature information obtained based on the second convolution kernel. For example, the size of the first convolution kernel is larger than the size of the second convolution kernel, and the size of the first feature information is larger than the size of the second feature information. Therefore, the detection device determines the first image region from the first three-dimensional CT image based on the first feature information and the second feature information, which can determine the first image region based on feature information of different sizes, and in turn can improve the accuracy of the first image region.

[0116] In Figure 1 In the multi-scale object detection method shown in the figure, the first three-dimensional CT image is obtained by a three-dimensional CT scanning device scanning the first object with a first layer thickness. The detection device determines the first convolution kernel and the second convolution kernel based on the first layer thickness, wherein the size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the first layer thickness, and the size of the first convolution kernel is different from the size of the second convolution kernel. The first three-dimensional CT image is processed using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the first feature information. The first three-dimensional CT image is processed using the second convolution kernel to obtain the second feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the second feature information. In this way, feature information of different sizes can be obtained. Then, based on the first feature information and the second feature information, a first image region corresponding to the first object is determined from the first three-dimensional CT image, which can improve the accuracy of the first image region.

[0117] Optionally, the size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the size of the first object, wherein the size of the first object is the size of the first object in the world coordinate system. In this way, when the first three-dimensional CT image is processed by using the first convolution kernel, the global information of the first object can be better utilized to obtain the first feature information. When the first three-dimensional CT image is processed by using the second convolution kernel, the global information of the first object can be better utilized to obtain the second feature information. For example, the first object is a sphere, and the size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the radius of the first object.

[0118] As an optional implementation, step 103 comprises: obtaining first related information and / or second related information based on the first three-dimensional CT image, wherein the first related information indicates the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of an adjacent pixel, and the second related information indicates the correlation between the gray value of a pixel in the first three-dimensional CT image and the gray value of an adjacent pixel. The first feature information is obtained based on the first related information and / or the second related information.

[0119] In the first three-dimensional CT image, the curvature of the surface of the first object is quite different from the curvature of a region other than the first object (for the sake of brevity, the region other than the first object is referred to as a background region below), i.e., the correlation between the curvature of a pixel belonging to the surface of the first object and the curvature of a pixel in the background region is low. For example, the first object is a sphere, and the surface of the first object is a spherical surface. The correlation between the curvatures of the pixels in the spherical surface is high. However, the correlation between the curvature of a pixel in the spherical surface and the curvature of a pixel in the background region is low. Therefore, based on the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of an adjacent pixel, it can be determined whether the pixel in the first three-dimensional CT image is a pixel belonging to the surface of the first object.

[0120] Optionally, the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of an adjacent pixel can refer to the following example: for example, the first three-dimensional CT image comprises a first pixel and a second pixel, wherein the first pixel is adjacent to the second pixel. The correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of an adjacent pixel comprises the correlation between the curvature of the first pixel and the curvature of the second pixel.

[0121] In the first three-dimensional CT image, the gray value of the surface of the first object is greatly different from the gray value of the background region, i.e., the correlation of the gray value of the pixel belonging to the surface of the first object and the gray value of the pixel in the background region is low. For example, the first object is a sphere, and the surface of the first object is a spherical surface. The correlation of the gray value of the pixel in the spherical surface is high. However, the correlation of the gray value of the pixel in the spherical surface and the gray value of the pixel in the background region is low. Therefore, based on the correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel, it can be determined whether the pixel in the first three-dimensional CT image is the pixel belonging to the surface of the first object.

[0122] Therefore, in the case that the first feature information is obtained based on the first correlation information and / or the second correlation information, the first feature information includes the first correlation information and / or the second correlation information. In this way, the first image region can be determined based on the first feature information, and the first image region can be determined based on the first correlation information and / or the second correlation information, so that the accuracy of the first image region can be improved.

[0123] In an optional implementation, the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel includes at least one of the following: the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in a transverse direction, the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in a longitudinal direction, and the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in a vertical direction, wherein the transverse direction is the direction of the horizontal axis of the image coordinate system of the first three-dimensional CT image, the longitudinal direction is the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image, and the vertical direction is the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image.

[0124] In some schemes, the first object is a sphere, for example, the first object is an optical marker sphere. For example, Figure 2 A schematic diagram in a first three-dimensional CT image of an embodiment of the present application is shown in FIG. 1. In FIG. 1, the first object is an optical marker sphere, and the second object is a human body. The first object is a sphere, and the surface of the first object is a spherical surface. The second object is a human body, and the surface of the second object is a human body surface. Figure 2 In the above-mentioned embodiment, the first object is an optical marker sphere, and the optical marker sphere has six in total. Since the correlation of the curvature of the adjacent two pixels in the surface of the sphere is high in any one of the following directions: the transverse direction, the longitudinal direction, and the vertical direction, if the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel includes the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in the transverse direction, the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in the longitudinal direction, and the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in the vertical direction, the accuracy of the determination of whether the pixel in the first three-dimensional CT image is the pixel belonging to the surface of the first object can be improved based on the correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel.

[0125] Optionally, the correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel comprises at least one of the following: a correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel in the transverse direction, a correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel in the longitudinal direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel in the vertical direction.

[0126] Optionally, the detection device performs convolution processing on the first three-dimensional CT image by using a first convolution kernel to obtain a first convolution processing result. The first convolution processing result is subjected to Hessian matrix analysis to obtain a first Hessian matrix of the pixel in the first three-dimensional CT image, for example, the first Hessian matrix is as follows: , wherein, represents a correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in the transverse direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel in the transverse direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image in the transverse direction and the curvature of the adjacent pixel in the longitudinal direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image in the transverse direction and the gray value of the adjacent pixel in the longitudinal direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image in the transverse direction and the curvature of the adjacent pixel in the vertical direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image in the transverse direction and the gray value of the adjacent pixel in the vertical direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image in the longitudinal direction and the curvature of the adjacent pixel in the transverse direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image in the longitudinal direction and the gray value of the adjacent pixel in the transverse direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image and the curvature of the adjacent pixel in the longitudinal direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image and the gray value of the adjacent pixel in the longitudinal direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image in the longitudinal direction and the curvature of the adjacent pixel in the vertical direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image in the longitudinal direction and the gray value of the adjacent pixel in the vertical direction. represents a correlation of the curvature of the pixel in the first three-dimensional CT image in the vertical direction and the curvature of the adjacent pixel in the transverse direction, and a correlation of the gray value of the pixel in the first three-dimensional CT image in the vertical direction and the gray value of the adjacent pixel in the transverse direction. a correlation of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with a curvature of an adjacent pixel in the vertical direction, a correlation of a gray value of a pixel in the first three-dimensional CT image in the vertical direction with a gray value of an adjacent pixel in the vertical direction. a correlation of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with a curvature of an adjacent pixel in the vertical direction, a correlation of a gray value of a pixel in the first three-dimensional CT image in the vertical direction with a gray value of an adjacent pixel in the vertical direction.

[0127] optionally, a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction. a first-order partial derivative of a curvature of a pixel in the first three-dimensional CT image in the vertical direction with respect to a curvature of an adjacent pixel in the vertical direction.

[0128] Based on the first Hessian matrix, a first eigenvalue, a second eigenvalue, and a third eigenvalue are obtained, wherein the first eigenvalue, the second eigenvalue, and the third eigenvalue are eigenvalues of the first Hessian matrix. The greater the absolute value of the first eigenvalue is, the greater the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of a neighboring pixel in a lateral direction is. The greater the absolute value of the second eigenvalue is, the greater the correlation between the curvature of the pixel in the first three-dimensional CT image and the curvature of the neighboring pixel in a longitudinal direction is. The greater the absolute value of the third eigenvalue is, the greater the correlation between the curvature of the pixel in the first three-dimensional CT image and the curvature of the neighboring pixel in a vertical direction is. Based on the first eigenvalue, the second eigenvalue, and the third eigenvalue, first feature information is obtained, wherein the first feature information includes the first eigenvalue, the second eigenvalue, and the third eigenvalue.

[0129] Optionally, the first object is a sphere. Based on the first eigenvalue, the second eigenvalue, and the third eigenvalue in the first feature information, it is determined whether the pixel in the first three-dimensional CT image is a pixel on a surface of the first object. In some schemes, in a case where the absolute value of the first eigenvalue, the absolute value of the second eigenvalue, and the absolute value of the third eigenvalue are all greater than or equal to a first preset value, it is determined that the pixel in the first three-dimensional CT image is the pixel on the surface of the first object. In other schemes, in a case where a sum of the absolute value of the first eigenvalue, the absolute value of the second eigenvalue, and the absolute value of the third eigenvalue is greater than or equal to a second preset value, it is determined that the pixel in the first three-dimensional CT image is the pixel on the surface of the first object. In yet other schemes, because the correlation between the curvature of two adjacent pixels on the surface of the sphere in three directions (the lateral direction, the longitudinal direction, and the vertical direction) differs less, in a case where any two of the first eigenvalue, the second eigenvalue, and the third eigenvalue differ by less than or equal to a third preset value, it is determined that the pixel in the first three-dimensional CT image is the pixel on the surface of the first object.

[0130] Optionally, in a case where the sum of the absolute value of the first eigenvalue, the absolute value of the second eigenvalue, and the absolute value of the third eigenvalue is greater than or equal to the second preset value, and any two of the first eigenvalue, the second eigenvalue, and the third eigenvalue differ by less than or equal to the third preset value, it is determined that the pixel in the first three-dimensional CT image is the pixel on the surface of the first object. Exemplarily, the second preset value is 20, and the third preset value is 5.

[0131] Optionally, the detection apparatus convolves the first three-dimensional CT image using a second convolution kernel to obtain a second convolution result. The second convolution result is subjected to a Hessian matrix analysis to obtain a second Hessian matrix of the pixel in the first three-dimensional CT image. Based on the second Hessian matrix, a fourth eigenvalue, a fifth eigenvalue and a sixth eigenvalue are obtained, wherein the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue are eigenvalues of the second Hessian matrix. Based on the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue, second feature information is obtained, wherein the second feature information includes the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue.

[0132] Optionally, the first object is a sphere. Based on the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue in the second feature information, it is determined whether the pixel in the first three-dimensional CT image is a pixel on the surface of the first object. Specifically, in a case where the absolute value of the fourth eigenvalue, the absolute value of the fifth eigenvalue and the absolute value of the sixth eigenvalue are all greater than or equal to a first preset value, it is determined that the pixel in the first three-dimensional CT image is a pixel on the surface of the first object.

[0133] Optionally, because the correlation of the curvatures of two adjacent pixels on the surface of the sphere in three directions (lateral direction, longitudinal direction and vertical direction) differs little, in a case where any two of the fourth eigenvalue, the fifth eigenvalue and the sixth eigenvalue differ by less than or equal to a second preset value, it is determined that the pixel in the first three-dimensional CT image is a pixel on the surface of the first object.

[0134] As an optional implementation, the first object is a sphere. Step 105 includes: based on the first feature information, determining a pixel in the first three-dimensional CT image corresponding to the center of the sphere of the first object to obtain n first candidate pixels, wherein n is a positive integer. Based on the first feature information, determining the radius of the first object in the first three-dimensional CT image to obtain n first candidate radii, wherein each first candidate radius corresponds to a first candidate pixel. Based on the second feature information, determining a pixel in the first three-dimensional CT image corresponding to the center of the sphere of the first object to obtain m second candidate pixels, wherein m is a positive integer. Based on the second feature information, determining the radius of the first object in the first three-dimensional CT image to obtain m second candidate radii, wherein each second candidate radius corresponds to a second candidate pixel. Based on the n first candidate pixels and the m second candidate pixels, determining a center pixel, wherein the center pixel is a pixel in the first three-dimensional CT image corresponding to the center of the sphere of the first object. Based on the n first candidate radii and the m second candidate radii, determining a target radius, wherein the target radius is the radius of the first object in the first three-dimensional CT image. Based on the center pixel and the target radius, determining a first image region from the first three-dimensional CT image. For example, Figure 3 A schematic view of a cross section of a first image region provided by an embodiment of the present application. Specifically,Figure 3 The cross section of the first image region shown is circular, i.e. the first image region is a sphere and the first object is a sphere.

[0135] Because the first object is a sphere, the first object can be determined by determining the sphere center of the first object and the first object radius. Based on this, in this embodiment, the detection device first determines n candidate sphere centers (i.e. n first candidate pixels), n candidate radii (i.e. n first candidate radii), m candidate sphere centers (i.e. m second candidate sphere centers), and m candidate radii (i.e. m second candidate radii) from the first three-dimensional CT image based on the first feature information and the second feature information respectively. Then, based on the n first candidate pixels and the m second candidate pixels, the sphere center of the first object (i.e. the sphere center pixel) is determined. Based on the n first candidate radii and the m second candidate radii, the radius of the first object (i.e. the target radius) is determined. Thus, the first image region can be determined from the first three-dimensional CT image based on the sphere center pixel and the target radius.

[0136] It should be understood that, in this embodiment, a first candidate pixel in the n first candidate pixels corresponds to a first candidate radius in the n first candidate radii one by one, i.e. one first candidate pixel and the corresponding one first candidate radius can be used to determine one candidate image region corresponding to the first object from the first three-dimensional CT image. Similarly, one second candidate pixel and the corresponding one second candidate radius can be used to determine one candidate image region corresponding to the first object from the first three-dimensional CT image.

[0137] In some schemes, based on the first feature information, the pixels in the first three-dimensional CT image corresponding to the sphere center of the first object are determined to obtain the n first candidate pixels, including: based on the first feature information, t first surface pixel sets are determined, wherein the first surface pixel set is a set of pixels belonging to the surface of the first object. The pixels in the t first surface pixel sets are respectively fitted to obtain t fourth candidate spheres. Based on the sphere centers of the t fourth candidate spheres, the n first candidate pixels are obtained.

[0138] As described above, based on the first feature information, it can be determined whether a pixel in the first three-dimensional CT image is a pixel belonging to the surface of the first object. Thus, based on the first feature information, t first surface pixel sets can be determined from the first three-dimensional CT image, wherein each first surface region includes a plurality of candidate pixels belonging to the surface of the first object, and t is an integer greater than or equal to n. Then, by fitting the pixels in each first surface pixel set respectively, t fourth candidate spheres are obtained. Thus, based on the sphere centers of the t fourth candidate spheres, the n first candidate pixels are obtained. Optionally, based on the radii of the t fourth candidate spheres, n first candidate radii are obtained.

[0139] Optionally, the detection device realizes the fitting of the pixels in each first surface pixel set by interpolating the pixels in each first surface pixel set, and obtains t fourth candidate spheres. In this way, the distance between two adjacent pixels in the surface of the fourth candidate sphere is less than the distance between two adjacent pixels in the first three-dimensional CT image, and the accuracy of the fourth candidate sphere can be improved, and thus the accuracy of the sphere center determined based on the fourth candidate sphere can be improved.

[0140] Optionally, in the case where the first feature information includes the first feature value, the second feature value and the third feature value, the detection device realizes the fitting of the pixels in each first surface pixel set by interpolating the feature values (including the first feature value, the second feature value and the third feature value) of the pixels in each first surface pixel set, and obtains t fourth candidate spheres. In this way, the accuracy of the fourth candidate sphere can be improved.

[0141] In other schemes, based on the second feature information, the pixel in the first three-dimensional CT image corresponding to the sphere center of the first object is determined, and m second candidate pixels are obtained. The implementation manner can refer to the implementation manner of determining the pixel in the first three-dimensional CT image corresponding to the sphere center of the first object based on the first feature information and obtaining n first candidate pixels, which will not be described herein.

[0142] As an optional implementation, the first feature information includes a first feature value, a second feature value, and a third feature value, and the second feature information includes a fourth feature value, a fifth feature value, and a sixth feature value. Based on the t fourth candidate sphere centers, obtaining the t first candidate pixels includes the following steps: for each of the t fourth candidate sphere centers, performing the following steps: based on a dispersion of the first feature value of the fourth candidate sphere center, the second feature value of the fourth candidate sphere center, and the third feature value of the fourth candidate sphere center, obtaining a sphericity of the fourth candidate sphere center, wherein the sphericity of the fourth candidate sphere center is negatively correlated with the dispersion. Based on a sum of an absolute value of the first feature value of the fourth candidate sphere center, an absolute value of the second feature value of the fourth candidate sphere center, and an absolute value of the third feature value of the fourth candidate sphere center, obtaining a feature value intensity of the fourth candidate sphere center, wherein the feature value intensity is positively correlated with the sum. Based on an absolute value of a difference between a radius of a fourth candidate sphere corresponding to the fourth candidate sphere center and the second candidate radius, obtaining a scale consistency of the fourth candidate sphere center, wherein the scale consistency is negatively correlated with the absolute value of the difference. Based on the sphericity of the fourth candidate sphere center, the feature value intensity of the fourth candidate sphere center, and the scale consistency of the fourth candidate sphere center, obtaining a confidence of the fourth candidate sphere center, wherein the sphericity of the fourth candidate sphere center, the feature value intensity of the fourth candidate sphere center, and the scale consistency of the fourth candidate sphere center are all positively correlated with the confidence of the fourth candidate sphere center. Based on the confidence of the t fourth candidate sphere centers, determining n sphere centers from the t fourth candidate sphere centers as the n first candidate pixels. Optionally, the n sphere centers with the largest confidence are taken as the n first candidate pixels.

[0143] As an optional implementation, the first object is attached to a surface of the second object, and the first three-dimensional CT image includes the second object. For example, the first object is an optical marker ball, and the second object is a human body, and at this time, the optical marker ball is attached to the surface of the human body. In this implementation, based on the n first candidate pixels and the m second candidate pixels, determining the sphere center pixel includes: determining a region corresponding to the surface of the second object from the first three-dimensional CT image to obtain a second image region. Based on the n first candidate pixels and the n first candidate radii, determining n first candidate spheres. Based on the n first candidate spheres, determining n third image regions from the second image region, wherein the first candidate spheres and the third image regions are in one-to-one correspondence, and the third image regions include image regions in contact with the first candidate spheres and image regions within the first candidate spheres. Based on the n third image regions, determining r third candidate pixels from the n first candidate pixels, wherein r is a positive integer less than or equal to n, the r third candidate pixels are sphere centers of r third candidate spheres in the n first candidate spheres, and a ratio of an area of the third image region corresponding to the third candidate sphere to a surface area of the third candidate sphere is less than a first threshold. Based on the r third candidate pixels and the m second candidate pixels, determining the sphere center pixel.

[0144] On the one hand, because the first object is attached to the surface of the second object, the surface of the first object and the surface of the second object are in contact with each other. On the other hand, because the first object is a sphere, there is a region in the surface of the first object that is not in contact with the second object. Therefore, the proportion of the region in the surface of the first object that is in contact with the second object should be less than a certain value. That is, the sphere constructed based on the center of the first object is in contact with the surface of the second object.

[0145] Based on this, in this embodiment, first, n first candidate spheres are determined based on n first candidate pixels and n first candidate radii. Then, based on the proportion of the region in the surface of the first object that is in contact with the second object in the surface of the first object being less than a certain value, it is determined whether the image region corresponding to the first candidate sphere can be used as the image region corresponding to the first object. Specifically, if the ratio of the area of the image region corresponding to the first candidate sphere to the surface area of the first candidate sphere is less than a first threshold value, it indicates that the image region corresponding to the first candidate sphere can be used as the image region corresponding to the first object. Conversely, if the ratio of the area of the image region corresponding to the first candidate sphere to the surface area of the first candidate sphere is greater than or equal to the first threshold value, it indicates that the image region corresponding to the first candidate sphere should not be used as the image region corresponding to the first object. Therefore, the detection device determines the center pixel based on the centers of the r third candidate spheres and the m second candidate pixels, thereby improving the accuracy of the center pixel.

[0146] As an optional embodiment, determining the center pixel based on the r third candidate pixels and the m second candidate pixels comprises: determining r two-dimensional CT images from the first three-dimensional CT image based on the r third candidate pixels, wherein the third candidate pixels correspond one-to-one to the two-dimensional CT images, the two-dimensional CT images do not include the third candidate pixels, the two-dimensional CT images intersect with a second surface pixel set, and the second surface pixel set is a set corresponding to the third candidate pixels in the t first surface pixel sets. The radii of the circles in the r two-dimensional CT images are determined respectively to obtain r circle radii. The distances between the r two-dimensional CT images and the corresponding third candidate pixels are determined respectively to obtain r distances. Based on the r circle radii and the r distances, r third candidate radii are obtained, wherein the third candidate radii are the radii of the first object in the first three-dimensional CT image, and the third candidate radii correspond one-to-one to the third candidate pixels. From the n first candidate radii, r fourth candidate radii corresponding to the r third candidate pixels are determined. Based on the r third candidate radii and the r fourth candidate radii, s fourth candidate pixels are determined from the r third candidate pixels, wherein s is a positive integer less than or equal to r, and the absolute value of the difference between the third candidate radius corresponding to the fourth candidate pixel and the fourth candidate radius corresponding to the fourth candidate pixel is less than a second threshold value. Based on the s fourth candidate pixels and the m second candidate pixels, the center pixel is determined.

[0147] Because for a sphere, the radius of any cross section passing through the center of the sphere, the center of the sphere, and the distance from the cross section to the center of the sphere form a right triangle, the radius of any two-dimensional cross section passing through the center of the sphere, the center of the sphere, and the distance from the two-dimensional cross section to the center of the sphere satisfy the Pythagorean theorem. Therefore, whether the r third candidate pixels can be candidate pixels of the center of the sphere of the first object can be verified based on this property of the sphere.

[0148] Based on this, in the embodiment, the detection device first determines r two-dimensional CT images (i.e., two-dimensional cross sections) from the first three-dimensional CT image based on the r third candidate pixels, then determines the radius of a circle in each of the r two-dimensional CT images to obtain r circle radii, and determines the distance between each of the r two-dimensional CT images and the corresponding third candidate pixel to obtain r distances. Then, based on the r circle radii and the r distances, the detection device obtains r third candidate radii. Because the r fourth candidate radii (i.e., the r radii of the n first candidate radii corresponding to the r third candidate pixels) corresponding to the r third candidate pixels are obtained based on the foregoing, the detection device can compare the r third candidate radii with the r fourth candidate radii. If the third candidate radii and the fourth candidate radii differ by a small amount, it indicates that the third candidate pixels can be candidate pixels of the center of the sphere of the first object, and otherwise, the third candidate pixels cannot be candidate pixels of the center of the sphere of the first object. Therefore, the detection device determines s fourth candidate pixels from the r third candidate pixels according to this principle, and determines the center of the sphere based on the s fourth candidate pixels and the m second candidate pixels, which can improve the accuracy of the center of the sphere.

[0149] Optionally, the detection device determines the circles in the r two-dimensional CT images based on a Hough transform. Then, the detection device determines the radius of each of the circles to obtain the r circle radii.

[0150] Optionally, the second threshold value is positively correlated with the manufacturing tolerance of the first object, and / or the second threshold value is negatively correlated with the quality of the first three-dimensional CT image, wherein the quality of the first three-dimensional CT image is negatively correlated with the area of artifacts in the first three-dimensional CT image and the resolution of the first three-dimensional CT image. For example, in the case where the second threshold value is positively correlated with the manufacturing tolerance of the first object, the second threshold value is less than or equal to 5% of the actual radius of the first object. In the case where the second threshold value is positively correlated with the manufacturing tolerance of the first object and the second threshold value is negatively correlated with the resolution of the first three-dimensional CT image, if the area of artifacts in the first three-dimensional CT image is greater than or equal to an area threshold value, and / or the resolution of the first three-dimensional CT image is less than or equal to a resolution threshold value, the ratio of the second threshold value to the radius of the first object is between 1.05 and 1.1.

[0151] As an optional implementation, the center pixel is determined based on the s fourth candidate pixels and the m second candidate pixels, including: converting coordinates of the s fourth candidate pixels in the image coordinate system of the first three-dimensional CT image into coordinates in the world coordinate system based on a first conversion relationship, to obtain s first coordinates, wherein the first conversion relationship is used to convert coordinates in the image coordinate system of the first three-dimensional CT image into coordinates in the world coordinate system. Converting coordinates of the m second candidate pixels in the image coordinate system of the first three-dimensional CT image into coordinates in the world coordinate system based on the first conversion relationship, to obtain m second coordinates. Clustering the s first coordinates and the m second coordinates to obtain at least one cluster. Obtaining a third coordinate based on coordinates in the at least one cluster. Converting the third coordinate into coordinates in the image coordinate system of the first three-dimensional CT image based on a second conversion relationship, to obtain a fourth coordinate, wherein the second conversion relationship is used to convert coordinates in the world coordinate system into coordinates in the image coordinate system of the first three-dimensional CT image. Determining the center pixel from the first three-dimensional CT image based on the fourth coordinate.

[0152] Optionally, obtaining the third coordinate based on the coordinates in the at least one cluster includes: determining a sum of a confidence degree of the second candidate pixel corresponding to the coordinates in each cluster and a confidence degree of the fourth candidate pixel corresponding to the coordinates in each cluster, to obtain a confidence degree of the cluster. Obtaining the third coordinate based on the coordinates in the cluster with the largest confidence degree.

[0153] Those skilled in the art can understand that, in the above method of the specific implementation, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0154] If the technical solution of the present application involves personal information, the product applying the technical solution of the present application has been explicitly informed of the personal information processing rules before processing the personal information, and has obtained the personal independent consent. If the technical solution of the present application involves sensitive personal information, the product applying the technical solution of the present application has obtained the personal independent consent before processing the sensitive personal information, and at the same time meets the requirement of "explicit consent". For example, at the personal information collection device such as camera, a clear and prominent sign is set to inform that the personal information collection range has been entered and the personal information will be collected. If the person voluntarily enters the collection range, it is regarded as agreeing to collect the personal information. Or, in the case of using obvious signs / information to inform the personal information processing rules on the device for processing personal information, the personal authorization is obtained through the pop-up information or by asking the person to upload his / her personal information. The personal information processing may include personal information processor, personal information processing purpose, processing method and personal information type, etc.

[0155] The above describes the method of the embodiments of the present application in detail, and the device of the embodiments of the present application is provided below.

[0156] Referring to Figure 4 , Figure 4 A structure diagram of a multi-scale target detection device provided by an embodiment of the present application. The multi-scale target detection device 1 comprises an acquisition unit 11 and a processing unit 12, wherein:

[0157] The acquisition unit 11 is configured to acquire a first three-dimensional CT image, wherein the first three-dimensional CT image is obtained by a three-dimensional CT scanning device by scanning a first object with a first layer thickness;

[0158] The processing unit 12 is configured to determine a first convolution kernel and a second convolution kernel based on the first layer thickness, wherein the size of the first convolution kernel and the size of the second convolution kernel are both positively correlated with the first layer thickness, and the size of the first convolution kernel is different from the size of the second convolution kernel;

[0159] The processing unit 12 is further configured to process the first three-dimensional CT image by using the first convolution kernel to obtain first feature information of pixels in the first three-dimensional CT image;

[0160] The processing unit 12 is further configured to process the first three-dimensional CT image by using the second convolution kernel to obtain second feature information of pixels in the first three-dimensional CT image;

[0161] The processing unit 12 is further configured to determine a first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information.

[0162] According to any one of the embodiments of the present application, the processing unit 12 is further configured to:

[0163] obtain first related information and / or second related information based on the first three-dimensional CT image, wherein the first related information indicates the correlation between the curvature of a pixel and the curvature of an adjacent pixel in the first three-dimensional CT image, and the second related information indicates the correlation between the gray value of a pixel and the gray value of an adjacent pixel in the first three-dimensional CT image;

[0164] obtain the first feature information based on the first related information and / or the second related information.

[0165] In combination with any of the embodiments of the present application, the correlation of the curvature of a pixel in the first three-dimensional CT image with the curvature of an adjacent pixel includes at least one of the following: a correlation of the curvature of a pixel in the first three-dimensional CT image with the curvature of an adjacent pixel in a transverse direction, a correlation of the curvature of a pixel in the first three-dimensional CT image with the curvature of an adjacent pixel in a longitudinal direction, and a correlation of the curvature of a pixel in the first three-dimensional CT image with the curvature of an adjacent pixel in a vertical direction, the transverse direction being a direction of a horizontal axis of an image coordinate system of the first three-dimensional CT image, the longitudinal direction being a direction of a vertical axis of the image coordinate system of the first three-dimensional CT image, and the vertical direction being a direction of a vertical axis of the image coordinate system of the first three-dimensional CT image.

[0166] In combination with any of the embodiments of the present application, the first object is a sphere; and the processing unit 12 is further configured to:

[0167] determine, based on the first feature information, a pixel in the first three-dimensional CT image corresponding to a sphere center of the first object, to obtain n first candidate pixels, the n being a positive integer;

[0168] determine, based on the first feature information, a radius of the first object in the first three-dimensional CT image, to obtain n first candidate radii, the first candidate radii corresponding to the first candidate pixels one by one;

[0169] determine, based on the second feature information, a pixel in the first three-dimensional CT image corresponding to the sphere center of the first object, to obtain m second candidate pixels, the m being a positive integer;

[0170] determine, based on the second feature information, a radius of the first object in the first three-dimensional CT image, to obtain m second candidate radii, the second candidate radii corresponding to the second candidate pixels one by one;

[0171] determine, based on the n first candidate pixels and the m second candidate pixels, a sphere center pixel, the sphere center pixel being the pixel in the first three-dimensional CT image corresponding to the sphere center of the first object;

[0172] determine, based on the n first candidate radii and the m second candidate radii, a target radius, the target radius being the radius of the first object in the first three-dimensional CT image;

[0173] determine, based on the sphere center pixel and the target radius, the first image region from the first three-dimensional CT image.

[0174] In combination with any of the embodiments of the present application, the first object is attached to a surface of a second object, and the first three-dimensional CT image includes the second object.

[0175] The processing unit 12 is further configured to:

[0176] determine, from the first three-dimensional CT image, a region corresponding to a surface of the second object, to obtain a second image region;

[0177] determine, based on the n first candidate pixels and the n first candidate radii, n first candidate spheres;

[0178] determine, based on the n first candidate spheres, n third image regions from the second image region, the first candidate spheres correspond to the third image regions one by one, and the third image regions include image regions in contact with the first candidate spheres and image regions within the first candidate spheres;

[0179] determine, based on the n third image regions, r third candidate pixels from the n first candidate pixels, the r is a positive integer less than or equal to the n, and the r third candidate pixels are sphere centers of r third candidate spheres in the n first candidate spheres, a ratio of an area of the third image region corresponding to the third candidate sphere to a surface area of the third candidate sphere is less than a first threshold value;

[0180] determine the sphere center pixel based on the r third candidate pixels and the m second candidate pixels.

[0181] In combination with any one of the embodiments of the present application, the processing unit 12 is further configured to:

[0182] determine, based on the first feature information, t first surface pixel sets, the first surface pixel set is a set of pixels belonging to a surface of the first object, and the t is an integer greater than or equal to the n;

[0183] fit pixels in the t first surface pixel sets respectively to obtain t fourth candidate spheres;

[0184] obtain the n first candidate pixels based on sphere centers of the t fourth candidate spheres.

[0185] In combination with any one of the embodiments of the present application, the processing unit 12 is further configured to:

[0186] determine, based on the r third candidate pixels, r two-dimensional CT images from the first three-dimensional CT image, the third candidate pixels correspond to the two-dimensional CT images one by one, the two-dimensional CT images do not include the third candidate pixels, the two-dimensional CT images intersect with a second surface pixel set, and the second surface pixel set is a set corresponding to the third candidate pixels in the t first surface pixel sets;

[0187] determining radii of circles in the r two-dimensional CT images respectively, to obtain r circle radii;

[0188] determining distances between the r two-dimensional CT images and the third candidate pixels respectively, to obtain r distances;

[0189] obtaining r third candidate radii based on the r circle radii and the r distances, the third candidate radii being radii of the first object in the first three-dimensional CT image, the third candidate radii corresponding to the third candidate pixels one by one;

[0190] determining r fourth candidate radii corresponding to the r third candidate pixels from the n first candidate radii;

[0191] determining s fourth candidate pixels from the r third candidate pixels based on the r third candidate radii and the r fourth candidate radii, the s being a positive integer less than or equal to the r, and an absolute value of a difference between the third candidate radius corresponding to the fourth candidate pixel and the fourth candidate radius corresponding to the fourth candidate pixel being less than a second threshold;

[0192] determining the ball center pixel based on the s fourth candidate pixels and the m second candidate pixels.

[0193] In the embodiments of the present application, the first three-dimensional CT image is obtained by a three-dimensional CT scanning device scanning the first object with a first layer thickness. The detection device determines a first convolution kernel and a second convolution kernel based on the first layer thickness, wherein the size of the first convolution kernel and the size of the second convolution kernel are positively correlated with the first layer thickness, and the size of the first convolution kernel is different from the size of the second convolution kernel. The first three-dimensional CT image is processed by using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the first feature information. The first three-dimensional CT image is processed by using the second convolution kernel to obtain the second feature information of the pixels in the first three-dimensional CT image, which can improve the accuracy of the second feature information. Thus, feature information of different scales can be obtained. Then, the first image region corresponding to the first object is determined from the first three-dimensional CT image based on the first feature information and the second feature information, which can improve the accuracy of the first image region.

[0194] In some embodiments, the apparatus provided by the embodiments of the present application has functions or includes modules which can be used to execute the methods described in the above method embodiments, and the specific implementation can be referred to the description of the above method embodiments. For briefness, details are not described herein.

[0195] Figure 5A hardware structure schematic diagram of an electronic device is provided in the embodiments of the present application. The electronic device 2 comprises a processor 21 and a memory 22. Optionally, the electronic device 2 further comprises an input device 23 and an output device 24. The processor 21, the memory 22, the input device 23 and the output device 24 are coupled through a connector, which comprises various interfaces, transmission lines or buses, etc. The embodiments of the present application do not make any limitation in this aspect. It should be understood that in various embodiments of the present application, coupling means mutual connection in a specific way, including direct connection or indirect connection through other devices, for example, various interfaces, transmission lines, buses, etc.

[0196] The processor 21 can be one or more graphics processing units (GPUs). In the case that the processor 21 is a GPU, the GPU can be a single-core GPU or a multi-core GPU. Optionally, the processor 21 can be a processor group composed of multiple GPUs, and the multiple processors are coupled with each other through one or more buses. Optionally, the processor can also be other types of processors, etc. The embodiments of the present application do not make any limitation in this aspect.

[0197] The memory 22 can be used to store computer program instructions, and various computer program codes for executing the schemes of the present application. Optionally, the memory includes but is not limited to a random access memory (RAM), a read-only memory (ROM), an erasable programmable read only memory (EPROM), or a compact disc read-only memory (CD-ROM), which is used for storing relevant instructions and data.

[0198] The input device 23 is used to input data and / or signals, and the output device 24 is used to output data and / or signals. The input device 23 and the output device 24 can be independent devices, or can be an integral device.

[0199] It can be understood that in the embodiments of the present application, the memory 22 can be used not only to store relevant instructions, but also to store relevant data. The embodiments of the present application do not make any limitation on the data stored in the memory.

[0200] It can be understood that, Figure 5Only a simplified design of an electronic device is shown. In actual applications, the electronic device can also include other necessary elements, including but not limited to any number of input / output devices, processors, memories, etc., and all electronic devices that can implement the embodiments of the present application are within the protection scope of the present application.

[0201] Those skilled in the art can understand that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0202] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here. Those skilled in the art can also clearly understand that each embodiment of the present application describes each focus, and for the convenience and brevity of the description, the same or similar parts in different embodiments can not be described in detail. Therefore, the parts not described or not described in detail in a certain embodiment can be referred to the description of other embodiments.

[0203] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division, and actual implementation can have another division manner. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0204] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0205] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0206] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable apparatus. The computer instructions can be stored in or transmitted by a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that includes one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital versatile disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)) and the like.

[0207] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be instructed by a computer program to complete the relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. The storage medium includes a read-only memory (ROM) or a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

Claims

1. A multi-scale target detection method, characterized in that, The method includes: Acquire a first three-dimensional CT image, wherein the first three-dimensional CT image is obtained by scanning a first object with a first slice thickness using a three-dimensional CT scanning device, and the first object is a sphere; Based on the first layer thickness, a first convolution kernel and a second convolution kernel are determined. The size of the first convolution kernel and the size of the second convolution kernel are both positively correlated with the first layer thickness, and the size of the first convolution kernel and the size of the second convolution kernel are different. The first three-dimensional CT image is processed using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image; The second convolution kernel is used to process the first three-dimensional CT image to obtain the second feature information of the pixels in the first three-dimensional CT image; Based on the first feature information and the second feature information, a first image region corresponding to the first object is determined from the first three-dimensional CT image; the determination of the first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information includes: based on the first feature information, determining the pixel in the first three-dimensional CT image corresponding to the center of the sphere of the first object, obtaining n first candidate pixels, where n is a positive integer; based on the first feature information, determining the radius of the first object in the first three-dimensional CT image, obtaining n first candidate radii, where the first candidate radii correspond one-to-one with the first candidate pixels; based on the second feature information, determining the first image region in the first three-dimensional CT image corresponding to the first object; The pixel corresponding to the center of an object is used to obtain m second candidate pixels, where m is a positive integer; based on the second feature information, the radius of the first object in the first three-dimensional CT image is determined, resulting in m second candidate radii, each corresponding one-to-one with a second candidate pixel; based on the n first candidate pixels and the m second candidate pixels, the center pixel is determined, where the center pixel is the pixel in the first three-dimensional CT image corresponding to the center of the first object; based on the n first candidate radii and the m second candidate radii, the target radius is determined, where the target radius is the radius of the first object in the first three-dimensional CT image; based on the center pixel and the target radius, the first image region is determined from the first three-dimensional CT image.

2. The method according to claim 1, characterized in that, The step of processing the first three-dimensional CT image using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image includes: Based on the first three-dimensional CT image, a first relevant information and / or a second relevant information are obtained. The first relevant information indicates the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of its neighboring pixels, and the second relevant information indicates the correlation between the gray value of a pixel in the first three-dimensional CT image and the gray value of its neighboring pixels. The first feature information is obtained based on the first relevant information and / or the second relevant information.

3. The method according to claim 2, characterized in that, The correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of adjacent pixels includes at least one of the following: the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of adjacent pixels in the horizontal direction, the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of adjacent pixels in the vertical direction, and the correlation between the curvature of a pixel in the first three-dimensional CT image and the curvature of adjacent pixels in the vertical direction, wherein the horizontal direction is the direction of the horizontal axis of the image coordinate system of the first three-dimensional CT image, the vertical direction is the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image, and the vertical direction is the direction of the vertical axis of the image coordinate system of the first three-dimensional CT image.

4. The method according to claim 1, characterized in that, The first object is attached to the surface of the second object, and the first three-dimensional CT image includes the second object; The step of determining the center pixel based on the n first candidate pixels and the m second candidate pixels includes: The region corresponding to the surface of the second object is determined from the first three-dimensional CT image to obtain the second image region; Based on the n first candidate pixels and the n first candidate radii, n first candidate spheres are determined; Based on the n first candidate balls, n third image regions are determined from the second image region. The first candidate balls correspond one-to-one with the third image regions. The third image regions include image regions that are in contact with the first candidate balls and image regions that are inside the first candidate balls. Based on the n third image regions, r third candidate pixels are determined from the n first candidate pixels, where r is a positive integer less than or equal to n, the r third candidate pixels are the centers of the r third candidate spheres among the n first candidate spheres, and the ratio of the area of ​​the third image region corresponding to the third candidate sphere to the surface area of ​​the third candidate sphere is less than a first threshold. The center pixel is determined based on the r third candidate pixels and the m second candidate pixels.

5. The method according to claim 4, characterized in that, Based on the first feature information, the pixel corresponding to the center of the sphere of the first object in the first three-dimensional CT image is determined, resulting in n first candidate pixels, including: Based on the first feature information, t sets of first surface pixels are determined, where the first surface pixel set is the set of pixels belonging to the surface of the first object, and t is an integer greater than or equal to n; Each of the t first surface pixel sets is fitted to obtain t fourth candidate spheres; Based on the centers of the t fourth candidate spheres, the n first candidate pixels are obtained.

6. The method according to claim 5, characterized in that, The step of determining the center pixel based on the r third candidate pixels and the m second candidate pixels includes: Based on the r third candidate pixels, r two-dimensional CT images are determined from the first three-dimensional CT image. The third candidate pixels correspond one-to-one with the two-dimensional CT images. The two-dimensional CT images do not include the third candidate pixels. The two-dimensional CT images intersect with the second surface pixel set. The second surface pixel set is the set of the t first surface pixel sets that corresponds to the third candidate pixels. The radii of the circles in the r two-dimensional CT images are determined respectively, resulting in r circle radii; The distances between the r two-dimensional CT images and the corresponding third candidate pixels are determined respectively, resulting in r distances; Based on the r circle radii and the r distances, r third candidate radii are obtained. The third candidate radii are the radii of the first object in the first three-dimensional CT image, and the third candidate radii correspond one-to-one with the third candidate pixels. From the n first candidate radii, determine r fourth candidate radii corresponding to the r third candidate pixels; Based on the r third candidate radii and the r fourth candidate radii, s fourth candidate pixels are determined from the r third candidate pixels, where s is a positive integer less than or equal to r, and the absolute value of the difference between the third candidate radius corresponding to the fourth candidate pixel and the fourth candidate radius corresponding to the fourth candidate pixel is less than a second threshold. The center pixel is determined based on the s fourth candidate pixels and the m second candidate pixels.

7. A multi-scale target detection device, characterized in that, The multi-scale target detection device includes: The acquisition unit is used to acquire a first three-dimensional CT image, wherein the first three-dimensional CT image is obtained by scanning a first object with a first layer thickness using a three-dimensional CT scanning device, and the first object is a sphere; The processing unit is configured to determine a first convolutional kernel and a second convolutional kernel based on the first layer thickness, wherein the size of the first convolutional kernel and the size of the second convolutional kernel are both positively correlated with the first layer thickness, and the size of the first convolutional kernel and the size of the second convolutional kernel are different. The processing unit is further configured to process the first three-dimensional CT image using the first convolution kernel to obtain the first feature information of the pixels in the first three-dimensional CT image. The processing unit is further configured to process the first three-dimensional CT image using the second convolution kernel to obtain the second feature information of the pixels in the first three-dimensional CT image; The processing unit is further configured to determine a first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information; the step of determining the first image region corresponding to the first object from the first three-dimensional CT image based on the first feature information and the second feature information includes: determining the pixel corresponding to the center of the sphere of the first object in the first three-dimensional CT image based on the first feature information, obtaining n first candidate pixels, where n is a positive integer; determining the radius of the first object in the first three-dimensional CT image based on the first feature information, obtaining n first candidate radii, where the first candidate radii correspond one-to-one with the first candidate pixels; and determining the first image region corresponding to the first object in the first three-dimensional CT image based on the second feature information. The pixels corresponding to the center of the first object in the image are used to obtain m second candidate pixels, where m is a positive integer; based on the second feature information, the radius of the first object in the first three-dimensional CT image is determined, resulting in m second candidate radii, each corresponding to a second candidate pixel; based on the n first candidate pixels and the m second candidate pixels, the center pixel is determined, where the center pixel is the pixel in the first three-dimensional CT image corresponding to the center of the first object; based on the n first candidate radii and the m second candidate radii, the target radius is determined, where the target radius is the radius of the first object in the first three-dimensional CT image; based on the center pixel and the target radius, a first image region is determined from the first three-dimensional CT image.

8. A surgical robot, characterized in that, Includes the multi-scale target detection device as described in claim 7.

9. An electronic device, characterized in that, include: A processor and a memory, the memory being used to store computer program code, the computer program code including computer instructions, wherein, when the processor executes the computer instructions, the electronic device performs the method as described in any one of claims 1 to 6.

Images