Intelligent tracking type ray inspection area self-adaptive shielding control system and method

Abstract

The invention belongs to the technical field of automatic control, and discloses an intelligent tracking type ray inspection area self-adaptive shielding control system and method. Comprising the steps of collecting a human body three-dimensional image in real time; recognizing and extracting a human body contour region from the human body three-dimensional image; carrying out fine segmentation on the human body contour region to obtain a tissue regions; performing intelligent analysis on each tissue area, and determining an occlusion area in the tissue area; obtaining area space information of each shielding area, and coordinating and controlling each protection component to shield the corresponding shielding area; detecting the movement track of the human body in the examination process in real time, and generating corresponding movement change data for each shielding area; according to the movement change data, the protection assembly corresponding to each shielding area is dynamically controlled; the radiation influence of rays on healthy tissues can be effectively reduced, and the protection safety and the examination efficiency in the ray examination process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control, and more specifically, to an intelligent tracking type ray inspection area adaptive shielding control system and method. Background Art

[0002] During medical imaging examinations and radiotherapy processes (such as X-ray examinations and radioactive treatments), the human body often needs to be exposed to a certain dose of ionizing radiation in order to obtain clear images or perform precise treatment on specific lesion areas. However, while the ray irradiates the target area, it often has an unnecessary radiation impact on surrounding healthy tissues. Although traditional inspection equipment is equipped with fixed protective screens or lead plates for shielding, due to the possible movement of the human body during the inspection process and the inability of the protective screen to follow and adjust in real time, some healthy tissues will still be exposed to the ray, posing a radiation safety hazard. That is, traditional ray inspection equipment mainly has deficiencies such as inflexible protection, high radiation risk, and low inspection efficiency. Therefore, there is an urgent need for a ray protection system based on intelligent tracking and adaptive shielding.

[0003] The patent with publication number CN111308993B discloses a human target following method based on monocular vision, including: starting the following system of the monocular vision robot through voice control and turning on the following function of the robot; detecting the target directly in front, using a human body recognition algorithm to detect multiple human body targets directly in front of the robot; selecting the target to be followed, and selecting the most suitable human body target from the obtained multiple human body detection frames as the target to be followed; setting the tracker state, using a combination of a main tracker and an auxiliary tracker to obtain the tracking information of the target to be followed; motion following control, setting corresponding following strategies according to the relationship between the target information and the following information; updating the following target information, taking the tracking information at the current moment as the target information at the next moment to achieve continuous tracking. This invention adopts an algorithm that can perform human target following with a monocular camera, and has the characteristics of strong real-time performance, high detection rate, and fast response.

[0004] However, although the above technology can achieve human target following, it only relies on monocular vision. Monocular vision can only capture two-dimensional image information and lacks depth perception ability. It is difficult to accurately judge the distance and spatial position between the human body and the protection component when the human body moves, resulting in inaccurate shielding by the protection module or delayed adjustment, thus affecting the protection effect. And monocular vision is relatively sensitive to external environmental changes such as light, angle, and obstacles, and is prone to tracking failure or loss of the target, resulting in the inability of the protection component to adjust the shielding area in real time, affecting the safety and effectiveness of the inspection.

[0005] In view of this, the present invention proposes an intelligent tracking type ray inspection area adaptive shielding control system and method to solve the above problems. Summary of the Invention

[0006] To overcome the above-mentioned defects of the prior art and achieve the above object, the present invention provides the following technical solution: An intelligent tracking type ray inspection area adaptive shielding control method, comprising: Real-time collect three-dimensional human body images; Perform grayscale analysis on the three-dimensional human body images, and identify and extract the human body contour area from the three-dimensional human body images; Based on an adaptive segmentation algorithm, perform refined segmentation on the human body contour area to obtain n tissue areas, where n is an integer greater than 1; Perform intelligent analysis on each tissue area respectively to determine the occlusion areas that do not require ray inspection within the tissue area; Obtain the regional spatial information of each occlusion area, and based on the regional spatial information, coordinately control each protection component to occlude the corresponding occlusion area; Adopt a fusion tracking technology to real-time detect the movement trajectory of the human body during the inspection process, and generate corresponding movement change data for each occlusion area; According to the movement change data, dynamically control each protection component corresponding to each occlusion area.

[0007] Further, the three-dimensional human body image is a three-dimensional image data including the surface shape, spatial position and depth information of the human body; The method for identifying and extracting the human body contour area from the three-dimensional human body image includes: Obtain the depth value of each point cloud point in the three-dimensional human body image, preset a depth range value, and compare the depth value of each point cloud point with the depth range value respectively; if the depth value is within the depth range value, mark the corresponding point cloud point as a candidate point; perform grayscale processing on each candidate point to obtain the grayscale value of each candidate point; preset a grayscale threshold, and compare the grayscale value of each candidate point with the grayscale threshold respectively; if the grayscale value is greater than or equal to the grayscale threshold, mark the corresponding candidate point as a contour point; if the grayscale value is less than the grayscale threshold, do not mark the corresponding candidate point; perform morphological processing on all contour points to obtain regional points, and use the region composed of all regional points as the human body contour area.

[0008] Further, the step of obtaining n tissue areas includes: Step S101: Screen out n regional points from all regional points as tissue points, and use the regional points that are not used as tissue points as segmentation points; Step S102: According to n tissue points, construct the corresponding a set of tissues, and calculate the Euclidean distance from each segmentation point to each tissue point in turn, and mark it as the point spacing; Step S103: Compare the point spacings corresponding to the same segmentation point, mark the tissue point corresponding to the smallest point spacing as the nearest point, and divide each segmentation point into the tissue set corresponding to the nearest point in turn; Step S104: Calculate the new tissue point corresponding to each tissue set, and replace each tissue point with the corresponding new tissue point; Step S105: Loop steps S102 to S104 until each new tissue point calculated in step S104 is the same as the corresponding new tissue point calculated in the previous loop process, then the loop ends, and obtain a set of tissues and the regional points in each tissue set; Step S106: Segment the human body contour area according to the area composed of the regional points in each tissue set, and obtain a set of tissue areas; In the said step S101, the steps of screening out regional points as tissue points include: Step S201: Randomly select a regional point as a tissue point; Step S202: Calculate the point spacing from each segmentation point to each tissue point, compare the point spacings corresponding to the same segmentation point, and take the point spacing with the smallest value as the minimum distance of the corresponding segmentation point; Step S203: Square the minimum distance of each segmentation point in turn to obtain the squared distance; add up the squared distances of each segmentation point in turn to obtain the sum of squared distances; divide the squared distance of each segmentation point by the sum of squared distances to obtain the screening probability corresponding to each segmentation point; Step S204: Screen out a regional point as a tissue point according to the screening probability; Step S205: Loop steps S202 to S204 until regional points are obtained, and the loop ends.

[0009] Furthermore, in the said step S104, the method for calculating the new tissue point includes: Count the number of regional points in each tissue area and mark it as the number of regional points; add up each regional point in each tissue area in turn, and then divide by the corresponding number of regional points to obtain the average point corresponding to each tissue area; calculate the point spacing from each regional point in each tissue area to the corresponding average point and mark it as the average distance; compare the average distances corresponding to the same tissue area, and take the regional point corresponding to the smallest average distance as the new tissue point corresponding to the corresponding tissue area; In the step S204, the method for screening out regional points according to the screening probability includes: Sort each regional point in descending order according to the corresponding screening probability, and sequentially add the screening probability of each regional point to the screening probability of the corresponding pre-point according to the positive order to obtain the cumulative probability of each regional point; calculate the probability range of each regional point according to the cumulative probability of each regional point; from the interval Generate a random number, mark the probability range into which the random number falls as the screening range, and screen out the regional points corresponding to the screening range; where, mark a regional point as the current point, then the pre-point of the current point is all the regional points ranked in front of the current point; the maximum value of the probability range corresponding to the current point is the cumulative probability corresponding to the current point, and the minimum value of the probability range corresponding to the current point is the cumulative probability corresponding to the regional point ranked one position in front of the current point.

[0010] Furthermore, the method for determining the occlusion area includes: Obtain the inspection area, where the inspection area is the tissue area that needs to be irradiated during the current X-ray inspection of the human body, and mark all tissue areas that are not the inspection area as non-inspection areas; obtain the protection feature data, where the protection feature data is the feature data of the protection component, and the protection feature data includes the component shape, component size, and component quantity; where, the component shape includes the shape of each protection component, the component size includes the size of each protection component, and the component quantity is the quantity of all protection components; Calculate the area of each non-inspection area and mark it as the area of the region; use the protection feature data and the area of the region as analysis data, and input the analysis data into the trained component division model to predict the corresponding component division matrix; where, the component division model is a deep neural network model, and the component division matrix is matrix, is the number of non-inspection areas, is the number of components; the elements in the component division matrix correspond one by one to the component labels, and the component labels are the digital labels of the protection components, and the digital labels corresponding to different protection components are all different, and the range of the component labels is , 0 indicates that there is no protection component, and the elements in each column of the component division matrix correspond one by one to the non-inspection areas; According to the component division matrix, obtain the protection component corresponding to each non-inspection area, and use the geometric segmentation method to perform regional segmentation on each non-inspection area to obtain the corresponding occlusion areas for each non-inspection area, is the number of protection components corresponding to a non-inspection area; The method for calculating the area of each non-inspection area is as follows: Obtain the area points of each non-inspection area and mark them as calculation points; use the Delaunay triangulation method to divide the calculation points corresponding to each non-inspection area into multiple triangles respectively; use the cross-product calculation formula to calculate the area of each triangle, and successively add the areas of the triangles corresponding to each non-inspection area to obtain the area of each non-inspection area.

[0011] Further, the method for obtaining the area space information is as follows: Obtain the area points of each occlusion area and mark them as occlusion points, use the boundary detection algorithm to analyze the occlusion points of each occlusion area respectively, and extract the corresponding boundary points from the occlusion points of each occlusion area. The boundary points are the occlusion points located on the outer contour or edge of the occlusion area; obtain the three-dimensional coordinates of each boundary point, and generate the area space information corresponding to each occlusion area according to the three-dimensional coordinates of the boundary points corresponding to each occlusion area; The method for coordinating and controlling the protection components includes: Randomly construct sets of different component position sets, where is an integer greater than 1, and each set of component position sets includes component positions, and each component position includes coordinate values, coordinate values, coordinate values, the rotation angle around the axis, the rotation angle around the axis, and the rotation angle around the axis. The component positions correspond to the protection components one by one; sequentially incrementally set digital labels for each set of component position sets and mark them as set labels, and the range of the set labels is ; Divide plus 1 by 2 as the initial control center, and divide minus 1 by 2 as the initial control radius; Define the iterative process. The iterative process is: Generate candidate solutions within the range of the set label. The candidate solutions correspond to the set label one by one, ; Sequentially calculate the effective occlusion coefficients corresponding to each candidate solution, move the control center to the candidate solution with the largest effective occlusion coefficient, and adjust the control radius; Execute the iterative process. When the number of iterations is greater than or equal to the preset iteration threshold, the iterative process ends. Mark the candidate solution corresponding to the control center as the optimal solution, obtain the set of component positions corresponding to the set label of the optimal solution and mark it as the optimal set, and coordinate and control each protection component according to the optimal set.

[0012] Further, randomly construct ​​The method for a set of different component position sets is as follows: Obtain a coordinate range and an angle range. The coordinate range includes the coordinate range, the coordinate range, and the coordinate range. The angle range is ; Randomly select a value from each range within the coordinate range and randomly select three values from the angle range to construct a set of component positions, and a total of sets of different component position sets are constructed; The method for generating candidate solutions is as follows: Generate a random coefficient within the interval ; Multiply the control radius by the random coefficient and add the control center to obtain a candidate solution. The method for adjusting the control radius is as follows: Randomly generate an adjustment coefficient within the interval ; Multiply the control radius by the adjustment coefficient to obtain an adjusted radius, and adjust the control radius according to the adjusted radius; The method for calculating the effective occlusion coefficient is as follows: Obtain the set label corresponding to the candidate solution and mark it as the current label. Obtain the set of component positions corresponding to the current label and mark it as the current position set; Use the regional spatial information of each occlusion area and the current position set as calculation data, and input the calculation data into the trained occlusion prediction model to predict the corresponding occlusion degree. The occlusion degree is the occlusion effect of each protective component on the corresponding occlusion area, and the range of the occlusion degree is ; Input the calculation data into the trained collision prediction model to predict the corresponding collision degree. The collision degree is the collision degree between each protective component, and the range of the collision degree is ; Subtract the collision degree from the occlusion degree to obtain the effective occlusion coefficient. The training processes of both the occlusion prediction model and the collision prediction model are the same as that of the component division model, and both are deep neural network models.

[0013] Furthermore, the method for generating movement change data includes: Obtain the boundary points corresponding to each occlusion area and use them as the reference frame point cloud data for the corresponding occlusion area; Create a corresponding filter for each occlusion area and initialize the filter parameters. The filter parameters include a state vector, a state transition matrix, an observation matrix, and an error covariance matrix; Among them, the state vector includes the regional position and the regional attitude. The calculation method for the regional position is as follows: Count the number of boundary points corresponding to each occlusion area and mark it as the number of boundary points. Add the three-dimensional coordinates of the boundary points corresponding to each occlusion area in sequence, and then divide by the number of boundary points to obtain the regional position of each occlusion area; The calculation method for the regional attitude is as follows: Use the least squares method to generate the normal vector of each occlusion area, and then use the Rodrigues rotation formula to convert the normal vector of each occlusion area into the corresponding rotation matrix, and use the rotation matrix of each occlusion area as the corresponding regional attitude; During the human body movement process, boundary points corresponding to each occluded area are obtained in real time and used as the point cloud data of the current frame of the corresponding occluded area; the iterative closest point algorithm is used to register the point cloud data of the current frame with the point cloud data of the reference frame, and the change data of each occluded area is calculated, where the change data includes displacement change and angle change; the change data of each occluded area is input into the corresponding filter for smoothing processing to generate corresponding smoothed data, and the smoothed data includes smoothed displacement change and smoothed angle change; the smoothed data of each occluded area is used as the corresponding movement change data.

[0014] Further, the method for dynamically controlling the protection components includes: In the obtained optimal set, the component position corresponding to each protection component is obtained and marked as the real-time position; the set of real-time positions and the movement change data are used as movement prediction data, and the movement prediction data is input into the trained component movement model to predict the corresponding movement position, where the movement position is the component position after the protection component moves; according to the movement position, the protection components corresponding to each occluded area are dynamically controlled; the training process of the component movement model is the same as that of the component division model, and both are deep neural network models.

[0015] An intelligent tracking type ray inspection area adaptive shielding control system, implementing the described intelligent tracking type ray inspection area adaptive shielding control method, includes: An image acquisition module, used for real-time acquisition of a three-dimensional human body image; A human body recognition module, used for performing grayscale analysis on the three-dimensional human body image, and recognizing and extracting the human body contour area from the three-dimensional human body image; An image segmentation module, used for performing refined segmentation on the human body contour area based on an adaptive segmentation algorithm to obtain tissue areas, where t is an integer greater than 1; An area determination module, used for respectively performing intelligent analysis on each tissue area to determine the occluded areas within the tissue area that do not require ray inspection; A protection control module, used for obtaining the regional space information of each occluded area, and based on the regional space information, coordinately controlling each protection component to shield the corresponding occluded area; A movement tracking module, used for adopting a fusion tracking technology to real-time detect the movement trajectory of the human body during the inspection process, and generating corresponding movement change data for each occluded area; A dynamic control module, used for dynamically controlling each protection component corresponding to each occluded area according to the movement change data.

[0016] Technical effects and advantages of an intelligent tracking type ray inspection area adaptive shielding control system and method of the present invention: By collecting human three-dimensional images in real time, combining grayscale analysis and adaptive segmentation algorithms, dividing the human three-dimensional images into tissue regions, and intelligently determining the occlusion regions that do not require ray irradiation, it provides a reliable information basis for subsequent precise occlusion; according to the regional space information of the protection components, using the iterative optimization algorithm to dynamically coordinate the positions and angles of each protection component, realizing precise adaptive occlusion of the occlusion regions, and minimizing the irradiation of rays on surrounding healthy tissues to the greatest extent; adopting the fusion tracking technology to monitor the movement trajectory of the human body during the inspection process in real time, and intelligently following and adjusting the protection components according to the dynamic changes of the occlusion regions to ensure the continuity and real-time nature of the protection and occlusion effect under the condition of human body movement; it can effectively reduce the radiation impact of rays on healthy tissues, improve the protection safety and inspection efficiency during the ray inspection process, and is of great significance for improving the medical service level and protecting the health of patients. Description of the Drawings

[0017] Figure 1 It is a schematic diagram of an intelligent tracking type ray inspection area adaptive shielding control system according to Embodiment 1 of the present invention; Figure 2 It is a flowchart of an intelligent tracking type ray inspection area adaptive shielding control method according to Embodiment 2 of the present invention. Detailed Embodiments

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Embodiment 1 Please refer to Figure 1 As shown, the intelligent tracking type ray inspection area adaptive shielding control system described in this embodiment includes an image acquisition module, a human body recognition module, an image segmentation module, a region determination module, a protection control module, a motion tracking module, and a dynamic control module; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0020] The image acquisition module is used to collect human three-dimensional images in real time.

[0021] The three-dimensional human body image is a three-dimensional image data containing the surface shape, spatial position, and depth information of the human body, usually represented in the form of point cloud data. The point cloud data includes a large number of point cloud points, and each point cloud point contains three-dimensional coordinates and color values. The coordinate value in the three-dimensional coordinates is the depth value; the three-dimensional human body image is obtained by a high-precision depth camera installed in the detection space (i.e., the place where the human body undergoes radiographic examination); among them, the human body surface shape is the external contour, surface characteristics, and local detail information of the human body; the spatial position is the specific position of the human body in the three-dimensional coordinate system established by the high-precision depth camera; the depth information is the distance information between the human body and the camera.

[0022] The human body recognition module is used to perform grayscale analysis on the three-dimensional human body image and identify and extract the human body contour area from the three-dimensional human body image.

[0023] The method for identifying and extracting the human body contour area from the three-dimensional human body image includes: Obtain the depth value of each point cloud point in the three-dimensional human body image, preset the depth range value, and compare the depth value of each point cloud point with the depth range value respectively; if the depth value is within the depth range value, mark the corresponding point cloud point as a candidate point; perform grayscale processing on each candidate point (i.e., convert the color value corresponding to each candidate point into a grayscale value) to obtain the grayscale value of each candidate point; preset the grayscale threshold, and compare the grayscale value of each candidate point with the grayscale threshold respectively; if the grayscale value is greater than or equal to the grayscale threshold, mark the corresponding candidate point as a contour point; if the grayscale value is less than the grayscale threshold, do not mark the corresponding candidate point; perform morphological processing (such as connected region analysis, edge smoothing, etc.) on all contour points to obtain region points, and use the region composed of all region points as the human body contour area; among them, the depth range value and the grayscale threshold are both preset by those skilled in the art according to the actual situation; the grayscale processing and the morphological processing are both existing technologies, and the specific process will not be elaborated here.

[0024] The image segmentation module is used to perform refined segmentation on the human body contour area based on an adaptive segmentation algorithm to obtain tissue regions, where t is an integer greater than 1.

[0025] The tissue region refers to a regional unit with specific functions or structural characteristics in human anatomy and is a common object of concern in radiographic examinations; each tissue region usually corresponds to a specific anatomical structure or organ system and has relatively independent medical diagnostic value; tissue regions include, for example, the cranial region, facial region, shoulder region, thigh region, heart region, upper abdominal region, etc.

[0026] The steps to obtain tissue regions include: Step S101: Screen out region points as tissue points from all region points, and use the region points that are not used as tissue points as segmentation points. The specific value of is preset by those skilled in the art according to the actual situation of performing radiographic examinations on the human body. Step S102: Construct corresponding tissue sets based on tissue points, and calculate the Euclidean distance from each segmentation point to each tissue point in turn, and mark it as the point distance; the Euclidean distance is prior art, and the specific calculation process will not be elaborated here. Step S103: Compare the point distances corresponding to the same segmentation point, mark the tissue point corresponding to the smallest point distance as the nearest point, and divide each segmentation point into the tissue set corresponding to the nearest point in turn. Step S104: Calculate the corresponding new tissue point for each tissue set, and replace each tissue point with the corresponding new tissue point. Step S105: Loop steps S102 to S104 until each new tissue point calculated in step S104 is the same as the corresponding new tissue point calculated in the previous loop process, then the loop ends, and tissue sets and the region points in each tissue set are obtained. Step S106: Segment the human body contour region according to the regions formed by the region points in each tissue set, and obtain

[0027] In the above step S101, the step of screening out region points as tissue points includes: Step S201: Randomly select a region point as a tissue point. Step S202: Calculate the point distance from each segmentation point to each tissue point, compare the point distances corresponding to the same segmentation point, and use the smallest point distance as the minimum distance corresponding to the segmentation point. Step S203: Square the minimum distance of each segmentation point in turn to obtain the squared distance; add the squared distances of each segmentation point in turn to obtain the sum of squared distances; divide the squared distance of each segmentation point by the sum of squared distances to obtain the screening probability corresponding to each segmentation point. Step S204: Screen out a region point as a tissue point according to the screening probability. Step S205: Loop steps S202 to S204 until region points are obtained, and the loop ends.

[0028] In the above step S104, the method for calculating the new tissue point includes: Count the number of regional points in each tissue region and label it as the number of regional points; sum up each regional point corresponding in each tissue region in sequence, and then divide by the corresponding number of regional points to obtain the average point corresponding to each tissue region; calculate the point distance from each regional point in each tissue region to the corresponding average point and label it as the average distance; compare the average distances of the same tissue regions, and use the regional point corresponding to the smallest average distance value as the new tissue point of the corresponding tissue region; it should be noted that since each regional point includes multiple dimensions (i.e., three coordinate values and grayscale values), when calculating the average point, it is necessary to perform summation and averaging processing on each dimension separately to ensure that the data on each dimension is fully considered.

[0029] In the above step S204, the method for screening regional points according to the screening probability includes: Sort each regional point according to the corresponding screening probability from high to low, and sequentially add the screening probability of each regional point to the screening probability of the corresponding previous point according to the positive order to obtain the cumulative probability of each regional point; calculate the probability range of each regional point according to the cumulative probability of each regional point; generate a random number from the interval and mark the probability range where the random number falls as the screening range, and screen out the regional points corresponding to the screening range; among them, if a regional point is marked as the current point, the previous point of the current point is all the regional points arranged in front of the current point; the maximum value of the probability range corresponding to the current point is the cumulative probability corresponding to the current point, and the minimum value of the probability range corresponding to the current point is the cumulative probability corresponding to the regional point ranked one position in front of the current point.

[0030] The region determination module is used to perform intelligent analysis on each tissue region respectively to determine the shielding regions in the tissue region that do not require ray inspection.

[0031] The method for determining the shielding region includes: Obtain the inspection region. The inspection region is the tissue region that needs to be irradiated during the current ray inspection of the human body. The inspection region is manually selected or input and confirmed by the doctor through the ray inspection equipment interface. Mark all tissue regions that are not the inspection region as non-inspection regions; obtain the protection feature data. The protection feature data is the feature data of the protection component. The protection feature data includes the component shape, component size, and component quantity; among them, the protection component is a movable device or device that prevents ray radiation from causing harm to the human body, used to shield the regions of the human body that do not need to be irradiated, usually made of lead or other radiation-proof materials, and can effectively absorb or block rays; the component shape includes the shape of each protection component, the component size includes the size of each protection component (such as length, width, etc.), and the component quantity is the quantity of all protection components; Calculate the area of each non-inspected area and label it as the area of the region; use the protection feature data and the area of the region as analysis data, and input the analysis data into the trained component division model to predict the corresponding component division matrix; wherein, the component division model is a deep neural network model, and the component division matrix is matrix, , is the number of non-inspected areas, is the number of components; the elements in the component division matrix correspond one by one to the component labels, and the component labels are the digital labels of the protection components, and the digital labels corresponding to different protection components are all different. The range of the component labels is , 0 indicates the absence of a protection component, and the elements in each column of the component division matrix correspond one by one to the non-inspected areas; According to the component division matrix, obtain the protection component corresponding to each non-inspected area, and use geometric segmentation methods (such as Voronoi segmentation, quad-tree, etc.) to perform area segmentation on each non-inspected area to obtain the corresponding occlusion areas, is the number of protection components corresponding to a non-inspected area.

[0032] Exemplarily, the number of components is 5, the number of non-inspected areas is 3, and the component division matrix is , so the first non-inspected area corresponds to the first and third protection components, the second non-inspected area corresponds to the fourth and fifth protection components, and the third non-inspected area corresponds to the second protection component.

[0033] The method for calculating the area of each non-inspected area is as follows: obtain the area points of each non-inspected area and label them as calculation points; use the Delaunay triangulation method to divide the calculation points corresponding to each non-inspected area into multiple triangles respectively; use the cross product calculation formula to calculate the area of each triangle, and add up the areas of the triangles corresponding to each non-inspected area in sequence to obtain the area of each non-inspected area; the Delaunay triangulation method and the cross product calculation formula are both existing technologies, and the specific process will not be elaborated here.

[0034] The training process of the component division model includes: Pre-collect groups of different analysis data, and set the corresponding component division matrix for groups of analysis data, is an integer greater than 1. Convert the analysis data and the corresponding component division matrix into a corresponding set of feature vectors; the component division matrix corresponding to the analysis data is collected by those skilled in the art during the process of historically determining the occlusion area A set of different analysis data, combined with the actual situation, different combinations of the protection components in each set of analysis data are carried out in turn, so that the combined protection components can completely block the corresponding non-inspection areas. According to the combination results, a component division matrix is constructed, and for each set of different analysis data, a corresponding component division matrix is set; Taking each set of feature vectors as the input of the component division model, the component division model takes a set of predicted component division matrices corresponding to each set of analysis data as the output, and takes the actual component division matrix corresponding to each set of analysis data as the prediction target. The actual component division matrix is the pre-set component division matrix corresponding to the analysis data; taking minimizing the sum of the prediction errors of all analysis data as the training target; among them, the calculation formula of the prediction error is where is the prediction error, is the group number of the feature vector corresponding to the analysis data, is the predicted component division matrix corresponding to the th set of analysis data,

[0035] The protection control module is used to obtain the regional space information of each occlusion area, and based on the regional space information, coordinate and control each protection component to occlude the corresponding occlusion area.

[0036] The regional space information is the specific position of the occlusion area in the three-dimensional space. The method for obtaining the regional space information is as follows: obtain the regional points of each occlusion area and mark them as occlusion points, and use boundary detection algorithms (such as convex hull algorithm, normal vector change detection algorithm, etc.) to analyze the occlusion points of each occlusion area respectively, and extract the corresponding boundary points from the occlusion points of each occlusion area. The boundary points are the occlusion points located on the outer contour or edge of the occlusion area; obtain the three-dimensional coordinates of each boundary point, and generate the regional space information corresponding to each occlusion area according to the three-dimensional coordinates of the boundary points corresponding to each occlusion area.

[0037] The method for coordinating and controlling the protection components includes: Randomly construct sets of different component position sets, is an integer greater than 1. Each set of component position sets includes component positions. Each component position includes coordinate values, coordinate values, coordinate values, the rotation angle around the axis, the rotation angle around the axis, and the rotation angle around The shaft rotation angle corresponds one-to-one with the component positions and the protection components; sequentially incrementally set digital tags for each set of component positions and mark them as set tags, and the range of the set tags is ; take add 1 and divide by 2 as the initial control center, and take subtract 1 and divide by 2 as the initial control radius; Define an iterative process, and the iterative process is: generate candidate solutions within the range of the set tags, and the candidate solutions correspond one-to-one with the set tags, ; sequentially calculate the effective occlusion coefficients corresponding to each candidate solution, move the control center to the candidate solution with the largest effective occlusion coefficient, and adjust the control radius; execute the iterative process. When the number of iterations is greater than or equal to the preset iteration threshold, the iterative process ends, mark the candidate solution corresponding to the control center as the optimal solution, obtain the set of component positions corresponding to the set tag of the optimal solution and mark it as the optimal set, and perform coordinated control on each protection component according to the optimal set; the iteration threshold is preset by those skilled in the art according to the actual situation.

[0038] The method for randomly constructing sets of different component position sets is: obtain the coordinate range and the angle range. The coordinate range includes coordinate range, coordinate range, and coordinate range. The coordinate range is obtained by those skilled in the art according to the specific position of the detection space in the three-dimensional coordinate system established by the high-precision depth camera. The angle range is ; randomly select a value from each range within the coordinate range and randomly select three values from the angle range to construct a set of component positions, and a total of sets of different component position sets are constructed.

[0039] The method for generating candidate solutions is: generate a random coefficient in the interval , multiply the control radius by the random coefficient and add the control center as a candidate solution; the method for adjusting the control radius is: randomly generate an adjustment coefficient in the interval , multiply the control radius by the adjustment coefficient to obtain the adjusted radius, and adjust the control radius according to the adjusted radius.

[0040] The calculation method of the effective occlusion coefficient is as follows: Obtain the set label corresponding to the candidate solution and mark it as the current label, and obtain the set of component positions corresponding to the current label and mark it as the current position set; Use the regional spatial information of each occlusion area and the current position set as calculation data, input the calculation data into the trained occlusion prediction model, and predict the corresponding occlusion degree. The occlusion degree is the occlusion effect of each protection component on the corresponding occlusion area, and the range of the occlusion degree is , where 0 means that the protection component has no occlusion on the corresponding occlusion area, and 1 means that the protection component completely occludes the corresponding occlusion area; Input the calculation data into the trained collision prediction model, and predict the corresponding collision degree. The collision degree is the collision degree between each protection component, and the range of the collision degree is , where 0 means that there is no collision between all protection components, and 1 means that all protection components are completely overlapped; Subtract the collision degree from the occlusion degree to obtain the effective occlusion coefficient; The training processes of the occlusion prediction model and the collision prediction model are both the same as the training process of the component division model, and both are deep neural network models.

[0041] The motion tracking module is used to adopt the fusion tracking technology to detect the moving trajectory of the human body during the inspection in real time, and generate corresponding movement change data for each occlusion area.

[0042] The method for generating movement change data includes: Obtain the boundary points corresponding to each occlusion area and use them as the reference frame point cloud data of the corresponding occlusion area; Create a corresponding filter (such as Kalman filter, particle filter, etc.) for each occlusion area, and initialize the filter parameters. The filter parameters include the state vector, state transition matrix, observation matrix, and error covariance matrix; Among them, the state vector includes the regional position and regional attitude. The calculation method of the regional position is: Count the number of boundary points corresponding to each occlusion area and mark it as the number of boundary points. Add the three-dimensional coordinates of the boundary points corresponding to each occlusion area in sequence, and then divide by the number of boundary points to obtain the regional position of each occlusion area; that is, add the coordinate values of the boundary points corresponding to each occlusion area in sequence, and then divide by the number of boundary points to obtain the coordinate values of the corresponding regional position of each occlusion area; Add the coordinate values of the boundary points corresponding to each occlusion area in sequence, and then divide by the number of boundary points to obtain the coordinate values of the corresponding regional position of each occlusion area; Add the coordinate values of the boundary points corresponding to each occlusion area in sequence, and then divide by the number of boundary points to obtain the Coordinate values; the calculation method of the regional attitude is as follows: the least squares method is used to generate the normal vector of each occluded area, and then the Rodrigues rotation formula is used to convert the normal vector of each occluded area into the corresponding rotation matrix, and the rotation matrix of each occluded area is used as the corresponding regional attitude; the creation method of the filter, the least squares method, and the Rodrigues rotation formula are all prior arts, and the specific process will not be elaborated here; the state transition matrix is used to describe how the state changes over time and connects the current state with the state at the next moment; the observation matrix is used to map the state vector to the observation space; the error covariance matrix is used to describe the uncertainty of the system state; the state transition matrix, the observation matrix, and the error covariance matrix are all preset by those skilled in the art according to the actual situation. During the human body movement, the boundary points corresponding to each occluded area are obtained in real time and used as the current frame point cloud data of the corresponding occluded area; the iterative closest point algorithm is used to register the current frame point cloud data with the reference frame point cloud data, and the change data of each occluded area is calculated. The change data includes displacement change (i.e., the spatial position change of the occluded area during the human body movement) and angle change (i.e., the rotation change of the occluded area during the human body movement); the change data of each occluded area is input into the corresponding filter for smoothing processing to generate the corresponding smoothed data. The smoothed data includes smoothed displacement change and smoothed angle change; the smoothed data of each occluded area is used as the corresponding movement change data; the iterative closest point algorithm is a prior art, and the specific process will not be elaborated here.

[0043] It should be understood that the fusion tracking technology can effectively combine the spatial information of the point cloud data and the state estimation ability of the filter to achieve target tracking and smooth prediction in a dynamic environment, show high robustness when dealing with dynamic targets and environmental changes, and can correct and optimize the movement trajectory of the target in real time, so as to achieve efficient target tracking and dynamic control in complex scenarios.

[0044] A dynamic control module is used to dynamically control each protection component corresponding to each occluded area according to the movement change data.

[0045] The method for dynamically controlling the protection component includes: Obtain the component position corresponding to each protection component in the optimal set and mark it as the real-time position; use each group of real-time positions and the movement change data as the movement prediction data, input the movement prediction data into the trained component movement model, and predict the corresponding movement position. The movement position is the component position after the protection component moves; dynamically control each protection component corresponding to each occluded area according to the movement position; the training process of the component movement model is the same as that of the component division model, and both are deep neural network models.

[0046] It should be noted that the protection component is a movable protective screen, which is composed of lead, tungsten, steel, polyethylene or other plastics; each protection component is connected to a robotic arm, and the robotic arm controls the movement of the protection component; the control principle of the robotic arm is to calculate the joint angles of the robotic arm through an accurate kinematic model to ensure that the protection component can reach the predetermined spatial position; during the control process, the robotic arm adjusts the motion trajectory and posture in real time according to the joint angle and position information feedback by the sensor to ensure that the protection component can accurately block the area that does not need to be irradiated during the ray inspection; when the human body moves, the robotic arm can dynamically adjust according to the movement change data, accurately move the protection component, so that the protection component can continuously and effectively block the rays and protect the surrounding healthy tissues from radiation effects.

[0047] In this embodiment, by collecting the three-dimensional human body image in real time, combining grayscale analysis and adaptive segmentation algorithm, the three-dimensional human body image is divided into tissue regions, and the occlusion regions that do not need to be irradiated by rays are intelligently determined, providing a reliable information basis for subsequent accurate occlusion; according to the regional spatial information of the protection components, the iterative optimization algorithm is used to dynamically coordinate the positions and angles of each protection component to achieve accurate adaptive occlusion of the occlusion regions, minimizing the irradiation of rays to the surrounding healthy tissues; the fusion tracking technology is adopted to monitor the motion trajectory of the human body during the inspection process in real time, and the protection components are intelligently followed and adjusted according to the dynamic changes of the occlusion regions to ensure the continuity and real-time nature of the protection and occlusion effect under the condition of human body movement; it can effectively reduce the radiation impact of rays on healthy tissues, improve the protection safety and inspection efficiency during the ray inspection process, and is of great significance for improving the medical service level and protecting the health of patients.

[0048] Embodiment 2 Please refer to Figure 2 As shown, for the parts not described in detail in this embodiment, refer to the description content of Embodiment 1. A method for adaptively shielding and controlling an intelligent tracking type ray inspection area is provided, and the method includes: Collect the three-dimensional human body image in real time; Perform grayscale analysis on the three-dimensional human body image, and identify and extract the human body contour area from the three-dimensional human body image; Based on the adaptive segmentation algorithm, perform refined segmentation on the human body contour area to obtain tissue regions, where t is an integer greater than 1; Perform intelligent analysis on each tissue region respectively to determine the occlusion regions within the tissue region that do not need to be subjected to ray inspection; Obtain the regional spatial information of each occlusion region, and based on the regional spatial information, coordinately control each protection component to occlude the corresponding occlusion region; Adopt the fusion tracking technology to detect the moving trajectory of the human body during the inspection in real time, and generate corresponding moving change data for each occluded area.

[0049] Embodiment 3 This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. Among them, computer-readable code is stored in the memory, and when the computer-readable code is run by one or more processors, it can execute an intelligent tracking type ray inspection area adaptive shielding control method as described above.

[0050] The method or system according to the embodiment of the present application can also be implemented by means of the architecture of the electronic device shown in the present application. The electronic device may include a bus, one or more CPUs, a ROM, a RAM, a communication port connected to the network, an input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or a hard disk, can store an intelligent tracking type ray inspection area adaptive shielding control method provided by the present application. Further, the electronic device may further include a user interface. Of course, the architecture shown in the present application is only exemplary, and when implementing different devices, one or more components shown in the electronic device of the present application can be omitted according to actual needs.

[0051] Embodiment 4 One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are run by a processor, they can execute an intelligent tracking type ray inspection area adaptive shielding control method according to the embodiment of the present application described with reference to the above drawings. The storage medium includes but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0052] In addition, according to the embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, and the non-transitory machine-readable storage medium stores machine-readable instructions, and the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, such as: an intelligent tracking type ray inspection area adaptive shielding control method. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0053] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims described above.

[0054] Finally: The above description is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent tracking type ray inspection area adaptive shielding control method, characterized in that: include: Real-time acquisition of human body three-dimensional images; Perform grayscale analysis on the three-dimensional image of the human body, and identify and extract the human body contour area from the three-dimensional image of the human body; Based on the adaptive segmentation algorithm, the human body contour area is segmented finely to obtain organizational area, is an integer greater than 1; Intelligent analysis is performed on each tissue area to determine the shielded areas within the tissue area that do not require radiographic inspection; Obtaining regional spatial information of each shielding area, and based on the regional spatial information, coordinating and controlling each protection component to shield the corresponding shielding area; Adopt fusion tracking technology to detect the movement trajectory of the human body in the inspection process in real time, and generate corresponding movement change data for each blocked area; According to the movement change data, the protection components corresponding to each shielding area are dynamically controlled.

2. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 1, characterized in that: The human body three-dimensional image is stereoscopic image data containing human body surface shape, spatial position and depth information; The method for identifying and extracting a human body contour area from a human body three-dimensional image comprises: The depth value of each point cloud point in the three-dimensional image of the human body is obtained, and a depth range value is preset, and the depth value of each point cloud point is compared with the depth range value; if the depth value is within the depth range value, the corresponding point cloud point is marked as a candidate point; each candidate point is grayed to obtain the gray value of each candidate point; a gray threshold is preset, and the gray value of each candidate point is compared with the gray threshold; if the gray value is greater than or equal to the gray threshold, the corresponding candidate point is marked as a contour point; if the gray value is less than the gray threshold, the corresponding candidate point is not marked; morphological processing is performed on all contour points to obtain regional points, and the area composed of all regional points is used as the human body contour area.

3. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 2, characterized in that: The acquisition The steps to organize a region include: Step S101: Filter out from all regional points The regional points are used as organizing points, and the regional points that are not used as organizing points are used as segmentation points; Step S102: According to The corresponding organization points are constructed A set of tissues is created, and the Euclidean distance from each segmentation point to each tissue point is calculated in turn, and marked as the point spacing; Step S103: comparing the point spacings of corresponding segmentation points, marking the tissue point corresponding to the smallest point spacing as the nearest point, and dividing each segmentation point into the tissue set corresponding to the nearest point in turn; Step S104: Calculate the new tissue point corresponding to each tissue set, and replace each tissue point with the corresponding new tissue point; Step S105: looping steps S102 to S104 until each new tissue point calculated in step S104 is consistent with the corresponding new tissue point calculated in the previous loop, the loop ends, and the tissue sets and regional points in each tissue set; Step S106: Segment the human body contour area according to the area composed of the regional points in each tissue set to obtain organizational area; In step S101, the The steps to use a regional point as an organizing point include: Step S201: randomly selecting a regional point as an organization point; Step S202: Calculate the point spacing from each segmentation point to each tissue point, compare the point spacings with the same value of corresponding segmentation points, and take the point spacing with the smallest value as the minimum distance of the corresponding segmentation point; Step S203: Squaring the minimum distance of each segmentation point in turn to obtain the distance square; adding the distance squares of each segmentation point in turn to obtain the distance square sum; dividing the distance square of each segmentation point by the distance square sum to obtain the screening probability corresponding to each segmentation point; Step S204: selecting a regional point as an organization point according to the screening probability; Step S205: loop through steps S202 to S204 until the The cycle ends.

4. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 3, characterized in that: In step S104, the method for calculating the new tissue point includes: Count the number of regional points in each tissue region and mark them as the number of regional points; add the corresponding regional points in each tissue region in turn, and then divide by the corresponding number of regional points to obtain the average point corresponding to each tissue region; calculate the point distance from each regional point to the corresponding average point in each tissue region, and mark it as the average distance; compare the same average distances of corresponding tissue regions, and use the regional point corresponding to the average distance with the smallest value as the new tissue point of the corresponding tissue region; In step S204, the method of selecting regional points according to the selection probability includes: Sort each regional point from high to low according to the corresponding screening probability, add the screening probability of each regional point to the screening probability of the corresponding preceding point in positive order, and obtain the cumulative probability of each regional point; calculate the probability range of each regional point according to the cumulative probability of each regional point; Generate a random number, mark the probability range that the random number falls into as the screening range, and screen out the regional points corresponding to the screening range; among them, mark a regional point as the current point, and the predecessor points of the current point are all regional points that are ranked before the current point; the maximum value of the probability range corresponding to the current point is the cumulative probability corresponding to the current point, and the minimum value of the probability range corresponding to the current point is the cumulative probability corresponding to the regional point ranked before the current point.

5. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 4, characterized in that: Methods for determining occluded areas include: Obtain an inspection area, which is a tissue area of ​​the human body that needs to be irradiated during this X-ray inspection, and mark all tissue areas that are not inspection areas as non-inspection areas; obtain protection feature data, which is feature data of protection components, including component shape, component size, and component quantity; wherein the component shape includes the shape of each protection component, the component size includes the size of each protection component, and the component quantity is the quantity of all protection components; The area of ​​each non-inspection area is calculated and marked as the area of ​​the area. The protection feature data and the area of ​​the area are used as analysis data, and the analysis data is input into the trained component partitioning model to predict the corresponding component partitioning matrix. The component partitioning model is a deep neural network model, and the component partitioning matrix is The matrix of , is the number of non-inspected areas, is the number of components; the elements in the component partition matrix correspond to the component labels one by one. The component labels are the digital labels of the protection components. Different protection components have different corresponding digital labels. The range of component labels is ,0 means there is no protection component, and the elements of each column in the component partition matrix correspond one-to-one to the non-inspection area; According to the component partitioning matrix, the corresponding protection components of each non-inspection area are obtained, and the geometric segmentation method is used to segment each non-inspection area to obtain the corresponding The occluded area, The number of protective components corresponding to a non-inspection area; The method for calculating the area of ​​each non-inspection area is as follows: obtain the area points of each non-inspection area and mark them as calculation points; use the Delaunay triangulation method to divide the calculation points corresponding to each non-inspection area into multiple triangles; use the cross product calculation formula to calculate the area of ​​each triangle, and add the areas of the triangles corresponding to each non-inspection area in turn to obtain the area of ​​each non-inspection area.

6. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 5, characterized in that: The method for obtaining regional spatial information is as follows: obtaining regional points of each occluded area and marking them as occluded points, respectively analyzing the occluded points of each occluded area using a boundary detection algorithm, extracting corresponding boundary points from the occluded points of each occluded area, where the boundary points are occluded points located at the outer contour or edge of the occluded area; obtaining the three-dimensional coordinates of each boundary point, and generating the regional spatial information corresponding to each occluded area according to the three-dimensional coordinates of the boundary points corresponding to each occluded area; Methods for coordinating control of protection components include: Random Build A different set of component locations, is an integer greater than 1, and each set of component positions includes component locations, each of which includes Coordinate value, Coordinate value, Coordinate value, Axis rotation angle, Axis rotation angle and The axis rotation angle, component position and protection component correspond one by one; set digital labels for each set of component position sets in ascending order and mark them as set labels. The range of the set label is ;Will Add 1 and divide by 2 as the initial control center. Subtract 1 and divide by 2 to get the initial control radius; Define the iteration process: Generate within the range of the collection tag candidate solutions, each of which corresponds to a set of labels. ; Calculate the effective occlusion coefficient corresponding to each candidate solution in turn, move the control center to the candidate solution with the largest effective occlusion coefficient, and adjust the control radius; The iterative process is executed. When the number of iterations is greater than or equal to the preset iteration threshold, the iterative process ends, the candidate solution corresponding to the control center is marked as the best solution, the component position set corresponding to the set label corresponding to the best solution is obtained, and marked as the best set, and each protection component is coordinated and controlled according to the best set.

7. The method for adaptive shielding control of an intelligent tracking ray inspection area according to claim 6, characterized in that: Random Build The method of grouping different component position sets is: Get the coordinate range and angle range. The coordinate range includes Coordinate range, Coordinate range and Coordinate range, angle range is ; Randomly select a value from each range in the coordinate range, and randomly select three values ​​from the angle range to construct a set of component positions, constructing a total of A set of different component locations; The method for generating candidate solutions is: Generate a random coefficient in the interval, multiply the control radius by the random coefficient and add the control center as a candidate solution; the method for adjusting the control radius is: A random adjustment coefficient is generated, the control radius is multiplied by the adjustment coefficient to obtain the adjustment radius, and the control radius is adjusted according to the adjustment radius; The calculation method of the effective occlusion coefficient is as follows: obtain the set label corresponding to the candidate solution and mark it as the current label, obtain the component position set corresponding to the current label and mark it as the current position set; The regional spatial information and current position set of each occluded area are used as calculation data, and the calculation data is input into the trained occlusion prediction model to predict the corresponding occlusion degree. The occlusion degree is the occlusion effect of each protective component on the corresponding occluded area. The range of occlusion degree is ; Input the calculated data into the trained collision prediction model to predict the corresponding collision degree. The collision degree is the collision degree between each protective component. The range of the collision degree is ; Subtract the collision degree from the occlusion degree to obtain the effective occlusion coefficient; The training process of the occlusion prediction model and the collision prediction model are consistent with the training process of the component partitioning model, and both are deep neural network models.

8. The method for controlling the adaptive shielding of an intelligent tracking ray inspection area according to claim 7, characterized in that: Methods for generating movement change data include: Get the boundary points corresponding to each occluded area and use them as the reference frame point cloud data of the corresponding occluded area; create a corresponding filter for each occluded area and initialize the filter parameters, which include the state vector, state transfer matrix, observation matrix and error covariance matrix; the state vector includes the area position and area posture, and the area position is calculated by counting the number of boundary points corresponding to each occluded area and marking them as the number of boundary points, adding the three-dimensional coordinates of the boundary points corresponding to each occluded area in sequence, and then dividing them by the number of boundary points to obtain the area position of each occluded area; the area posture is calculated by using the least squares method to generate the normal vector of each occluded area, and then using the Rodrigues rotation formula to convert the normal vector of each occluded area into the corresponding rotation matrix, and using the rotation matrix of each occluded area as the corresponding area posture; During human body movement, the boundary points corresponding to each occluded area are obtained in real time and used as the current frame point cloud data of the corresponding occluded area; the iterative nearest point algorithm is used to align the current frame point cloud data with the reference frame point cloud data, and the change data of each occluded area is calculated, and the change data includes displacement change and angle change; the change data of each occluded area is input into the corresponding filter for smoothing to generate corresponding smoothed data, and the smoothed data includes smoothed displacement change and smoothed angle change; the smoothed data of each occluded area is used as the corresponding movement change data.

9. The method for controlling the adaptive shielding of an intelligent tracking ray inspection area according to claim 8, characterized in that: Methods for dynamically controlling protection components include: Obtain the component position corresponding to each protection component in the best set and mark it as the real-time position; use each set of real-time position and movement change data as movement prediction data, input the movement prediction data into the trained component movement model, and predict the corresponding movement position, which is the component position after the protection component moves; dynamically control the protection component corresponding to each occluded area according to the movement position; the training process of the component movement model is consistent with the training process of the component partitioning model, and both are deep neural network models.

10. An intelligent tracking type ray inspection area adaptive shielding control system, implementing an intelligent tracking type ray inspection area adaptive shielding control method according to any one of claims 1 to 9, characterized in that: include: An image acquisition module, used for acquiring three-dimensional images of the human body in real time; A human body recognition module is used to perform grayscale analysis on a three-dimensional human body image, and to recognize and extract a human body contour area from the three-dimensional human body image; The image segmentation module is used to perform fine segmentation of the human body contour area based on the adaptive segmentation algorithm to obtain organizational area, is an integer greater than 1; A region determination module is used to perform intelligent analysis on each tissue region to determine the shielded regions within the tissue region that do not require radiographic inspection; A protection control module is used to obtain the regional spatial information of each shielding area, and based on the regional spatial information, coordinate and control each protection component to shield the corresponding shielding area; The motion tracking module is used to detect the movement trajectory of the human body during the inspection process in real time using fusion tracking technology, and generate corresponding movement change data for each occluded area; The dynamic control module is used to dynamically control the protection components corresponding to each shielding area according to the movement change data.

Images