WO2023138544A1

WO2023138544A1 - Capsule endoscope intestinal image-based recognition and positioning method, storage medium, and device

Info

Publication number: WO2023138544A1
Application number: PCT/CN2023/072477
Authority: WO
Inventors: 李鹏; 胡峰; 谢浩生; 刘揆亮; 刘一君
Original assignee: 江苏势通生物科技有限公司
Priority date: 2022-01-18
Filing date: 2023-01-17
Publication date: 2023-07-27
Also published as: CN114066781A; CN114066781B

Abstract

Disclosed are a capsule endoscope intestinal image-based recognition and positioning method, a storage medium, and a device. The method comprises: sequentially performing rotation correction processing on original intestinal images captured by a capsule endoscope to obtain corrected images; sequentially performing cropping and measurement processing on the corrected images to obtain intestinal inner wall annular images corresponding to the corrected images and obtain physical widths of the intestinal inner wall annular images; unfolding the intestinal inner wall annular images and then sequentially splicing the unfolded intestinal inner wall annular images end to end in a pixel width direction to form an intestinal panoramic image; marking a region of interest in the intestinal panoramic image according to an obtained instruction, and determining pixel coordinates of the region of interest in the intestinal panoramic image; and determining the physical position of the region of interest in the intestinal tract according to the pixel coordinates of the region of interest, and the pixel widths and the physical widths of the intestinal inner wall annular images. Therefore, the image reading and diagnosis processes are visual and convenient, and the clinical position can be quickly and accurately determined.

Description

Recognition and positioning method, storage medium and equipment for intestinal image of capsule endoscope

This application claims the priority of the Chinese patent application with application number 202210052705.2, the contents of which are incorporated in this application by reference.

【Technical field】

The invention belongs to the technical field of medical equipment imaging, and in particular relates to a method for identifying and locating intestinal images of a capsule endoscope, a computer-readable storage medium, and computer equipment.

【Background technique】

Capsule endoscope is a kind of medical equipment. The capsule endoscope integrates core functions such as image acquisition and wireless transmission into a capsule that can be swallowed by the human body. During the inspection process, the capsule endoscope is swallowed into the body, and the endoscope collects images of the digestive tract in the body and transmits them synchronously to the outside of the body for medical inspection and diagnosis based on the obtained image data.

Capsule endoscopes acquire and transmit tens of thousands of images as they work in the small intestine. The traditional diagnostic method and process is to present these tens of thousands of images to medical workers for observation through picture playback or video playback. The whole process takes a long time, and the location of suspicious lesions identified and obtained by medical workers in the intestinal tract cannot be determined, which greatly affects the use and diagnostic efficiency of intestinal endoscopy.

【Content of invention】

(1) technical problem to be solved by the present invention

The technical problem solved by the invention is: how to quickly and efficiently identify the region of interest in the intestinal tract and determine the position of the region of interest.

(2) The technical scheme adopted in the present invention

A method for identifying and positioning a capsule endoscope intestinal image, the identifying and positioning method comprising:

Sequentially performing rotation correction processing on each original intestinal image captured by the capsule endoscope to obtain a corrected image, wherein the angles of view of each corrected image are the same;

Perform interception and measurement processing on each of the corrected images in turn to obtain a circular image of the inner wall of the intestinal tract corresponding to each of the corrected images and obtain a physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract;

Stitching each circular image of the inner wall of the intestinal tract sequentially end to end according to the pixel width direction to form a panoramic image of the intestinal tract;

Marking the region of interest in the intestinal panorama image according to the acquired instruction, and determining the pixel coordinates of the interest region in the intestinal panoramic image;

The physical position of the region of interest in the intestinal tract is determined according to the pixel coordinates of the region of interest, the pixel width of each circular image of the inner wall of the intestinal tract, and the physical width corresponding to the pixel width of the circular images of the inner wall of the intestinal tract.

Preferably, the method of performing rotation correction processing on each original intestinal tract image captured by the capsule endoscope to obtain a corrected image includes:

performing feature point matching on the image to be corrected and a corrected image adjacent to the image to be corrected, to obtain a first set of feature points of the image to be corrected and a second set of feature points of the corrected image;

determining a relative rotation angle between the image to be corrected and the corrected image according to the first set of feature points of the image to be corrected and the second set of feature points of the corrected image;

Correcting each pixel point value of the image to be corrected according to the difference between the first included angle and the second included angle to obtain a corrected image.

Acquiring the first gravitational acceleration value at the acquisition moment of the image to be corrected and the second gravitational acceleration value at the acquisition moment of the corrected image adjacent to the image to be corrected;

determining a relative rotation angle between the image to be corrected and the corrected image according to the first gravitational acceleration value and the second gravitational acceleration value;

Correcting each pixel point value of the image to be corrected according to the relative rotation angle to obtain a corrected image.

Preferably, the correction image is subjected to interception and measurement processing, and the method for obtaining the ring image of the inner wall of the intestinal tract corresponding to the correction image includes:

determining the circular edge of the intercepted measurement image adjacent to the measurement image to be intercepted, and extracting the target edge of the circular edge in the measurement image to be intercepted;

The difference between the radius of the circular edge and the radius of the target edge is used as the interception width, and a partial image is intercepted from the measurement image to be intercepted according to the interception width, as an annular image of the inner wall of the intestinal tract, and the pixel width of the annular image of the inner wall of the intestinal tract is equal to the interception width

Preferably, said determining the circular edge of the intercepted measurement image adjacent to the measurement image to be intercepted, The method for extracting the target edge of the circular edge in the measurement image to be intercepted includes:

calculating the first discrete quantized pixel gray value sequence of the circular edge;

Selecting a candidate edge in the measurement image to be intercepted, and calculating a second discrete quantized pixel gray value sequence of the candidate edge;

calculating the Levenstein distance between the first discrete quantized pixel gray value sequence and the second discrete quantized pixel gray value sequence;

The radius of the candidate edge is adjusted sequentially until the Levenstein distance satisfies the predetermined condition, and the candidate edge at this time is taken as the target edge.

Preferably, the method for obtaining the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract comprises:

According to the acquired radius of the front cover of the capsule endoscope, the distance between the equivalent optical center of the capsule endoscope and the bottom of the front cover, the radius of the circular edge and the radius of the target edge, the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract is calculated.

Preferably, the method for determining the physical position of the region of interest in the intestinal tract according to the pixel coordinates of the region of interest, the pixel width of each annular image of the inner wall of the intestinal tract, and the physical width corresponding to the pixel width of the annular images of the inner wall of the intestinal tract includes:

Determine the picture number of the current circular image of the intestinal wall corresponding to the pixel coordinates according to the pixel coordinates of the region of interest and the pixel width of each of the circular images of the intestinal wall;

Calculate the sum of the pixel widths and the sum of the physical widths of the annular images of the inner wall of the intestinal tract before the sequence number of the pictures;

The physical position of the region of interest in the intestinal tract is determined according to the pixel coordinates, the sum of the pixel widths, the sum of the physical widths, and the physical width of the current circular image of the inner wall of the intestinal tract.

Preferably, the method for marking the region of interest in the intestinal panorama image according to the acquired instruction includes:

Operating the intestinal panorama image according to at least one of a scrolling instruction, a sliding instruction, a zoom-in instruction, and a zoom-out instruction, to determine a region of interest;

The region of interest is marked according to one of a single-click instruction, a double-click instruction, and a frame selection instruction.

The present application also discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a recognition and positioning program of the capsule endoscope intestinal image, and when the capsule endoscope intestinal image recognition and positioning program is executed by a processor, the above-mentioned capsule endoscope intestinal image recognition and positioning method is realized.

The present application also discloses a computer device. The computer device includes a computer-readable storage medium, a processor, and a program for identifying and locating the intestinal tract images of the capsule endoscope stored in the computer-readable storage medium. When the program for identifying and locating the intestinal tract images of the capsule endoscope is executed by the processor, the above-mentioned method for identifying and locating the intestinal tract images of the capsule endoscope is implemented.

(3) Beneficial effects

Compared with the traditional method, a method for identifying and locating the intestinal tract image of the capsule endoscope disclosed by the present invention has the following technical effects:

By splicing the original intestinal pictures to form a panoramic picture of the intestinal tract, the whole process of film reading and diagnosis is more intuitive and convenient. In addition, according to the pixel coordinates of the region of interest and the physical length of each circular image of the inner wall of the intestinal tract, the clinical location of the region of interest can be quickly and accurately determined.

【Description of drawings】

1 is a flowchart of a method for identifying and locating a capsule endoscope intestinal image according to Embodiment 1 of the present invention;

FIG. 2 is a schematic diagram of image rotation correction in Embodiment 1 of the present invention;

FIG. 3 is a schematic diagram of target edge extraction according to Embodiment 1 of the present invention;

Fig. 4 is a schematic diagram of the physical width measurement of the annular image of the inner wall of the intestinal tract according to Embodiment 1 of the present invention;

Fig. 5 is a schematic diagram of interception and expansion of the annular image of the inner wall of the intestinal tract in Embodiment 1 of the present invention;

Fig. 6 is a schematic diagram of splicing and reconstruction of multiple circular images of the inner wall of the intestinal tract according to Embodiment 1 of the present invention;

FIG. 7 is a schematic diagram of computer equipment according to Embodiment 2 of the present invention.

【Detailed ways】

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

Before describing the various embodiments of the present application in detail, first briefly describe the technical concept of the present application: in the prior art, identification and diagnosis are performed based on intestinal images taken by capsule endoscopes, mainly through video or image reading. Due to the large number of images, the entire process of reading images takes a long time, and for areas of interest such as suspicious lesions, the specific location in the intestinal tract cannot be determined. To this end, the identification and positioning method of the capsule endoscope intestinal image provided by this application, after performing rotation correction processing on the original intestinal image, After the interception and measurement of each image, it is expanded and spliced to form a panoramic image of the intestinal tract. According to the relevant operation instructions, the film is directly read in the panoramic image of the intestinal tract, and the region of interest is identified and marked. Combined with the pixel coordinates of the region of interest and the pixel width and physical width obtained from the interception and measurement of each image, the physical position of the region of interest in the intestinal tract is determined. The whole process of image reading and diagnosis is more intuitive and convenient, and the clinical location of the region of interest can be quickly and accurately determined.

Specifically, as shown in FIG. 1, the identification and positioning method of the capsule endoscope intestinal image in the first embodiment includes the following steps:

Step S10: Perform rotation correction processing on the original intestinal images captured by the capsule endoscope to obtain corrected images, wherein the angle of view of each corrected image is the same.

Step S20: Intercepting and measuring each corrected image sequentially to obtain the annular image of the inner wall of the intestine corresponding to each corrected image and the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract.

Step S30 : unfolding the circular images of the inner wall of the intestinal tract and sequentially splicing them according to the pixel width direction to form a panoramic image of the intestinal tract.

Step S40: mark the ROI in the intestinal panorama image according to the acquired instructions, and determine the pixel coordinates of the ROI in the intestinal panorama image.

Step S50: Determine the physical position of the region of interest in the intestinal tract according to the pixel coordinates of the region of interest, the pixel width of each annular image of the inner wall of the intestinal tract, and the corresponding physical width of the pixel width of the annular image of the inner wall of the intestinal tract.

Specifically, in step S10, when using the capsule endoscope to take images of the intestinal tract, since the capsule endoscope still rotates around its own axis during the movement process, the captured image will be deflected, wherein the self-axis here refers to the forward direction of the capsule endoscope, so rotation correction processing is required. For the convenience of description, the original image set composed of each original intestinal image I _n is denoted as {I _n }, 1≤n≤N, N represents the total number of acquired images, and the resolution of the original intestinal image is D×D. Exemplarily, the photographing direction of the capsule endoscope in this embodiment is opposite to the moving direction, that is, the capsule endoscope photographs the inner wall of the intestinal tract behind it during the advancing process. The rotation correction process can be carried out in two ways, one is to correct based on the characteristics of the image itself, and the other is to correct based on sensor data.

The method for correcting based on the image itself includes the following steps:

Step S101, performing feature point matching on the image to be corrected and the corrected image adjacent to the image to be corrected, to obtain the first feature point set of the image to be corrected and the second feature point set of the corrected image set.

Assuming that the corrected image in the original image set {I _n } is I _n → I' _n , the image to be corrected is I _n+1 , where the first image (n=1) does not need to be corrected, the image to be corrected is I n ₊ ₁ and the corrected image I' _n , and the feature points are matched to obtain M pairs of feature points, where the set of M pairs of feature points in the image to be corrected I _n+1 is represented as the first feature point set {G _m } _n+1 , 1≤m≤M, M pairs of feature points in the already The set of corrected images I' n is expressed as the second feature point set {G _m } _n , 1≤m≤M,

Step S102. Determine the relative rotation angle between the image to be corrected and the corrected image according to the first set of feature points of the image to be corrected and the second set of feature points of the corrected image.

As shown in FIG. 2 , the first center point of the first feature point set is denoted as C _n+1 , and the second center point of the second feature point set is denoted as C _n . In the image to be corrected is I _n+1 , the angle between the line connecting the first center point C _n+1 to the image center point and the horizontal direction of the image center is α _n+1 ; in the image to be corrected is I' _n , the angle between the line connecting the second center point C _n to the image center point and the horizontal direction of the image center is α _n . The relative rotation angle between the image to be corrected and the corrected image is:

Δα＝α _n+1 -α _n

Among them, when Δα>0, it means that the image to be corrected is I _n+1 and needs to be rotated clockwise to achieve perspective alignment with the image to be corrected which is I'_n; when Δα<0, it means that the image to be corrected is I _n+1 and needs to be rotated counterclockwise to align with the image to be corrected which is I' _n .

Step S103 , correcting each pixel value of the image to be corrected according to the difference between the first included angle and the second included angle, to obtain a corrected image.

The image to be corrected is I _n+1 and the image after rotation correction is expressed as I′ _n+1 . For any pixel value p'(u',v')∈I' _n+1 , the corresponding pixel value p(u,v)∈I _n+1 can be corrected according to the following formula,

Among them, round(□) means rounding. In other implementation manners, interpolation may also be used to obtain the corrected image pixel value.

The method for correcting based on sensor data includes the following steps:

Step S111 , acquiring the first gravitational acceleration value at the acquisition time of the image to be corrected and the second gravitational acceleration value at the acquisition time of the corrected image adjacent to the image to be corrected.

Assuming that the corrected image in the original image set {I _n } is I _n → I' _n , the image to be corrected is I _n+1 , where the first image (n=1) does not need to be corrected, the image to be corrected is I _n+1 and the corrected image I' _n . Obtain the first gravitational acceleration value measured by the gravity sensor at the acquisition moment of the corrected image I' _n :

The second gravitational acceleration value measured by the gravity sensor at the acquisition moment of the image I _n+1 to be corrected:

in, are the gravity acceleration readings of the gravity sensor on the u-axis, v-axis and r-axis at the time when the corrected image I' _n is collected, respectively. are the readings of the gravity acceleration of the gravity sensor on the u-axis, v-axis and r-axis at the acquisition time of the image I _n+1 to be corrected, respectively.

Step S112. Determine the relative rotation angle between the image to be corrected and the corrected image according to the first gravitational acceleration value and the second gravitational acceleration value.

Specifically, the relative rotation angle between the image to be corrected and the corrected image is calculated according to the following formula:

Step S113 , correcting each pixel point value of the image to be corrected according to the relative rotation angle to obtain a corrected image.

Among them, round(□) means rounding. In other implementation manners, interpolation may also be used to obtain the corrected image pixel value. The corrected image set obtained after rotation correction is denoted as {I' _n }, 1≤n≤N.

In step S20, the correction image is intercepted and measured, and the method for obtaining the ring image of the inner wall of the intestinal tract corresponding to the correction image includes the following steps:

Step S21 , determining the circular edge of the captured measurement image adjacent to the measurement image to be captured, and extracting the target edge of the circular edge in the measurement image to be captured.

Step S22, taking the difference between the radius of the circular edge and the radius of the target edge as the interception width, and intercepting a part of the image from the measurement image to be intercepted according to the interception width, as an annular image of the inner wall of the intestinal tract, and the pixel width of the annular image of the inner wall of the intestinal tract is equal to the interception width.

Wherein, the corrected image set obtained after rotation correction is expressed as {I' _n }, 1≤n≤N. No processing is performed on the first image, as shown in Figure 3, assuming that the circular edge of the intercepted measurement image _I'n is represented as 201, the position of the circular edge 201 in the measurement image I'n ₊₁ to be intercepted is the target edge 202. The circular edge 201 is a circle with the center of the intercepted measurement image as the center and a radius of _r1 ; the target edge 202 is a circle with the center of the measurement image to be intercepted as the center and a radius of _r2 . Therefore, the process of extracting the target edge 202 can be regarded as a process of determining the radius r ₂ . Among them, r ₁ and r ₂ are pixel lengths.

Specifically, step S21 includes the following processes:

Step S211, calculating the first discrete quantized pixel gray value sequence E _n of the circular edge, the calculation formula is as follows:

Step S212, selecting a candidate edge in the measurement image to be intercepted, and calculating the second discrete quantized pixel gray value sequence of the candidate edge Calculated as follows:

Wherein, r _v represents the radius of the candidate edge, the center of the candidate edge is the center of the measurement image to be intercepted, 1≤Q≤255 is the discrete quantization scale, the larger the Q value corresponds to the faster calculation efficiency, but the accuracy will decrease, preferably, Q=26.

Step S213, calculating the Levenstein distance between the first discrete quantized pixel gray value sequence and the second discrete quantized pixel gray value sequence Calculated as follows:

Among them, Lev(a,b) means to calculate the Levenstein distance between sequences a and b.

Step S214 , adjusting the radius of the candidate edge in turn until the Levenstein distance satisfies the predetermined condition, and the candidate edge at this time is taken as the target edge.

when r _v starts to decrease from r ₁ -1, if it satisfies:

Select r _v at this time as r ₂ , that is, the corresponding candidate edge is the target edge 202 .

If reduced to If the above conditions still cannot be met, select The corresponding r _v is used as the position of the target edge 202 , that is, r ₂ =r _v .

After the radius r ₂ of the target edge 202 is determined, the interception width W=r ₁ -r ₂ can be obtained, according to the interception width, a partial image can be intercepted from the measurement image I'n ₊₁ to be intercepted as the annular image I″ _n+1 of the inner wall of the intestinal tract.

Further, the method for obtaining the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract is as follows:

As shown in FIG. 4 , it is assumed that the radius of the front cover 301 of the capsule endoscope is R, and the distance between the lens equivalent optical center 303 and the bottom diameter of the front cover 301 is d. CO-1 is the maximum viewing angle edge line of sight of the lens, corresponding to the imaging edge whose radius is r1 in the measurement image I'n ₊₁ to be intercepted, CO-2 is the line of sight corresponding to the imaging circle whose radius is _r2 in the measurement image I'n ₊₁ to be intercepted, and the distance to be measured ₃₀₄ is the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract. Given that the maximum viewing angle of the lens is FOV _max , the length of the distance to be measured 304 can be calculated as:

l _n+1 ＝|φ ₁ -φ ₂ | _rad R

in,

and

The obtained circular image set of the inner wall of the intestinal tract is expressed as {I” _n }, 1≤n≤N, and the obtained physical width data set is expressed as {l _n+1 }, 1≤n≤N-1.

Further, in step S30 , the circular images of the inner wall of the intestinal tract are unfolded and spliced sequentially according to the pixel width direction to form a panoramic image of the intestinal tract.

Specifically, as shown in Fig. 5, the resolution of the measurement image I'n ₊₁ to be intercepted is D*D, the radius of the circular edge is r ₁ =D/2, and the resolution of the ring image I″ _n+1 of the inner wall of the intestinal tract is W*H, where:

W=r ₁ -r ₂

H=round(2πD)

For any pixel P" _n+1 (u", v") of the annular image I" _n+1 of the inner wall of the intestinal tract, its value is obtained from the corresponding pixel P' _n+1 (u', v') in the measurement image I' _n+1 to be intercepted, where

In other embodiments, the value of P" _n+1 (u", v") can be obtained from P' _n+1 (u', v') through interpolation.

As shown in Fig. 6, for the obtained intestinal inner wall ring image collection representation {I” _n }, 1≤n≤N, the pixel height of each intestinal inner wall ring image I” _n is H _n , the pixel width is W _n , and the physical width is l _n . Wherein, the unit of H _n and W _n is the number of pixels, and the unit of H n is millimeter or decimeter.

Wherein, H ₁ =H ₂ =...=H _n =H _N =round(2πD)

The intestinal inner wall ring image set {I” _n }, 1≤n≤N each image is spliced end to end according to the pixel width direction to form a intestinal panoramic image I _F , the pixel height of the intestinal panoramic image I _F is H _F , and the pixel width is W _F , and

H _F = round(2πD)

In step S40, the method for marking the region of interest in the intestinal panorama image according to the acquired instruction includes the following steps:

Step S41: Operate the intestinal panorama image according to at least one of a scrolling instruction, a sliding instruction, a zoom-in instruction and a zoom-out instruction, so as to determine a region of interest.

Since the image resolution of the reconstructed intestinal panorama image exceeds the resolution of the display screen, the operator can operate the image by scrolling, sliding, zooming in and zooming out, etc., which is convenient for film reading and diagnosis, and can quickly determine clinical positioning marks, suspected lesions and other areas of interest.

Step S42: Mark the region of interest according to one of a single-click instruction, a double-click instruction, and a frame selection instruction.

Wherein, the number of regions of interest can be multiple, and after the image reading is completed, K regions of interest {T _k } can be formed, 1≤k≤K, and the pixel coordinates of each region of interest T _k in the intestinal panorama image _IF are in

Further, in step S50, according to the pixel coordinates of the region of interest T _k The method for determining the physical position L _k of the region of interest T _k in the intestinal _tract by determining the pixel width W _n of each annular image of the inner wall of the intestinal tract and the corresponding physical width l n of the pixel width W _n of the annular images of the inner wall of the intestinal tract includes the following steps:

Step S51, according to the pixel coordinates of the region of interest T _k and the pixel width W _n of each intestinal wall ring image is determined with the pixel coordinates Corresponding to the current ring image of the inner wall of the intestinal tract image number Specifically, it is determined according to the following formula:

Step S52, calculate the serial number of the picture The sum of the pixel widths of the previous ring images of the inner wall of the intestinal tract and the sum of the physical width

Step S53, according to the pixel coordinates sum of pixel widths sum of physical widths And the current circular image of the inner wall of the intestine physical width of Determining the physical location of the region of interest in the gut The position L _k is specifically determined according to the following formula:

Among them, the above-mentioned processing is performed on all regions of interest {T _k }, 1≤k≤K, and the original picture number set of the region of interest can be obtained The positioning data set of the corresponding ROI is {L _k }, 1≤k≤K.

Embodiment 2 of the present application also discloses a computer-readable storage medium. The computer-readable storage medium stores a recognition and positioning program for capsule endoscope intestinal images. When the capsule endoscope intestinal image recognition and positioning program is executed by a processor, the above-mentioned capsule endoscope intestinal image recognition and positioning method is realized.

Embodiment 3 also discloses a computer device. At the hardware level, as shown in FIG. 7 , the computer device includes a processor 12 , an internal bus 13 , a network interface 14 , and a computer-readable storage medium 11 . The processor 12 reads the corresponding computer program from the computer-readable storage medium and executes it, forming a request processing device on a logical level. Of course, in addition to software implementations, one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each logic unit, and may also be hardware or logic devices. The computer-readable storage medium 11 stores a capsule endoscope intestinal image recognition and positioning program, which realizes the capsule endoscope intestinal image recognition and positioning method when executed by the processor.

Computer-readable storage media includes both volatile and non-permanent, removable and non-removable media by any method or technology for storage of information. Information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum memory, graphene-based storage media or other magnetic storage device or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

The specific embodiments of the present invention have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that without departing from the claims and their equivalents, the Under the condition of the principle and spirit of the present invention within the scope, modifications and improvements can be made to these embodiments, and these modifications and improvements should also be within the protection scope of the present invention.

Claims

A method for identifying and positioning capsule endoscope intestinal images, characterized in that the method for identifying and positioning comprises:

Sequentially performing rotation correction processing on each original intestinal image captured by the capsule endoscope to obtain a corrected image, wherein the angles of view of each corrected image are the same;

Perform interception and measurement processing on each of the corrected images in turn to obtain a circular image of the inner wall of the intestinal tract corresponding to each of the corrected images and obtain a physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract;

Stitching each circular image of the inner wall of the intestinal tract sequentially end to end according to the pixel width direction to form a panoramic image of the intestinal tract;

Marking the region of interest in the intestinal panorama image according to the acquired instruction, and determining the pixel coordinates of the interest region in the intestinal panoramic image;

The physical position of the region of interest in the intestinal tract is determined according to the pixel coordinates of the region of interest, the pixel width of each circular image of the inner wall of the intestinal tract, and the physical width corresponding to the pixel width of the circular images of the inner wall of the intestinal tract.
The method for identifying and locating intestinal images of a capsule endoscope according to claim 1, wherein the method of performing rotation correction processing on each original intestinal image captured by a capsule endoscope to obtain a corrected image comprises:

performing feature point matching on the image to be corrected and a corrected image adjacent to the image to be corrected, to obtain a first set of feature points of the image to be corrected and a second set of feature points of the corrected image;

determining a relative rotation angle between the image to be corrected and the corrected image according to the first set of feature points of the image to be corrected and the second set of feature points of the corrected image;

Correcting each pixel point value of the image to be corrected according to the difference between the first included angle and the second included angle to obtain a corrected image.
The method for identifying and locating intestinal images of a capsule endoscope according to claim 1, wherein the method of performing rotation correction processing on each original intestinal image captured by a capsule endoscope to obtain a corrected image comprises:

Acquiring the first gravitational acceleration value at the acquisition moment of the image to be corrected and the second gravitational acceleration value at the acquisition moment of the corrected image adjacent to the image to be corrected;

determining the image to be corrected according to the first gravitational acceleration value and the second gravitational acceleration value relative rotation angle to the corrected image;

Correcting each pixel point value of the image to be corrected according to the relative rotation angle to obtain a corrected image.
The method for identifying and locating the intestinal tract image of the capsule endoscope according to claim 1, wherein the method of intercepting and measuring the corrected image, and obtaining the circular image of the inner wall of the intestinal tract corresponding to the corrected image comprises:

determining the circular edge of the intercepted measurement image adjacent to the measurement image to be intercepted, and extracting the target edge of the circular edge in the measurement image to be intercepted;

The difference between the radius of the circular edge and the radius of the target edge is used as the interception width, and a partial image is intercepted from the measurement image to be intercepted according to the interception width, as an annular image of the inner wall of the intestinal tract, and the pixel width of the annular image of the inner wall of the intestinal tract is equal to the interception width.
The method for identifying and locating the intestinal tract image of the capsule endoscope according to claim 4, wherein the method of determining the circular edge of the intercepted measurement image adjacent to the measurement image to be intercepted, and extracting the target edge of the circular edge in the measurement image to be intercepted comprises:

calculating the first discrete quantized pixel gray value sequence of the circular edge;

Selecting a candidate edge in the measurement image to be intercepted, and calculating a second discrete quantized pixel gray value sequence of the candidate edge;

calculating the Levenstein distance between the first discrete quantized pixel gray value sequence and the second discrete quantized pixel gray value sequence;

The radius of the candidate edge is adjusted sequentially until the Levenstein distance satisfies the predetermined condition, and the candidate edge at this time is taken as the target edge.
The method for identifying and positioning the intestinal tract image of the capsule endoscope according to claim 4, wherein the method for obtaining the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract comprises:

According to the acquired radius of the front cover of the capsule endoscope, the distance between the equivalent optical center of the capsule endoscope and the bottom of the front cover, the radius of the circular edge and the radius of the target edge, the physical width corresponding to the pixel width of the annular image of the inner wall of the intestinal tract is calculated.
The method for identifying and locating intestinal images of the capsule endoscope according to claim 4, wherein the method for determining the physical position of the region of interest in the intestinal tract according to the pixel coordinates of the region of interest, the pixel width of each annular image of the inner wall of the intestinal tract, and the physical width corresponding to the pixel width of the annular images of the inner wall of the intestinal tract comprises:

According to the pixel coordinates of the region of interest and the pixel width of each circular image of the inner wall of the intestinal tract Determine the picture number of the current circular image of the inner wall of the intestinal tract corresponding to the pixel coordinates;

Calculate the sum of the pixel widths and the sum of the physical widths of the annular images of the inner wall of the intestinal tract before the sequence number of the pictures;

The physical position of the region of interest in the intestinal tract is determined according to the pixel coordinates, the sum of the pixel widths, the sum of the physical widths, and the physical width of the current circular image of the inner wall of the intestinal tract.
The identification and positioning method of the intestinal tract image of the capsule endoscope according to claim 4, wherein the method of marking the region of interest in the intestinal panoramic image according to the acquired instruction comprises:

Operating the intestinal panorama image according to at least one of a scrolling instruction, a sliding instruction, a zoom-in instruction, and a zoom-out instruction, to determine a region of interest;

The region of interest is marked according to one of a single-click instruction, a double-click instruction, and a frame selection instruction.
A computer-readable storage medium, characterized in that the computer-readable storage medium stores a program for identifying and locating intestinal images of a capsule endoscope, and when the program for identifying and locating intestinal images of a capsule endoscope is executed by a processor, the method for identifying and locating intestinal images of a capsule endoscope according to any one of claims 1 to 8 is implemented.
A computer device, characterized in that the computer device includes a computer-readable storage medium, a processor, and a capsule endoscope intestinal image recognition and positioning program stored in the computer-readable storage medium, and when the capsule endoscope intestinal image recognition and positioning program is executed by the processor, the capsule endoscope intestinal image recognition and positioning method according to any one of claims 1 to 8 is implemented.