Intelligent path planning system for medical transfer robot

Abstract

The invention relates to the technical field of path planning, in particular to a medical transfer robot path intelligent planning system which comprises an environment image module used for collecting continuous frames of environment images; the region extraction module is used for carrying out region segmentation on the environment image, screening out suspected obstacle regions and extracting a plurality of highlight regions from the suspected obstacle regions; the fuzzy processing module is used for screening out an obstacle region in the suspected obstacle region, and further performing deblurring processing on the obstacle region by using gradient and gray level distribution characteristics corresponding to edge information in the obstacle region to obtain an enhanced environment image; and the path planning module is used for performing path planning on the medical transfer robot by utilizing the enhanced environment image. According to the method, the barrier edge resolution is greatly improved through self-adaptive deblurring, the collision risk of the medical transfer robot is reduced, and the transportation safety of medical equipment in the complex environment of a hospital is further guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to an intelligent path planning system for a medical handling robot. Background Art

[0002] Medical handling robots are of great significance in modern hospitals. They greatly improve the work efficiency of hospitals and reduce the burden on medical staff by automatically handling medicines, medical equipment, bed sheets and other items. At the same time, they reduce contact between people, reduce the risk of cross-infection, and ensure the safety of the medical environment. Robots also effectively avoid labor injuries caused by manual handling, improve the accuracy and safety of item handling, promote the intelligent development of hospitals, and improve the overall quality of medical services and patient experience.

[0003] Normally, medical handling robots use visual sensors to perceive the surrounding environment and provide necessary information for path planning. However, if the visual sensors of medical handling robots move or vibrate violently during shooting, especially in a fast-moving environment, the image may become blurred due to relative motion. Since the degree of influence of motion blur varies in different areas, a unified deblurring method may not achieve the desired effect, further resulting in unsatisfactory path planning results for medical handling robots. Summary of the invention

[0004] The present invention provides a medical handling robot path intelligent planning system to solve the existing problems.

[0005] The medical handling robot path intelligent planning system of the present invention adopts the following technical solutions: An embodiment of the present invention provides a medical handling robot path intelligent planning system, which includes the following modules: An environmental image module, used to collect continuous frames of environmental images through a visual sensor during the movement of the medical handling robot; The region extraction module is used to segment any frame of the environment image and filter out suspected obstacle regions based on the color and area of ​​different regions, and extract several highlight regions from the suspected obstacle regions based on the relative brightness of different regions; A fuzzy processing module is used to utilize the spatial distribution characteristics of the highlight area in the suspected obstacle area and the comprehensive results of the regularity of the edge information in the suspected obstacle area to screen out the obstacle area in the suspected obstacle area, further utilize the gradient and grayscale distribution characteristics corresponding to the edge information in the obstacle area, and perform deblurring processing on the obstacle area in combination with the comprehensive results to obtain an enhanced environment image; The path planning module is used to plan the path of the medical transport robot using enhanced environmental images.

[0006] Preferably, the method for performing region segmentation on any frame of environmental image and screening out suspected obstacle regions based on the colors and areas of different regions includes the following specific steps: Perform clustering processing on the pixel points in each frame of environmental image using the DBSCAN clustering algorithm according to the distances and gray values between pixel points to obtain a number of initial connected regions; For the number of initial connected regions in any environmental image, perform clustering on all connected regions using the K-means clustering algorithm according to the areas of the connected regions to obtain a number of clusters; For any frame of environmental image, calculate the average area of all initial connected regions included in each cluster in the environmental image, and denote all the initial connected regions corresponding to the cluster with the largest average area as irrelevant regions; denote all the connected regions in the environmental image except the irrelevant regions as suspected obstacle regions.

[0007] Preferably, the method for extracting a number of high-light regions from the suspected obstacle regions based on the relative brightness of different regions includes the following specific steps: Convert each frame of environmental image in RGB format into an HSV environmental image, denoted as the HSV environmental image, and use A window of size divides each suspected obstacle region in each frame of HSV environmental image into a number of regions, denoted as equal-divided regions, where is a preset window parameter; According to the difference between the overall V-channel level of each equal-divided region in any HSV environmental image and the overall V-channel level of the environmental image, use it as the possibility that the equal-divided region is a high-light region; Use the magnitude of the possibility that the equal-divided region is a high-light region to screen out a number of high-light regions.

[0008] Preferably, the method for screening out the obstacle regions in the suspected obstacle regions by using the comprehensive result of the spatial distribution characteristics of the high-light regions in the suspected obstacle regions to which they belong and the regularity degree of the edge information in the suspected obstacle regions includes the following specific steps: Use the distribution situation comprehensively reflected by the area proportion characteristics of the high-light regions in the suspected obstacle regions to which they belong and the positional relationship between the high-light regions and other high-light regions in the suspected obstacle regions to which they belong as the distribution discreteness of the high-light regions in the suspected obstacle regions; Obtain the edge information in any suspected obstacle region and fit the straight-line information through the edge information, and use the proportion of the straight-line information in the edge information as the regularity degree of the suspected obstacle region; Comprehensively consider the regularity degree of each suspected obstacle region and the distribution discreteness of the high-light regions in each frame of HSV environmental image to screen out the obstacle regions in the HSV environmental image.

[0009] Preferably, the specific method for obtaining the distribution discreteness of the high-light regions in the suspected obstacle region is as follows: Taking the vertex at the lower left corner of each frame of HSV environmental image as the origin, with the horizontal right direction as the positive direction of the horizontal axis and the vertical upward direction as the positive direction of the vertical axis, a rectangular coordinate system is constructed to obtain the rectangular coordinate system of each frame of HSV environmental image, and the minimum circumscribed circle of each suspected obstacle region in each frame of HSV environmental image is obtained; For any frame of HSV environmental image, any high-light region is denoted as the target high-light region, and the high-light region closest to the target high-light region among the high-light regions other than the target high-light region is obtained and denoted as the high-light reference region of the target high-light region; Using the area ratio of the high-light region on the corresponding minimum circumscribed circle of the suspected obstacle region to which it belongs, and the distance between the high-light region and the corresponding high-light reference region, calculate the distribution discreteness of the high-light regions in the suspected obstacle region, where the distribution discreteness is negatively correlated with the area ratio and positively correlated with the distance.

[0010] Preferably, the specific method for obtaining the regularity degree of the suspected obstacle region is as follows: By using the Canny edge detection algorithm, several edge lines in each suspected obstacle region of each frame of HSV environmental image are obtained, and the Hough line detection algorithm is used to perform line detection on each edge line to obtain several line segments of each suspected obstacle region in each frame of HSV environmental image; Obtain the differences in quantity and length between the edge lines and the corresponding line segments in the suspected obstacle region as the regularity degree of the suspected obstacle region.

[0011] Preferably, the specific method for obtaining the obstacle region is as follows: Combining the regularity degree of the suspected obstacle region and the distribution discreteness of the high-light regions in the suspected obstacle region, obtain the possibility of the suspected obstacle region, where the possibility is positively correlated with both the regularity degree and the distribution discreteness; If the obstacle possibility of the suspected obstacle region is greater than or equal to a preset obstacle region threshold, then the suspected obstacle region is denoted as the obstacle region.

[0012] Preferably, the specific method for using the gradient and gray-scale distribution characteristics corresponding to the edge information in the obstacle region and combining the comprehensive result to perform de-blurring processing on the obstacle region to obtain an enhanced environmental image includes: For any obstacle region in the HSV environmental image, obtain the degree of blurriness of the obstacle region according to the gradient distribution of the edge pixel points in the obstacle region and the average difference between the gray level of each edge pixel point and the overall gray level of all edge pixel points. Take the sum value between the degree of blurriness and the obstacle possibility of each obstacle region in each frame of the HSV environmental image as the sharpening factor, and combine the sharpening factor with the image sharpening algorithm, so as to perform a de-blurring operation on each obstacle region in each frame of the HSV environmental image, and obtain several frames of enhanced environmental images.

[0013] Preferably, the method of combining the sharpening factor with the image sharpening algorithm to perform a de-blurring operation on each obstacle region in each frame of the HSV environmental image includes the following specific method: Use the USM sharpening algorithm to perform sharpening processing on each obstacle region, and use the sharpening factor corresponding to each obstacle region to multiply and adjust the scaling factor in the USM sharpening algorithm during the sharpening process.

[0014] Preferably, the method of using the enhanced environmental image to perform path planning for the medical handling robot includes the following specific method: Use the SLAM technology to analyze consecutive frames of the enhanced environmental image, construct an environmental model of the hospital, and use the Dijkstra algorithm to calculate the travel path of the medical handling robot in the environmental model.

[0015] The beneficial effects of the technical solution of the present invention are as follows: Through multi-modal environmental perception and intelligent optimization algorithms, the navigation accuracy and safety in complex medical scenarios are significantly improved. The system adopts dynamic environmental image acquisition and hierarchical processing technology. First, it screens suspected obstacle regions through color, area, and brightness features, and then based on a dual verification mechanism of spatial distribution and edge regularity, effectively solves the misjudgment problem caused by high-light reflection in traditional visual navigation. In the hospital environment, the system can accurately distinguish real obstacles from optical interference, and greatly improves the edge resolution of obstacles through adaptive de-blurring. The system effectively improves the obstacle avoidance success rate of the medical handling robot during the peak period of the emergency channel, reduces the collision risk of the medical handling robot, and further ensures the transportation safety of medical equipment in the complex environment of the hospital. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is the structural block diagram of the intelligent path planning system for the medical handling robot of the present invention; Figure 2 It is the step flow chart of the area extraction module; Figure 3 It is the step flow chart of the fuzzy processing module. Specific implementation manner

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following combines the drawings and preferred embodiments to detail the specific implementation manner, structure, features and effects of the intelligent path planning system for the medical handling robot proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0020] The following specifically describes the specific solution of the intelligent path planning system for the medical handling robot provided by the present invention with reference to the drawings.

[0021] Please refer to Figure 1 , which shows the step flow chart of the intelligent path planning system for the medical handling robot provided by an embodiment of the present invention. The system includes the following modules: Environmental image module 001: used to collect continuous frames of environmental images through a vision sensor during the movement of the medical handling robot.

[0022] It should be noted that the medical handling robot usually uses machine vision technology to analyze the environment. However, due to the influence of relative motion and the interference of the environment on the movement of the medical handling robot, there is a problem of motion blur in the environmental images collected by the medical handling robot. As a result, when the medical handling robot uses the environmental images for image analysis to achieve path planning, the path planning result has problems. Therefore, the embodiment of the present invention selects to perform deblurring processing on the continuous frame environmental images collected by the medical handling robot to reduce the influence of motion blur on path planning.

[0023] Specifically, in order to implement the intelligent path planning system for the medical handling robot proposed in this embodiment, it is first necessary to collect the environmental images of the medical handling robot during movement. The specific process is as follows: First, use the vision sensor installed on the medical handling robot to collect images of the environment, and obtain continuous frames of images during the movement, which are recorded as environmental images. The preset image acquisition frequency during the image acquisition process is 0.1 s.

[0024] Then, use mean filtering to filter the environmental image of each frame.

[0025] It should be noted that mean filtering is a well-known technology, so the specific filtering process will not be elaborated in the embodiments of the present invention.

[0026] So far, the continuous-frame environmental images of the medical handling robot during movement are obtained through the above method.

[0027] Region extraction module 002: It is used to perform region segmentation on any frame of environmental image and screen out suspected obstacle regions based on the colors and areas of different regions, and extract several high-light regions from the suspected obstacle regions through the relative brightness of different regions.

[0028] It should be noted that the medical handling robot relies on visual sensors to identify the environment, such as obstacles, etc., to assist in path planning. However, when moving quickly, the visual sensors may vibrate, causing the captured images to become blurred, affecting the image quality, and thus affecting the path planning. Since the blurring degrees of different regions are different, the effect of a unified de-blurring method is limited. Therefore, in the embodiments of the present invention, by analyzing the characteristics of the obstacle regions, de-blurring processing is performed specifically to focus on solving the blurring problem of the obstacle regions, as Figure 2 shown in the step flow chart of the region extraction module.

[0029] There may be obstacles on the road surface where the medical handling robot travels, and protruding parts such as fire-fighting equipment and handrails on the wall may also hinder the progress of the robot. Moreover, the colors of the road surface and the wall are usually relatively single. Use the K-means clustering image segmentation algorithm to operate on each environmental image, and mark the region with a larger connected domain area (wall or road surface) obtained as an irrelevant region. And the connected domains other than the irrelevant regions are marked as suspected obstacle regions. However, there may be stickers on the wall or road surface, interfering with the screening of the obstacle regions; therefore, according to the characteristics of the obstacle regions, calculate the obstacle possibility of each suspected obstacle region, screen out several obstacle regions, and then calculate the blurring degree of each obstacle region.

[0030] Specifically, in step S201, the scene image is segmented through clustering segmentation, and suspected obstacle regions are screened out based on the colors and areas of the connected domains after the segmentation processing.

[0031] As an optional embodiment, the method for obtaining the suspected obstacle regions includes: First, perform clustering processing on the pixel points in each frame of environmental image using the DBSCAN clustering algorithm according to the distances and gray values between the pixel points to obtain several initial connected domains.

[0032] Then, for several initial connected components in any environmental image, all the connected components are clustered using the K-means clustering algorithm according to the area of the connected components to obtain several clusters.

[0033] It should be noted that since the K-means clustering algorithm is a well-known technology, the specific clustering process of the K-means clustering algorithm in the embodiments of the present invention will not be elaborated. The clustering number parameter of the preset K-means clustering algorithm is 2, and this is taken as an example for description. The specific preset value is an empirical value and can be adaptively adjusted according to the actual situation.

[0034] Finally, for any frame of environmental image, calculate the average area of all the initial connected components included in each cluster in the environmental image, and mark all the initial connected components corresponding to the cluster with the largest average area as the irrelevant area; mark all the connected components in the environmental image except the irrelevant area as the suspected obstacle area, and obtain several suspected obstacle areas in each frame of environmental image.

[0035] It should be noted that: Since usually the colors of the road surface and the wall are relatively single, and the road surface and the wall area, as a background area, have a much larger corresponding area than the obstacles, so in the embodiments of the present invention, clustering is performed according to the area of each connected component to screen out several irrelevant areas (road surface or wall area), and the areas outside the irrelevant areas may be obstacle areas.

[0036] Step S202, in the environmental image in HSV format, extract the highlight area from the suspected obstacle area according to the horizontal difference in the V-channel value between different ranges.

[0037] It should be noted that through the above process, several suspected obstacle areas of each frame of environmental image are obtained. Stickers may exist on the wall or the road surface, which will interfere with the screening of the obstacle area; so the differences in features between the obstacle area and the sticker are used for screening. Among them, since the obstacle has a three-dimensional structure, multiple highlight areas will be generated on its surface under the action of light. These highlight areas usually appear in different parts of the obstacle, and due to factors such as the light angle, the surface material and shape of the object, their distribution is often scattered and irregular. Therefore, within the obstacle area, there may be multiple scattered and discontinuous highlight areas; according to the above characteristics, several highlight areas of each suspected obstacle area are screened out.

[0038] As a preferred embodiment, the specific method for obtaining the highlight area includes: First, convert each frame of environmental image in RGB format into an HSV environmental image, denoted as the HSV environmental image, and use A window of a certain size divides each suspected obstacle area in each frame of the HSV environmental image into several areas, denoted as equal division areas, where is a preset window parameter.

[0039] Among them, the technology of mutual transformation between RGB and HSV is a well-known technology, so it will not be specifically described in the embodiments of the present invention; in addition, in the embodiments of the present invention, the window parameter is preset to be 10 according to experience, and can be adjusted according to the actual situation, and the embodiments of the present invention do not specifically limit it.

[0040] Then, according to the difference between the overall V-channel level of each equal division area in any HSV environmental image and the overall V-channel level of the environmental image, as the possibility that the equal division area is a highlight area, the highlight areas are screened out.

[0041] As an optional embodiment, taking the th frame of the HSV environmental image as an example, obtain the th frame of the HSV environmental image, the th suspected obstacle area, the th equal division area is the possibility of a highlight area. The specific calculation formula is: Among them, represents the th frame of the HSV environmental image, the th suspected obstacle area, the th equal division area is the possibility of a highlight area; represents the linear normalization function; represents the th frame of the HSV environmental image, the th suspected obstacle area, the th equal division area contains the number of pixel points; represents the th frame of the HSV environmental image, the th suspected obstacle area, the th equal division area, the th pixel point in the V channel value; represents the th frame of the HSV environmental image, the average value of all pixel points in the V channel value.

[0042] It should be noted that: in the HSV image, the value corresponding to the pixel point in the V channel reflects the brightness characteristics of the pixel point. Therefore, in the embodiments of the present invention, It reflects the brightness level of the equally divided area relative to the ambient image. Therefore, the greater the difference between the V channel value of the pixel points in the equally divided area and the average value of the V channel value of all pixels in the entire ambient image, the greater the brightness of the equally divided area compared to the overall brightness of the ambient image, and the greater the possibility that the equally divided area is a highlight area.

[0043] By performing calculations on each equally divided area of ​​each suspected obstacle area of ​​each frame of the HSV environment image according to the above process, the possibility that each equally divided area of ​​each suspected obstacle area of ​​each frame of the HSV environment image is a highlight area is obtained.

[0044] As an optional embodiment, a plurality of highlight regions are selected based on the probability that the equally divided regions are highlight regions. The specific process is as follows: a highlight probability threshold is preset. , if the probability that the equally divided area is a highlight area is greater than or equal to , then the equally divided area is recorded as the highlight area.

[0045] So far, several highlight areas in any frame environment image are obtained through the above method.

[0046] Blur processing module 003: It is used to utilize the comprehensive results of the spatial distribution characteristics of the highlight area in the suspected obstacle area and the regularity of the edge information in the suspected obstacle area to screen out the obstacle area in the suspected obstacle area, further utilize the gradient and grayscale distribution characteristics corresponding to the edge information in the obstacle area, and deblur the obstacle area in combination with the comprehensive results to obtain an enhanced environment image.

[0047] It should be noted that, in order to accurately extract the features of obstacles that usually appear in the complex environment of the hospital, the embodiment of the present invention takes into account that the spatial distribution of different areas of obstacles in the environment image and the gradient and grayscale information on the corresponding edges are different from those of other non-obstacles. Therefore, in order to ensure that the corresponding environment model can be accurately established according to the position of the obstacle in the image during subsequent path planning and improve the success rate of obstacle avoidance, the embodiment of the present invention deblurs the obstacle area in the environment image through the analysis results of the above features, such as Figure 3 Shown is a flow chart of the steps of the fuzzy processing module.

[0048] Specifically, step S301 uses the distribution situation comprehensively reflected by the area ratio characteristics of the highlight area in the suspected obstacle area to which it belongs and the positional relationship between the highlight area and other highlight areas in the suspected obstacle area to which it belongs as the distribution discreteness of the highlight area in the suspected obstacle area.

[0049] It should be noted that: Since obstacles usually have complex three-dimensional structures, with different inclination angles and shapes on their surfaces, when light shines on their surfaces, reflection and refraction occur, and the distribution of multiple specular highlight regions formed is usually uneven and scattered. Therefore, according to the above characteristics, in the embodiments of the present invention, the distribution discreteness of the specular highlight regions in each suspected obstacle region in each frame of HSV environmental image is calculated.

[0050] As a preferred embodiment, the process of obtaining the distribution discreteness of the specular highlight regions in each suspected obstacle region includes: First, taking the vertex at the lower left corner of each frame of HSV environmental image as the origin, with the horizontal right direction as the positive direction of the horizontal axis and the vertical upward direction as the positive direction of the vertical axis, a rectangular coordinate system is constructed to obtain the rectangular coordinate system of each frame of HSV environmental image, and the minimum circumscribed circle of each suspected obstacle region in each frame of HSV environmental image is obtained.

[0051] Then, for any frame of HSV environmental image, any specular highlight region is denoted as the target specular highlight region, and the specular highlight region that is the closest to the target specular highlight region among the specular highlight regions other than the target specular highlight region is obtained and denoted as the specular highlight reference region of the target specular highlight region.

[0052] Finally, using the area ratio of the specular highlight region on the minimum circumscribed circle corresponding to the suspected obstacle region to which it belongs, and the distance between the specular highlight region and the corresponding specular highlight reference region, the distribution discreteness of the specular highlight regions in the suspected obstacle region is calculated, where the distribution discreteness is negatively correlated with the area ratio and positively correlated with the distance.

[0053] As an alternative embodiment, for the th suspected obstacle region in the th frame of HSV environmental image, the distribution discreteness of the specular highlight regions in the th suspected obstacle region in the th frame of HSV environmental image is obtained. The specific calculation formula is: where represents the distribution discreteness of the specular highlight regions in the th suspected obstacle region in the th frame of HSV environmental image; represents the linear normalization function; represents the area of the minimum circumscribed circle of the th suspected obstacle region in the th frame of HSV environmental image; represents the number of specular highlight regions in the th suspected obstacle region in the th frame of HSV environmental image; Indicates the area of the th highlight area in the th suspected obstacle area in the th frame of the HSV environment image; Indicates the distance between the th highlight area and the corresponding highlight reference area in the

[0054] It should be noted that: is the proportion of the area of the highlight area on the minimum circumscribed circle corresponding to the suspected obstacle area to which it belongs. Since and are reciprocal to each other, in the embodiments of the present invention, the distribution discreteness is negatively correlated with the area proportion. The smaller the area of the highlight area in the suspected obstacle area, and the larger the area of the minimum circumscribed circle of the suspected obstacle area, it indicates that the distribution of the highlight areas in the suspected obstacle area is more discrete, and the distribution discreteness of the highlight areas in the suspected obstacle area is greater. The greater the distance between the center of the highlight area in the suspected obstacle area and its corresponding highlight reference area, the greater the distribution discreteness of the highlight areas in the suspected obstacle area.

[0055] So far, by performing operations on each suspected obstacle area of each frame of the HSV environment image according to the above process, the distribution discreteness of the highlight areas of each suspected obstacle area of each frame of the HSV environment image is obtained.

[0056] Step S302, obtain the edge information in any suspected obstacle area and fit the straight line information through the edge information, and use the proportion of the straight line information in the edge information as the regularity degree of the suspected obstacle area.

[0057] It should be noted that: In order to further improve the accuracy of screening the obstacle area, other features are used for further screening. Among them, stickers or signs usually contain various characters, and the shapes of these characters often show a certain degree of irregularity, which may be curves, sharp corners, or complex structures intertwined together. In contrast, the structures of obstacles are usually more regular, and they often present clear geometric shapes, such as straight lines, or rectangles, such as medical cabinets and trolleys. According to the above characteristics, the regularity degree of each suspected obstacle area is calculated.

[0058] First, obtain several edge lines in each suspected obstacle area of each frame of the HSV environment image through the Canny edge detection algorithm, and use the Hough line detection algorithm to perform straight line detection on each edge line to obtain several straight line segments in each suspected obstacle area of each frame of the HSV environment image.

[0059] It should be noted that the Hough line detection algorithm is a well-known algorithm, so the specific process of the algorithm will not be described in detail in the embodiments of the present invention.

[0060] Then, obtain the differences in quantity and length between the edge lines and the corresponding line segments in the suspected obstacle area as the regularity degree of the suspected obstacle area.

[0061] As an optional embodiment, for the th suspected obstacle area in the th frame of the HSV environment image, obtain the regularity degree of the th suspected obstacle area in the th frame of the HSV environment image. The specific calculation formula is as follows: Where represents the regularity degree of the th suspected obstacle area in the th frame of the HSV environment image; represents the number of edge lines of the th suspected obstacle area in the th frame of the HSV environment image; represents the length of the th edge line of the th suspected obstacle area in the th frame of the HSV environment image; represents the number of line segments of the th suspected obstacle area in the th frame of the HSV environment image; represents the length of the th line segment of the th suspected obstacle area in the th frame of the HSV environment image.

[0062] It should be noted that: the longer the lengths of all the line segments in the suspected obstacle area, the more regular the suspected obstacle area is, and the greater the regularity degree of the suspected obstacle area.

[0063] So far, by performing operations on each suspected obstacle area in each frame of the HSV environment image according to the above process, the regularity degree of each suspected obstacle area in each frame of the HSV environment image is obtained.

[0064] Step S303: Screen out the obstacle areas in the HSV environment image by comprehensively considering the regularity degree of each suspected obstacle area in each frame of the HSV environment image and the distribution discreteness of the highlight areas.

[0065] First, the possibility of the suspected obstacle area is obtained by combining the regularity degree of the suspected obstacle area and the distribution discreteness of the highlight areas in the suspected obstacle area, where the possibility is positively correlated with both the regularity degree and the distribution discreteness.

[0066] As an optional embodiment, for the th suspected obstacle area in the th frame of the HSV environment image, the obstacle possibility of the th suspected obstacle area in the th frame of the HSV environment image is obtained. The specific calculation formula is: Where, represents the obstacle possibility of the th suspected obstacle area in the th frame of the HSV environment image; represents the distribution discreteness of the highlight areas of the th suspected obstacle area in the th frame of the HSV environment image; represents the regularity degree of the th suspected obstacle area in the th frame of the HSV environment image; represents the linear normalization function.

[0067] It should be noted that: the greater the distribution discreteness of the highlight areas of the suspected obstacle area, and the greater the regularity degree of this suspected obstacle area, the greater the obstacle possibility of this suspected obstacle area.

[0068] Then, if the obstacle possibility of the suspected obstacle area is greater than or equal to the preset obstacle area threshold, the suspected obstacle area is recorded as an obstacle area.

[0069] Thus, several obstacle areas in any HSV environment image are obtained.

[0070] Step S304: Using the gradient and gray-scale distribution characteristics corresponding to the edge information in the obstacle area, and combining with the obstacle possibility corresponding to the obstacle area as a sharpening factor, so as to perform de-blurring processing on the obstacle area to obtain an enhanced environment image.

[0071] It should be noted that through the above process, in the embodiments of the present invention, several obstacle regions of each frame of HSV environmental image are obtained. Since vibration will cause different degrees of blurring of different obstacle regions, it is necessary to calculate the blurring degrees of different obstacles. Image blurring will cause the standard deviation of the gradient amplitude of edge pixel points to decrease and the difference between the gray values of all pixel points to also decrease. According to the above characteristics, the blurring degree of each obstacle region of each frame of HSV environmental image is calculated.

[0072] As a preferred embodiment, the specific method for obtaining the enhanced environmental image includes: First, for the obstacle region in any HSV environmental image, according to the gradient distribution of the edge pixel points in the obstacle region and the average difference between the gray value of each edge pixel point and the overall gray level of all edge pixel points, the blurring degree of the obstacle region is obtained.

[0073] Then, the sum value between the blurring degree and the obstacle possibility of each obstacle region of each frame of HSV environmental image is used as the sharpening factor, and the sharpening factor is combined with the image sharpening algorithm, so as to perform a de-blurring operation on each obstacle region of each frame of HSV environmental image, and several frames of enhanced environmental images are obtained.

[0074] As an alternative embodiment, for the th obstacle region of the th frame of HSV environmental image, the blurring degree of the th obstacle region of the th frame of HSV environmental image is obtained. The specific calculation formula is: Among them, represents the blurring degree of the th obstacle region of the th frame of HSV environmental image; represents the exponential function with the natural constant as the base; represents the th frame of HSV environmental image of the th obstacle region of all edge pixel point gradient amplitude variances; represents the th frame of HSV environmental image of the th obstacle region of the number of pixel points; represents the absolute value function; represents the th frame of HSV environmental image of the th obstacle region of the th edge pixel point gray value; represents the th frame of HSV environmental image of the The average gray value of all edge pixel points of an obstacle area.

[0075] It should be noted that: the smaller the variance of the gradient magnitudes of all edge pixel points of an obstacle area, the greater the degree of blurriness of the obstacle area; the smaller the difference between the gray values of all edge pixel points of an obstacle area, the greater the degree of blurriness of the obstacle area.

[0076] As an optional embodiment, the specific method for performing de-blurring operations on each obstacle area of each frame of HSV environmental image is: using the USM sharpening algorithm to perform sharpening processing on each obstacle area, and during the sharpening process, using the sharpening factor corresponding to each obstacle area to multiply and adjust the scaling factor in the USM sharpening algorithm.

[0077] It should be noted that the USM (Unsharp Masking) sharpening algorithm is a well-known technology. Therefore, the specific process of using the USM sharpening algorithm to perform sharpening processing on the obstacle area will not be elaborated in the embodiments of the present invention. Through the above process, the embodiments of the present invention obtain the degree of blurriness of each obstacle area of each frame of HSV environmental image and the possibility of being an obstacle area. Different degrees of de-blurring operations are performed on each obstacle area according to these two characteristic values, realizing the adaptive enhancement of different regions in the environmental image.

[0078] So far, each frame of de-blurred HSV environmental image, that is, the enhanced environmental image, is obtained through the above method.

[0079] Path planning module 004: used to perform path planning for the medical handling robot using the enhanced environmental image.

[0080] Specifically, use the SLAM technology to analyze consecutive frames of the enhanced environmental image, construct an environmental model of the hospital, and use the Dijkstra algorithm to calculate the travel path of the medical handling robot in the environmental model.

[0081] It should be noted that: SLAM (Simultaneous Localization and Mapping) is simultaneous localization and mapping. It is a technology that enables a robot to perform self-localization and construct an environmental map in real time through sensor data without a prior map in an unknown environment. It combines the two tasks of localization and mapping, usually using a camera or other sensors to perceive the surrounding environment. SLAM is widely used in fields such as autonomous driving, service robots, and drones to help robots navigate autonomously, avoid obstacles, and make decisions. The Dijkstra algorithm is a classic algorithm for calculating the shortest path between two nodes in a graph, applicable to graphs with non-negative edge weights, and thus realizes the calculation of the shortest path by gradually expanding the shortest path tree.

[0082] So far, this embodiment is completed.

[0083] It should be noted that the model used in this embodiment is only used to represent the negative correlation relationship and restrict the result of the model output to be within the interval. In specific implementation, it can be replaced with other models with the same purpose. This embodiment only takes the model as an example for description and does not specifically limit it, where refers to the input of the model.

[0084] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Intelligent path planning system for medical handling robots, characterized in that: The system includes the following modules: An environmental image module is used to collect continuous frames of environmental images through a visual sensor during the movement of the medical handling robot; The region extraction module is used to segment any frame of the environment image and filter out suspected obstacle regions based on the color and area of ​​different regions, and extract several highlight regions from the suspected obstacle regions based on the relative brightness of different regions; A fuzzy processing module is used to utilize the spatial distribution characteristics of the highlight area in the suspected obstacle area and the comprehensive results of the regularity of the edge information in the suspected obstacle area to screen out the obstacle area in the suspected obstacle area, further utilize the gradient and grayscale distribution characteristics corresponding to the edge information in the obstacle area, and perform deblurring processing on the obstacle area in combination with the comprehensive results to obtain an enhanced environment image; The method for obtaining the obstacle area specifically includes: The distribution situation comprehensively reflected by the area ratio characteristics of the highlight region in the suspected obstacle region and the positional relationship between the highlight region and other highlight regions in the suspected obstacle region is used as the distribution discreteness of the highlight region in the suspected obstacle region; Obtain edge information in any suspected obstacle area and fit straight line information through the edge information, and use the proportion of the straight line information in the edge information as the regularity of the suspected obstacle area; The obstacle areas in the HSV environment image are screened out by comprehensively considering the regularity of each suspected obstacle area and the distribution discreteness of the highlight area in each frame of the HSV environment image; The path planning module is used to plan the path of the medical transport robot using enhanced environmental images.

2. According to claim 1, the medical handling robot path intelligent planning system is characterized in that: The specific method of performing region segmentation on any frame environment image and screening out suspected obstacle regions based on the colors and areas of different regions includes: The pixels in each frame of the environment image are clustered using the DBSCAN clustering algorithm according to the distance and gray value between the pixels to obtain several initial connected domains; For several initial connected domains in any environment image, all connected domains are clustered using the K-means clustering algorithm according to the area of ​​the connected domains to obtain several clusters; For any frame of the environmental image, the average area of ​​all initial connected domains contained in each cluster in the environmental image is calculated, and all initial connected domains corresponding to the cluster with the largest average area are recorded as irrelevant areas; all connected domains in the environmental image except the irrelevant areas are recorded as suspected obstacle areas.

3. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The specific method of extracting several highlight areas from the suspected obstacle area by using the relative brightness of different areas is as follows: Convert each frame of the RGB environment image to an HSV environment image and record it as an HSV environment image. The window of size divides each suspected obstacle area in each frame of HSV environment image into several areas, which are recorded as equal areas. is the preset window parameter; According to the difference between the overall V channel level of each equally divided area in any HSV environment image and the overall V channel level of the environment image, the possibility that the equally divided area is a highlight area is determined; Several highlight areas are screened out using the likelihood that the equally divided areas are highlight areas.

4. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The specific method for obtaining the distribution discreteness of the highlight area in the suspected obstacle area is: A rectangular coordinate system is constructed with the vertex of the lower left corner of each frame of the HSV environment image as the origin, the horizontal right direction as the positive direction of the horizontal axis, and the vertical upward direction as the positive direction of the vertical axis to obtain the rectangular coordinate system of each frame of the HSV environment image, and obtain the minimum circumscribed circle of each suspected obstacle area in each frame of the HSV environment image; For any frame of HSV environment image, any highlight area is recorded as a target highlight area, and a highlight area closest to the target highlight area among highlight areas other than the target highlight area is obtained and recorded as a highlight reference area of ​​the target highlight area; The distribution discreteness of the highlight area in the suspected obstacle area is calculated by using the area ratio of the highlight area on the corresponding minimum circumscribed circle of the suspected obstacle area and the distance between the highlight area and the corresponding highlight reference area, wherein the distribution discreteness is negatively correlated with the area ratio, and the distribution discreteness is positively correlated with the distance.

5. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The specific method for obtaining the regularity of the suspected obstacle area is: The Canny edge detection algorithm is used to obtain several edge lines in each suspected obstacle area of ​​each frame of the HSV environment image, and the Hough line detection algorithm is used to perform line detection on each edge line to obtain several straight line segments in each suspected obstacle area of ​​each frame of the HSV environment image; The difference in quantity and length between the edge lines and the corresponding straight line segments in the suspected obstacle area is obtained as the regularity of the suspected obstacle area.

6. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The specific method for obtaining the obstacle area is: Combining the regularity of the suspected obstacle area and the distribution discreteness of the highlight area in the suspected obstacle area, the possibility of the suspected obstacle area is obtained, wherein the possibility is positively correlated with both the regularity and the distribution discreteness; If the obstacle possibility of the suspected obstacle area is greater than or equal to a preset obstacle area threshold, the suspected obstacle area is recorded as an obstacle area.

7. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The method of utilizing the gradient and grayscale distribution characteristics corresponding to the edge information in the obstacle area and combining the comprehensive result to perform deblurring processing on the obstacle area to obtain an enhanced environment image includes the following specific methods: For an obstacle area in any HSV environment image, the blur degree of the obstacle area is obtained according to the gradient distribution of edge pixels in the obstacle area and the average difference between the grayscale of each edge pixel and the overall grayscale level of all edge pixels; The sum of the blur degree and the obstacle possibility of each obstacle area in each frame of the HSV environment image is used as the sharpening factor, and the sharpening factor is combined with the image sharpening algorithm to perform a deblurring operation on each obstacle area in each frame of the HSV environment image to obtain several frames of enhanced environment images.

8. The medical handling robot path intelligent planning system according to claim 7, characterized in that: The sharpening factor is combined with the image sharpening algorithm to perform a deblurring operation on each obstacle area of ​​each frame of the HSV environment image, including the following specific methods: The USM sharpening algorithm is used to sharpen each obstacle area. During the sharpening process, the sharpening factor corresponding to each obstacle area is used to multiply the scaling factor in the USM sharpening algorithm.

9. The medical handling robot path intelligent planning system according to claim 1, characterized in that: The specific method of using the enhanced environment image to plan the path of the medical handling robot includes: The SLAM technology is used to analyze the enhanced environmental images of continuous frames, to construct an environmental model of the hospital, and the Dijkstra algorithm is used to calculate the travel path of the medical transport robot in the environmental model.

Images