Radioactive article automatic identification method based on machine vision

Abstract

The invention belongs to the technical field of machine vision automatic identification, and particularly relates to a radioactive article automatic identification method based on machine vision, and the method comprises the steps: labels on radioactive articles are automatically identified and clamped, and the labels are square labels with the same size, different from the background color and different in color; according to the method, the radioactive article is measured, the radiation data of the radioactive article is measured, and the radiation field is reconstructed according to the radiation data, so that the problems that working personnel are affected and the measured and acquired data are insufficient when the space dose rate in the retired environment is acquired are solved. According to the radioactive substance clamping and measuring device, the complexity and danger of radioactive substance processing in the nuclear industry are dealt with, clamping and measuring of radioactive substances are carried out through the mechanical arm, the clamping accuracy is improved, the working efficiency is greatly improved, and the safety of workers is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine vision automatic identification, and in particular is a method for automatic identification of radioactive objects based on machine vision. Background Art

[0002] Nuclear facility decommissioning management is an important part of the construction of an advanced nuclear industry system, a frontier area in the development of international nuclear science, a typical high-tech industry, and an important issue related to the overall safe and sustainable development of the nuclear industry.

[0003] During the long-term operation and decommissioning of nuclear facilities, many issues such as sampling, measuring, sorting, and decontamination of radioactive materials are involved. Some of these items may have high radioactivity. In order to minimize the radiation dose received by workers during decommissioning operations, it is necessary to plan their movement trajectories reasonably and obtain the spatial dose rate in the decommissioning environment. Traditional radiation monitoring methods mainly rely on the deployment of a limited number of point detectors. Although they can obtain radiation readings at local locations, it is difficult to reflect the distribution of the radiation field in the entire three-dimensional space, which brings great difficulties in path planning. Summary of the invention

[0004] The present invention provides a method for automatically identifying radioactive materials based on machine vision, which solves the problems of affecting workers and insufficient data obtained by measurement when acquiring spatial dose rate in a decommissioning environment.

[0005] The present invention is achieved in this way.

[0006] A method for automatic detection of spatial dose of radioactive articles based on machine vision, the method comprising:

[0007] Automatically identify and pick up labels on radioactive items, where the labels are square labels of the same size, different from the background color and different color;

[0008] Place the radioactive items in the measurement area and measure the radiation data of the radioactive items, and reconstruct the radiation field based on the radiation data;

[0009] Among them: Automatic identification and gripping of labels on radioactive items include:

[0010] Use a robot to carry a camera to take images of radioactive items and convert the images from RGB to HSV color space;

[0011] Find the color comparison table of HSV color space, set the color threshold to extract the required color, build a mask based on the threshold, and find the target color block;

[0012] Perform closing and opening operations on the target color block in sequence to obtain the object outline;

[0013] And draw the minimum right rectangular frame in the original image to obtain the circumscribed rectangle of the object outline;

[0014] Calculate the coordinates of the four vertices, length and width of the minimum rectangular box that encloses the object outline;

[0015] Based on the length and width, determine whether the robot can grasp the experimental object.

[0016] Furthermore, when there is only one radioactive object in the image of the radioactive object taken, the area of ​​the circumscribed rectangle is calculated and compared in size, and the circumscribed rectangle with the largest area is used as the target label area, and the circumscribed rectangle with the largest area is used as the circumscribed rectangle of the object contour.

[0017] Furthermore, based on the circumscribed rectangle of the object outline corresponding to the radioactive item returned in real time, the position information is determined and the robotic arm is controlled to drive the robotic hand to approach the target radioactive item to be grasped, so that the coordinates and area of ​​the circumscribed rectangle of the object outline finally reach the threshold range, and the grasping action is executed.

[0018] Further, the determining and obtaining the location information includes:

[0019] The horizontal and vertical coordinates of the center point of the circumscribed rectangle of the object outline are used to determine the position deviation of the robot up, down, left, and right, and the area is used to determine the position deviation of the robot near and far.

[0020] When the coordinates and the rectangular area of ​​the circumscribed rectangle of the object outline are both within the threshold range, the robot arm performs a grasping action.

[0021] Furthermore, when the image of radioactive items contains multiple radioactive items, multiple circumscribed rectangles are found, the areas of the circumscribed rectangles are calculated and sorted from large to small, and the circumscribed rectangles corresponding to the first N areas are found, where N is the number of radioactive items, and the sizes of the first N areas are determined. If the area of ​​the label attached to the target radioactive item is not the smallest among the three, it is considered that there is no occlusion, and the manipulator is controlled to directly grab the target radioactive item;

[0022] If the area of ​​the label attached to the radioactive item is the smallest among the three, it is considered that there is obstruction.

[0023] Furthermore, when the robot is grabbing, in order to determine which one is blocking the target radioactive item, it is necessary to calculate the horizontal coordinates of the center points of the N labels respectively, and subtract the horizontal coordinates of the center points of the labels of the two non-target radioactive items from the horizontal coordinates of the center points of the label of the target radioactive item, and take the absolute value. If the absolute value of the difference with the horizontal coordinate of the target radioactive item is small, it is considered that the distance to the target radioactive item is close and it is the radioactive item that blocks the target radioactive item. The robot is controlled to first clamp the radioactive item that blocks the target radioactive item and put it aside. After all subsequent operations on the target radioactive item are completed, the radioactive item that blocks the target radioactive item is returned to its place.

[0024] Furthermore, the radioactive material measurement area is set up to measure the radiation data of the radioactive material, and the radiation field is reconstructed according to the radiation data, including:

[0025] Detection point coordinate generation: Divide the radiation field into 100×100×100 grids, with a total of 10 6 The grid space is a grid space with a size of 5cm×5cm×5cm. The prediction area is set as a 95×95×95 cubic area with grid space coordinates (5,5,5) and (100,100,100) as diagonal vertices. The detection point is one block for every 8 blocks in each dimension. A total of [(95 / / 8)+1] 3 =1728 known detection points;

[0026] Prediction point coordinate generation: The prediction point is one block for every 3 blocks in each dimension, a total of [(95 / / 3)+1] 3 =32768 points;

[0027] Measure the radiation value according to the coordinates of the detection point;

[0028] Import the data measured at the detection point and use the Kriging algorithm to calculate the radiation data of the predicted point;

[0029] The radiation field is constructed based on the measured data at the detection point and the radiation data at the prediction point.

[0030] 8. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 1, characterized in that:

[0031] The Kriging algorithm is used to calculate the predicted point radiation data including:

[0032] The radiation data of the predicted point is estimated by weighted summation of the data measured by all detection points in the network space. The formula is as follows:

[0033]

[0034] in is the point (x o ,y o ), i.e., z o = z(x o ,y o ), z i is the measured value of the known point i, and n is the total number of known points used to estimate the unknown point;

[0035] λ i is the weight coefficient, which is the factor that satisfies the point (x o ,y o ) and the true value z o The optimal set of coefficients with the smallest difference is

[0036]

[0037] Var is the variance;

[0038] Satisfy the conditions for unbiased estimation:

[0039]

[0040] E is expectation;

[0041] The assumption of Kriging interpolation is that the spatial attribute z is uniform, and for any point (x, y) in space, there is the same expectation c and variance σ 2 ;

[0042] That is, for any point (x, y)

[0043] E[z(x,y)]=E[z]=c

[0044] Var[z(x,y)]=σ 2

[0045] The value z(x,y) at any point is composed of the regional average c and the random deviation R(x,y) of the point, that is,

[0046] z(x,y)=E[z(x,y)]+R(x,y)]=c+R(x,y);

[0047] Where R(x,y) represents the deviation at the point (x,y), and its variance is a constant:

[0048] Var[R(x,y)]=σ 2 .

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The method of the present invention is used to control the manipulator to clamp, which can cope with the complexity and danger of handling radioactive materials in the nuclear industry. By using the manipulator to clamp and measure radioactive materials, the accuracy of clamping is improved, the work efficiency is greatly improved and the safety of the staff is guaranteed. At the same time, the dose rate distribution in the three-dimensional radiation field is spatially inverted and analyzed by using the Kriging method. With its efficient computing power and good interpolation effect, the source term distribution in the three-dimensional space is accurately inverted based on limited detection point data. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flowchart of a method provided by an embodiment of the present invention;

[0052] Figure 2 The color space conversion of an image from RGB to HSV provided in an embodiment of the present invention, wherein (a) an RGB color space image, (b) an HSV color space image;

[0053] Figure 3 An image after mask operation provided by an embodiment of the present invention;

[0054] Figure 4 The connected domain obtained after closing and opening operations provided by the embodiment of the present invention;

[0055] Figure 5 A comparison chart of the predicted value and the actual value provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0057] See also Figure 1 As shown, a method for automatic identification of radioactive materials based on machine vision, the method comprising:

[0058] Each radioactive item has a square label of the same size, different from the background color and different color; in order to accurately obtain the location information of the radioactive item, each radioactive item has a square label of the same size, different from the background color and different color. The camera sends the captured image to the Raspberry Pi for processing, extracts the radioactive item label, and determines the location of the radioactive item for subsequent operations.

[0059] The target object can be identified in the image based on color, texture, shape and other attributes, among which color is the most widely used, so the color-based object recognition method is used to determine the target position. In this embodiment, red, orange and yellow are selected as the label colors of radioactive items.

[0060] In the processing process based on the method of the present invention, the OpenCV library is used in conjunction with the Raspberry Pi, and the video captured by the Raspberry Pi camera can be directly transmitted to the OpenCV library for various computer vision developments such as image search and edge detection. The method of the present invention is based on the system that uses the Raspberry Pi in conjunction with the Python language and the OpenCV library to realize the recognition of different color labels, thereby locating the target radioactive items.

[0061] Place the radioactive items in the measurement area and measure the radiation data of the radioactive items, and reconstruct the radiation field based on the radiation data;

[0062] See also Figure 2 As shown, the image of the radioactive material is captured by a robot carrying a camera, and the image is converted from RGB to HSV color space; Figure 2 (a) RGB color space image, Figure 2 (b) HSV color space image;

[0063] Find the color table in the HSV color space, set the color threshold to extract the required color, and construct a mask based on the threshold. Figure 2 , find the target color block;

[0064] The HSV color comparison table in this embodiment is shown in Table 1:

[0065] Table 1 HSV color comparison table

[0066]

[0067]

[0068] Since the labels of the experimental objects used in the experiment are red, orange, and yellow, the color thresholds are set as:

[0069] lower_red=[0,43,46], upper_red=[8,255,255];

[0070] lower_orange=[11,43,46], upper_orange=[25,255,255]lower_yellow=

[0071] [26,43,46]、upper_yellow=[34,255,255]

[0072] Perform closing and opening operations on the target color block in turn to obtain the connected domain of the object contour. Figure 4 As shown;

[0073] And draw the minimum right rectangular frame in the original image to obtain the circumscribed rectangle of the object outline;

[0074] Calculate the coordinates of the four vertices, length and width of the minimum rectangular box that encloses the object outline;

[0075] According to the length and width, determine whether the robot can grasp the experimental object.

[0076] Based on the circumscribed rectangle of the object outline corresponding to the radioactive item returned in real time, the position information is determined and the robotic arm is controlled to drive the manipulator to approach the target radioactive item to be grasped, so that the coordinates and area of ​​the circumscribed rectangle of the object outline finally reach the threshold range and the grasping action is executed.

[0077] After grayscale and binarization preprocessing, the collected images can be obtained through a series of operations to obtain the center point coordinates and area of ​​the circumscribed rectangle of the radioactive item label. The area information of the circumscribed rectangle of the label corresponding to the radioactive item is used to judge the distance between the manipulator and the radioactive item. The area of ​​the circumscribed rectangle is inversely proportional to the relative distance between the manipulator and the radioactive item. The center point coordinates of the circumscribed rectangle are calculated, and its ordinate is used to judge whether the radioactive item is higher or lower than the manipulator. The average value is taken from multiple experiments, and the center point coordinates of the label when the manipulator can clamp the radioactive item are measured, set as (x, y). If the ordinate is less than y, it means that the manipulator is higher, and if the ordinate is greater than y, it means that the manipulator is lower; the abscissa of the center point is used to judge whether the radioactive item is to the left or right relative to the manipulator. If the abscissa is greater than x, it means that the manipulator is to the left, and if the abscissa is less than x, it means that the manipulator is to the right.

[0078] Extract the label's circumscribed rectangle and calculate the center point coordinates through the findContours function in the OpenCV library. The function of the findContours function is to find the image contour. After obtaining the object contour, the boundingRect function that calculates the rectangular bounding box can be used to obtain the coordinates of the four vertices of the smallest regular rectangle covering this contour as well as the length and width, and the rectangle function that draws the rectangular box can be used to draw the rectangle in the original image to obtain the circumscribed rectangle of the contour.

[0079] However, due to the presence of interfering colors in the background, there may still be a small amount of interfering areas after the image is opened. At this time, when using the findContours function to find the contour, multiple contours will be obtained (the contour of the target label and the contour of each interfering area). These contours will be recognized by the boundingRect function, thereby obtaining multiple circumscribed rectangles.

[0080] In order to avoid the above situation, when there is only one radioactive object in the image of the radioactive object taken, the area of ​​the bounding rectangle is calculated and compared, and the bounding rectangle with the largest area is used as the target label area, and the bounding rectangle with the largest area is used as the bounding rectangle of the object contour.

[0081] The location information obtained includes:

[0082] The horizontal and vertical coordinates of the center point of the circumscribed rectangle of the object outline are used to determine the position deviation of the robot up, down, left, and right, and the area is used to determine the position deviation of the robot near and far.

[0083] When the coordinates and the rectangular area of ​​the circumscribed rectangle of the object outline are both within the threshold range, the robot arm performs a grasping action.

[0084] Since the position of radioactive items is not fixed, there may be a situation where the target radioactive items are blocked by non-target radioactive items. And because the clamping method of the manipulator is horizontal clamping, if the manipulator needs to clamp the blocked radioactive items, it needs to move the radioactive items in the front row away and then clamp the target radioactive items.

[0085] When the image of radioactive items contains multiple radioactive items, find multiple circumscribed rectangles, calculate the areas of the circumscribed rectangles and sort them from large to small, find the circumscribed rectangles corresponding to the first N areas, where N is the number of radioactive items, and judge the sizes of the first N areas. If the area of ​​the label attached to the target radioactive item is not the smallest among the three, it is considered that there is no occlusion, and the robot is controlled to directly grab the target radioactive item;

[0086] If the area of ​​the label attached to the radioactive item is the smallest among the three, it is considered that there is obstruction.

[0087] When the robot is grabbing, in order to determine which one is blocking the target radioactive item, it is necessary to calculate the horizontal coordinates of the center points of the N labels respectively, and subtract the horizontal coordinates of the center points of the labels of the two non-target radioactive items from the horizontal coordinates of the center points of the label of the target radioactive item, and take the absolute value. If the absolute value of the difference with the horizontal coordinate of the target radioactive item is small, it is considered that the distance to the target radioactive item is close and it is the radioactive item that blocks the target radioactive item. The robot is controlled to first clamp the radioactive item that blocks the target radioactive item and put it aside. After all subsequent operations on the target radioactive item are completed, the radioactive item that blocks the target radioactive item will be returned to its place.

[0088] The radioactive items are measured in the measurement area and the radiation data of the radioactive items are measured, and the radiation field is reconstructed based on the radiation data, including:

[0089] Detection point coordinate generation: Divide the radiation field into 100×100×100 grids, with a total of 10 6 The grid space is a grid space with a size of 5cm×5cm×5cm. The prediction area is set as a 95×95×95 cubic area with grid space coordinates (5,5,5) and (100,100,100) as diagonal vertices. The detection point is one block for every 8 blocks in each dimension. A total of [(95 / / 8)+1] 3 =1728 known detection points;

[0090] Prediction point coordinate generation: The prediction point is one block for every 3 blocks in each dimension, a total of [(95 / / 3)+1] 3 =32768 points;

[0091] Measure the radiation value according to the coordinates of the detection point;

[0092] Import the data measured at the detection point and use the Kriging algorithm to calculate the radiation data of the predicted point;

[0093] The radiation field is constructed based on the measured data at the detection point and the radiation data at the prediction point.

[0094] The Kriging algorithm is used to calculate the predicted point radiation data including:

[0095] The radiation data of the predicted point is estimated by weighted summation of the data measured by all detection points in the network space. The formula is as follows:

[0096]

[0097] in is the point (x o ,y o ), i.e., z o = z(x o ,yo ), z i is the measured value of the known point i, and n is the total number of known points used to estimate the unknown point;

[0098] λ i is the weight coefficient, which is the factor that satisfies the point (x o ,y o ) and the true value z o The optimal set of coefficients with the smallest difference is

[0099]

[0100] Var is the variance;

[0101] Satisfy the conditions for unbiased estimation:

[0102]

[0103] E is expectation;

[0104] The assumption of Kriging interpolation is that the spatial attribute z is uniform, and for any point (x, y) in space, there is the same expectation c and variance σ 2 ;

[0105] That is, for any point (x, y)

[0106] E[z(x,y)]=E[z]=c

[0107] Var[z(x,y)]=σ 2

[0108] The value z(x,y) at any point is composed of the regional average c and the random deviation R(x,y) of the point, that is,

[0109] z(x,y)=E[z(x,y)]+R(x,y)]=c+R(x,y);

[0110] Where R(x,y) represents the deviation at the point (x,y), and its variance is a constant:

[0111] Var[R(x,y)]=σ 2 .

[0112] See also Figure 5 As shown in the figure, it is a comparison chart between the predicted value and the true value. In the case of , the dose rate prediction accuracy can reach 86.19%. Traditional three-dimensional interpolation algorithms, such as the inverse distance weighted algorithm, generally have an accuracy of only about 70% under the same training prediction ratio, which is much lower than the Kriging method. This shows the superiority of the Kriging algorithm in three-dimensional radiation field reconstruction.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for automatic detection of spatial dose of radioactive materials based on machine vision, characterized in that: The method includes: Automatically identify and pick up labels on radioactive items, where the labels are square labels of the same size, different from the background color and different color; Place the radioactive items in the measurement area and measure the radiation data of the radioactive items, and reconstruct the radiation field based on the radiation data; Among them: Automatic identification and gripping of labels on radioactive items include: Use a robot to carry a camera to take images of radioactive items and convert the images from RGB to HSV color space; Find the color comparison table of HSV color space, set the color threshold to extract the required color, build a mask based on the threshold, and find the target color block; Perform closing and opening operations on the target color block in sequence to obtain the object outline; And draw the minimum right rectangular frame in the original image to obtain the circumscribed rectangle of the object outline; Calculate the coordinates of the four vertices, length and width of the minimum rectangular box that encloses the object outline; Based on the length and width, determine whether the robot can grasp the experimental object.

2. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 1, characterized in that: When there is only one radioactive object in the image of the radioactive object taken, the area of ​​the bounding rectangle is calculated and compared in size, and the bounding rectangle with the largest area is taken as the target label area, and the bounding rectangle with the largest area is taken as the bounding rectangle of the object contour.

3. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 1, characterized in that: According to the circumscribed rectangle of the object outline corresponding to the radioactive item returned in real time, the position information is determined and the robotic arm is controlled to drive the manipulator to approach the target radioactive item to be grasped, so that the coordinates and area of ​​the circumscribed rectangle of the object outline finally reach the threshold range and the grasping action is executed.

4. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 3, characterized in that: The determining and obtaining the location information comprises: The horizontal and vertical coordinates of the center point of the circumscribed rectangle of the object outline are used to determine the position deviation of the robot up, down, left, and right, and the area is used to determine the position deviation of the robot near and far. When the coordinates and the rectangular area of ​​the circumscribed rectangle of the object outline are both within the threshold range, the robot arm performs a grasping action.

5. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 3 is characterized in that: When the image of radioactive items contains multiple radioactive items, find multiple circumscribed rectangles, calculate the areas of the circumscribed rectangles and sort them from large to small, find the circumscribed rectangles corresponding to the first N areas, where N is the number of radioactive items, and judge the sizes of the first N areas. If the area of ​​the label attached to the target radioactive item is not the smallest among the three, it is considered that there is no occlusion, and the robot is controlled to directly grab the target radioactive item; If the area of ​​the label attached to the radioactive item is the smallest among the three, it is considered that there is obstruction.

6. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 5, characterized in that: When the robot is grabbing, in order to determine which one is blocking the target radioactive item, it is necessary to calculate the horizontal coordinates of the center points of the N labels respectively, and subtract the horizontal coordinates of the center points of the labels of the two non-target radioactive items from the horizontal coordinates of the center points of the label of the target radioactive item, and take the absolute value. If the absolute value of the difference with the horizontal coordinate of the target radioactive item is small, it is considered that the distance to the target radioactive item is close and it is the radioactive item that blocks the target radioactive item. The robot is controlled to first clamp the radioactive item that blocks the target radioactive item and put it aside. After all subsequent operations on the target radioactive item are completed, the radioactive item that blocks the target radioactive item will be returned to its place.

7. The method for automatic detection of radioactive material space dose based on machine vision according to claim 1, characterized in that: The radioactive items are measured in the measurement area and the radiation data of the radioactive items are measured, and the radiation field is reconstructed based on the radiation data, including: Detection point coordinate generation: Divide the radiation field into 100×100×100 grids, with a total of 10 6 The grid space is a grid space with a size of 5cm×5cm×5cm. The prediction area is set as a 95×95×95 cubic area with grid space coordinates (5,5,5) and (100,100,100) as diagonal vertices. The detection point is one block for every 8 blocks in each dimension. A total of [(95 / / 8)+1] 3 =1728 known detection points; Prediction point coordinate generation: The prediction point is one block for every 3 blocks in each dimension, a total of [(95 / / 3)+1] 3 =32768 points; Measure the radiation value according to the coordinates of the detection point; Import the data measured at the detection point and use the Kriging algorithm to calculate the radiation data of the predicted point; The radiation field is constructed based on the measured data at the detection point and the radiation data at the prediction point.

8. The method for automatic detection of spatial dose of radioactive materials based on machine vision according to claim 1, characterized in that: The Kriging algorithm is used to calculate the predicted point radiation data including: The radiation data of the predicted point is estimated by weighted summation of the data measured by all detection points in the network space. The formula is as follows: in is the point (x o ,y o ), that is, z o = z(x o ,y o ), z i is the measured value of the known point i, and n is the total number of known points used to estimate the unknown point; λ i is the weight coefficient, which is the factor that satisfies the point (x o ,y o ) and the true value z o The optimal set of coefficients with the smallest difference is Var is the variance; Satisfy the conditions for unbiased estimation: E is expectation; The assumption of Kriging interpolation is that the spatial attribute z is uniform, and for any point (x, y) in space, there is the same expectation c and variance σ 2 ; That is, for any point (x, y) E[z(x,y)]=E[z]=c Var[z(x,y)] = σ 2 The value z(x,y) at any point is composed of the regional average c and the random deviation R(x,y) of the point, that is, z(x,y)=E[z(x,y)]+R(x,y)]=c+R(x,y); Where R(x,y) represents the deviation at the point (x,y), and its variance is a constant: Var[R(x,y)]=σ 2 。

Images