Evaluation method for spacing and binding rate of steel bar mesh based on node space analysis

Abstract

The present application can recognize the pixel coordinates and the binding state information of the steel bar net nodes based on the machine vision method. The present application proposes a "cross intersection" node space analysis method based on pattern recognition and a growing clustering numbering algorithm for the steel bar net nodes, and establishes a mathematical model for calculating the uniformity of the steel bar net spacing and the binding rate. The present application can identify the unqualified nodes in the steel bar net, calculate the binding rate of the steel bar net, generate the optimal binding scheme for the unqualified nodes, and visually output the quantitative evaluation results of the uniformity of the steel bar net spacing. The present application only uses the node information of the steel bar net, has a simplified algorithm, fast running speed, accurate recognition effect and stable evaluation results. The application of the present application will greatly improve the construction acceptance efficiency of the steel reinforced concrete warehouse, standardize the work flow of the acceptance, and improve the informatization and intelligentization in the field of civil construction engineering.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of machine vision recognition technology and civil construction technology, and particularly relates to a method for comprehensively evaluating the binding quality of a steel mesh through a node space analysis method. BACKGROUND

[0002] Reinforced concrete structures are widely used in the fields of civil engineering, water conservancy, roads, bridges, etc. In a concrete structure, the pouring preparation quality acceptance evaluation is an important link in engineering construction, including steel mesh binding rate calculation and steel mesh binding uniformity evaluation. The traditional manual inspection and acceptance method mainly relies on manual visual inspection, which has the problems of large measurement workload, time-consuming and laborious, large human error, slow efficiency, low precision, and the inspection data is mainly recorded in paper, which is not convenient for subsequent data storage and retrieval, and the informatization level is insufficient.

[0003] With the rapid development of intelligentization and informatization, based on the current industry situation, using computer technology to replace manual work can improve the work precision and work efficiency. Computer vision recognition technology is increasingly applied in the field of engineering technology, and can also realize the quality detection of civil and water conservancy engineering construction. Through computer vision recognition technology to obtain the node space information of the steel mesh, and then use the computational geometry design space relationship analysis and recognition mathematical model to analyze and calculate the node space of the steel mesh, so as to realize the rapid evaluation of the binding quality of the steel mesh.

[0004] However, there is currently a lack of an efficient, fast, and accurate evaluation method that can provide relevant binding schemes. SUMMARY

[0005] In view of the above technical problems, the present application provides a steel mesh spacing and binding rate evaluation method based on node space analysis. The method uses existing machine vision training results to identify the nodes and their binding states in the steel mesh image, uses computational geometry algorithm to analyze the spatial relationship of the steel mesh node identification information, and establishes a mathematical model according to the node space analysis result to realize the comprehensive evaluation of the steel mesh binding rate and spacing uniformity, so as to quickly and efficiently complete the pouring preparation acceptance of reinforced concrete structures in the water conservancy construction industry, and can recommend the optimal binding scheme according to the evaluation result of the binding.

[0006] The method of the present application has three key technical innovations in realizing the comprehensive evaluation of the binding quality of the steel mesh:

[0007] Firstly, a "cross intersection" mode recognition of the spatial relationship around the steel bar mesh node is proposed, which can identify the four nodes adjacent to a certain steel bar mesh node in space, and can identify the nodes located at the edge of the image through mode extraction feature recognition, the "cross intersection" mode recognition makes full use of the characteristics of the steel bar mesh, the algorithm flow is simple, the recognition result is stable, and the recognition efficiency is high;

[0008] Secondly, a steel bar mesh node numbering method based on a growth algorithm is proposed, the growth algorithm is combined with the "cross intersection" mode recognition, the node spatial analysis recognition is realized in the growth process, and node row and column numbering is established, and the overall spatial relationship analysis of the steel bar mesh is completed when the growth process is completed;

[0009] Thirdly, a mathematical mode for evaluating the steel bar mesh binding rate and the steel bar mesh binding uniformity is proposed, according to the node row and column numbering results, the binding state of the nodes under the interval binding can be quickly searched, the unqualified nodes are identified, the steel bar mesh binding rate evaluation is completed, and the optimal binding scheme of the unqualified nodes is generated. v The optimization objective function method is used to restore the plane steel bar mesh image to the orthographic projection state, and the steel bar mesh binding uniformity evaluation is realized.

[0010] The steel bar mesh binding quality comprehensive evaluation method based on node spatial analysis is proposed, the node and the binding state thereof in the steel bar mesh image are identified by using the machine vision recognition result, the pixel coordinates of the node recognition frame and the Boolean value of the node binding state are taken as the input data of the method, and the algorithm is established to complete the binding rate comprehensive evaluation.

[0011] The specific implementation steps of the present application are as follows:

[0012] Step S1: the steel bar mesh nodes and the node binding information in the image are identified according to the machine vision technology, the "cross intersection" mode recognition method is used to analyze the local spatial relationship between the steel bar mesh nodes;

[0013] Step S2: the mode recognition method proposed in step S1 is used to analyze the overall spatial relationship of the steel bar mesh nodes by establishing the execution flow of the growth numbering algorithm, and the row and column numbering of all the steel bar mesh nodes is established;

[0014] Step S3: according to the spatial analysis result of the steel bar mesh nodes and the binding state of the nodes, it is judged that the steel bar mesh binding mode is all binding or interval binding; when the binding mode is interval binding, the node binding qualified condition is searched according to the node spatial relationship, the unqualified nodes are identified, the steel bar mesh node binding rate is calculated, and the optimal binding scheme of the unqualified nodes is generated; when the binding mode is all binding, the binding rate can be directly calculated;

[0015] Step S4: According to the steel bar node space analysis results and the scale change rule in photogrammetry, the deviation coefficient C is established v The objective function optimization mathematical model is used for reducing the length of the steel bar segment of the plane steel bar mesh to the orthographic projection state, analyzing the deviation degree of the length of the steel bar segment in the steel bar mesh, evaluating the binding uniformity of the steel bar mesh, and visually outputting the uniformity evaluation result.

[0016] Further, the step S1 is to realize the steel bar node space analysis, identify whether a steel bar node is located at the image edge, and if the node is not located at the image edge, identify the four nodes around the node which constitute a steel bar segment with the node. The specific implementation steps are as follows:

[0017] A node which is subjected to the surrounding space relationship analysis is defined as a center node, and the eight nodes closest to the center node are selected as candidate nodes. Four nodes are randomly selected from the candidate nodes to form a "cross intersection" combination with the center node, and there are 70 combination modes. Among the 70 combinations, one combination is the expected identification result, and the four surrounding nodes in the combination are the nodes at the "up, down, left and right" four positions adjacent to the center node in space.

[0018] Further, the extraction rule of the "cross intersection" combination mode identification is that the four surrounding nodes in each combination are sorted counterclockwise as No. 1-4 nodes, the No. 1 node and the No. 3 node, and the No. 2 node and the No. 4 node are connected as straight lines, the intersection of the two straight lines is the "cross intersection", and the coordinates of the intersection point are calculated. The distance between the intersection point and the center node is calculated, and the distance is less than 0.22 times the average distance between the four surrounding nodes of the "cross intersection" and the center node, which is the identification pass.

[0019] According to the mode extraction rule, if the node is not located at the image edge, the node will be left with three basic combination types after mode extraction, as shown in FIG. 3, which can enter the next step of screening to obtain the expected "cross intersection" combination; if the node is located at the image edge, the node will be left with less than three combination types after mode extraction, which can be determined as an edge node and does not enter the next step of screening, but the space relationship of the node can be embodied by the adjacent internal nodes. Figure 3

[0020] Further, in the step S1, the method for screening the expected identification result through the intersection angle and the average distance of the "cross intersection" includes the following two methods:

[0021] (1) The three basic combinations are sorted by the intersection angle from large to small, the top two combinations are sorted by the average distance, and the smaller average distance is the expected identification combination. The identification result of this method is defined as the reference "cross intersection".​​

[0022] (2) Set the "cross intersection" angle of the adjacent node of the node to be solved and the node that has completed the spatial relationship analysis as a, extract any "cross intersection" intersection angle as a i , the average distance is B i , let the parameter A i = abs(a-a i ), find the minimum value of A i is A min ; Then sort the remaining "cross intersection" combinations according to the size of B i , from small to large, and correct them one by one (A-A min )<10 If the condition is met, it means that the combination meets the A and B parameter requirements at the same time, and the combination is the expected recognition result. The recognition result of this method is defined as a positive "cross intersection".

[0023] In fact, the arrangement type of the reference "cross intersection" and the positive "cross intersection" is the same, that is, the center node is the center, and the four nodes around it point to the four directions of the horizontal and vertical extension of the reinforcement mesh. The reference "cross intersection" is used to grow other positive "cross intersections".

[0024] Further, in the step S1, the machine recognition technology adopts the method of ZL202011328747.1. The patent proposes a target detection model capable of identifying the state of the steel bar section and the node binding state between the steel bar mesh nodes. The method of the present application identifies the nodes and node binding state in the steel bar mesh image based on the target detection model, and realizes visual intelligent evaluation of the uniformity of the steel bar mesh spacing and the binding rate.

[0025] Further, step S2 is to establish the execution flow of the growth number coordination algorithm, to establish the row and column numbers for all steel bar mesh nodes, and to complete the overall spatial analysis of the steel bar mesh at the end of the growth process, including the following substeps:

[0026] P1: First growth: first, according to the pattern recognition method of step S1, randomly select nodes from the numerous steel bar mesh nodes for local spatial analysis according to the screening method, so as to establish the first reference "cross intersection". The center point of the reference "cross intersection" is taken as the growth reference point, and the row and column numbers of the steel bar mesh nodes are set as (x, y). The node number is (0, 0), and the row and column numbers of the points 1-4 around the growth reference point are (0, -1), (1, 0), (0, 1), and (-1, 0) in turn, wherein the No. 1 node is the node with the smallest pixel y coordinate.

[0027] P2: Second round of growth: the surrounding nodes of the (0, 0) reference point are taken as new reference points, the surrounding nodes are screened out by referring to the "cross intersection" according to the point sequence, the node local space analysis and the row and column numbering are carried out, and the identified positive "cross intersection" is updated as the reference "cross intersection".

[0028] P3: the P2 operation step is repeatedly executed, the growth reference point is updated to enter the next round of growth, and the growth algorithm ends until all nodes are recorded or grown to the image edge, so that the row and column numbering is established for all nodes.

[0029] After the growth algorithm in the above step S2 is executed, there are still a small number of nodes without row and column numbering, which can be determined by using the row and column numbering characteristics of the numbered nodes.

[0030] Further, the determination method of the point sequence of the sub-step P2 of the step S2 is as follows:

[0031] The i-numbered node is subjected to node local space analysis, and the positive "cross intersection k" established by the i-numbered node is compared with the reference "cross intersection" of the growth reference point, so that the point sequence of each node of the "cross intersection k" is consistent with the reference point.

[0032] Further, the determination method of the point sequence of the sub-step P2 of the step S2 is as follows:

[0033] Taking the growth in the direction of the i-numbered node as an example: the i-numbered node is subjected to node local space analysis, and the "cross intersection k" established by the i-numbered node is compared with the reference "cross intersection" of the growth reference point according to the screening method of the step S1, and the point sequence of each node of the "cross intersection k" is also determined by comparison.

[0034] The specific point sequence comparison method is that the four surrounding points on the "cross intersection" are sorted in a counterclockwise direction, and the point sequence of only the first node is determined to determine the point sequences of the four nodes. The angle J0 between the center point of the reference "cross intersection" pointing to the direction of the first node and the positive direction of the pixel coordinate y-axis is calculated, and the angles J1, J2, J3 and J4 between the center point of the "cross intersection k" pointing to the directions of the four points and the positive direction of the pixel coordinate y-axis are calculated, wherein the direction represented by the value closest to J0 is defined as the first node point sequence method. The point sequence is strictly determined to ensure the consistency of the growth direction and the correctness of the node row and column numbering.

[0035] The row and column numbering rules of the four points on the "cross intersection" of the expected result of the i-numbered node space analysis are as follows: taking the row and column numbering of the i-numbered node as a reference, assuming that its value is (x, y), the x value of the first node is unchanged, and the y value is reduced by 1; the x value of the second node is increased by 1, and the y value is unchanged; the x value of the third node is unchanged, and the y value is increased by 1; the x value of the fourth node is reduced by 1, and the y value is unchanged; the row and column numbering of the four nodes is obtained according to the rules.

[0036] The nodes 1-4 perform the i-node steps, and the row and column numbers of all adjacent nodes around the nodes 1-4 can be obtained.

[0037] Further, the step S3 establishes a mathematical model according to the row and column number results of the reinforcement mesh nodes after the step S2 is performed, identifies the unqualified nodes of the reinforcement mesh binding, and calculates the reinforcement mesh binding rate. The specific steps are as follows:

[0038] J1: Calculate the binding rate

[0039] J1.1: According to the identification result based on the machine vision method, it is assumed that the binding mode is full binding, and the reinforcement mesh binding rate is directly calculated. According to the probability analysis, when the binding rate is greater than 75%, it can be judged that the binding mode is full binding, the binding rate is directly calculated, and the unqualified nodes of the binding are identified. Otherwise, it is considered that the binding mode is interval binding.

[0040] J1.2: When the binding mode is interval binding, according to the binding rate rule, the node number (x, y) is set to be inspected. When the condition 1: the node numbered (x, y) is bound, or the condition 2: the nodes numbered (x-1, y), (x+1, y), (x, y+1), and (x, y-1) are all bound, it can be judged that the node numbered (x, y) is bound qualified. Otherwise, the node is recorded as an unqualified node of the binding.

[0041] J2: Generate the optimal binding scheme

[0042] In the case of interval binding, the sum of the row coordinate and the column coordinate (x+y) of the actually bound node (x, y) is calculated, and the number of odd and even numbers is counted. If the number of odd numbers is greater than the number of even numbers, it is marked as class one, otherwise it is marked as class two.

[0043] The optimal binding scheme of class one: the sum of the row coordinate and the column coordinate of the unbound unqualified node is calculated, and the binding scheme is to bind the nodes with odd sum;

[0044] The optimal binding scheme of class two: the sum of the row coordinate and the column coordinate of the unbound unqualified node is calculated, and the binding scheme is to bind the nodes with even sum.

[0045] In interval binding, taking a node as a reference, if the node is correctly bound, the next node to be bound should be the node corresponding to the row coordinate plus 2 or minus 2 or the column coordinate plus 2 or minus 2. Therefore, in the ideal case, the sum of the row and column coordinates of the bound nodes has the same parity. In the binding process, there will be incorrect binding. Only the nodes with dominant parity need to be counted to determine the optimal binding scheme. Comparing the standard binding scheme (sum of row and column coordinates is odd or even) with the unqualified nodes of the binding can determine the binding scheme of the unqualified nodes of the binding, which is the optimal binding scheme.

[0046] According to the visualization output of the unqualified node pixel coordinates, the unqualified node and the optimal binding scheme of the unqualified node are bound.

[0047] Further, the step S4 establishes the variation coefficient C v The objective function optimization mathematical model reduces the length of the steel bar segment in the plane reinforcement mesh to the orthographic projection state, evaluates and visualizes the reinforcement mesh binding rate, and the specific steps are as follows:

[0048] K1: According to the row and column numbering of the reinforcement mesh node obtained in step S2, the row numbering is clustered to obtain the horizontal steel bar segment group; the column numbering is clustered to obtain the vertical steel bar group. For a steel bar after clustering, the node divides the steel bar into segments, which is defined as a steel bar segment. The pixel distance between adjacent nodes on the steel bar is calculated to form a sequence L={l0, l1,…,l n}, n represents the nth segment, one end of the steel bar is defined as a reference point, the distance from the midpoint of each steel bar segment to the reference point is calculated to form a sequence T={t0, t1,…,t n}, the elements in the sequences L and T correspond one by one to form complete steel bar segment information;

[0049] K2: Each steel bar segment on a steel bar is reduced to the same scale size, and the reduced length l Ri can be calculated by the following formula:

[0050]

[0051] In the formula: k l is the scale coefficient, which determines the size of the reduced scale; b is the coordination coefficient; i is the ith segment;

[0052] The initial value of k l and b is determined, and the reduced steel bar segment length sequence L R ={l R0 ,l R1 ,…,l Rn} is calculated according to the following formula: v

[0053]

[0054] l Ri is the average of each reduced length;

[0055] Taking the minimum variation coefficient C v as the optimization target, the parameter b is solved, and finally the steel bar segment is reduced to the same scale size;

[0056] ​K3: According to the reduction result of K2, the average value of the length of each steel bar section is obtained Calculate the length of each steel bar section x i The relative deviation degree of the average value is defined as the steel bar section deviation index According to the deviation index, the color scale corresponding to the deviation degree of each steel bar section is established, and the deviation degree of each steel bar section of the reinforcement mesh is visualized and output.

[0057] According to the steel bar section deviation index definition output evaluation result: calculate the deviation index for all transverse steel bar groups, the deviation index I is greater than 0, indicating that the length of the steel bar section is too large, the deviation index I is less than 0, indicating that the length of the steel bar section is too small, the transverse deviation index is the average value of all deviation indexes greater than 0, and the transverse deviation index is the average value of all deviation indexes less than 0; The transverse deviation index and the transverse deviation index are defined in the same way.

[0058] The beneficial effects of the present application are as follows:

[0059] (1) The present application is based on the steel bar mesh node and the binding state information recognized by the machine vision method, and proposes a "cross intersection" node space analysis method based on pattern recognition and a steel bar mesh node growth clustering numbering algorithm, and establishes a steel bar mesh uniformity evaluation and binding rate calculation mathematical model.

[0060] (2) The present application can identify the unqualified nodes of the steel bar mesh, calculate the binding rate of the steel bar mesh, and generate the optimal binding scheme of the unqualified nodes, and quantitatively evaluate and visually output the binding uniformity of the steel bar mesh.

[0061] (3) The method of the present application only uses the node information of the steel bar mesh, the algorithm is simple, the running speed is fast, the recognition effect is accurate, and the evaluation result is stable.

[0062] (4) The application of the method of the present application will greatly improve the construction acceptance efficiency of the steel reinforced concrete warehouse, standardize the work flow of the acceptance, and improve the informatization and intelligentization of the civil construction engineering field. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 The expected recognition result of the node local space analysis described in step S1 of the present embodiment, wherein the circular frame node is the center node, and the rectangular frame is the four nodes adjacent to the center node in space.

[0064] Figure 2 The 70 kinds of "cross intersection" do not satisfy the mode extraction feature described in step S1 of the present embodiment.

[0065] Figure 3 The three basic combination types remaining after the mode extraction of the 70 kinds of "cross intersection" described in step S1 of the present embodiment, wherein type three is the expected recognition result.

[0066] Figure 4 The rectangular frame represents the reinforcement mesh node identified by the machine vision technology.

[0067] Figure 5 The schematic diagram is determined for the point sequence.

[0068] Figure 6 The spatial analysis result of the reinforcement mesh after the second growth of the embodiment is shown. The blue circle frame is the growth reference point, and the surrounding space of the 1-4th node around the growth reference point has been analyzed. The 1-4th node point sequence of the "cross intersection" formed by the 1st node point sequence is shown in the 1.1 sequence number in the figure.

[0069] Figure 7 The visualization output result of the unqualified node in the reinforcement mesh binding rate analysis of the embodiment is shown in the black frame, the upper 6 black frames.

[0070] Figure 8 The optimal binding scheme of the unqualified node in the reinforcement mesh binding rate analysis of the embodiment is shown in the black frame, the upper 3 black frames.

[0071] Figure 9 The visualization output result of the reinforcement mesh binding uniformity is shown, in which the depth of the warm color represents the degree of overlength of the reinforcement segment, and the depth of the cold color represents the degree of underlength of the reinforcement segment.

[0072] Figure 10 The output (a) unqualified reinforcement mesh node and (b) uniformity evaluation result. DETAILED DESCRIPTION

[0073] The technical solutions of the method of the application will be clearly and completely explained by the following embodiments, and the outstanding features and significant progress of the method of the application will be further illustrated.

[0074] In the following embodiments, the machine vision recognition technology adopts the method disclosed in ZL202011328747.1.

[0075] EMBODIMENT

[0076] Based on the recognition result of the reinforcement mesh image by the machine vision technology, the embodiment provides a reinforcement mesh spacing and binding rate evaluation method based on node spatial analysis. The image selected in the example is a plane reinforcement image taken at a large inclination angle, and the taken image is shown in Figure 4 The specific steps and detailed explanations are as follows:

[0077] The specific implementation steps of the application are as follows:

[0078] Step S1: Based on machine vision technology, the node-to-node binding information of the reinforcement mesh in the image is identified, and a "cross intersection" pattern recognition method is used. For a certain reinforcement mesh node, the local spatial relationship with the surrounding nodes is analyzed. The expected result of the local spatial analysis is shown in FIG. 8. Figure 1

[0079] The above step S1 is to realize the spatial analysis of the reinforcement mesh node, and to identify whether a certain reinforcement mesh node is located at the edge of the image. If the node is not located at the edge of the image, the four nodes around the node that form the reinforcement segment are identified. The specific implementation steps are as follows:

[0080] For a certain node that is subjected to surrounding spatial relationship analysis, the center node is defined as the center node. The minimum square of the side length is selected as the center node. The nodes on the sides of the square are selected as the surrounding nodes (8 nodes) for screening. Four surrounding nodes are selected at random to form a "cross intersection" combination with the center node, and there are kinds of combination modes. Among the 70 combinations, one combination is the expected identification result. The four surrounding nodes in this combination are the nodes at the "up, down, left and right" positions adjacent to the center node in space.

[0081] The extraction rule of the "cross intersection" combination pattern recognition is that the four surrounding nodes in each combination are sorted counterclockwise as nodes 1-4. Nodes 1 and 3, and nodes 2 and 4 are connected as straight lines, respectively. The intersection of the two straight lines is the "cross intersection", and the coordinates of the intersection point are calculated. The distance between the intersection point and the center node is calculated. If the distance is less than 0.22 times the average distance between the four surrounding nodes of the cross intersection and the center node, the identification is passed. The combination that does not satisfy the feature extraction rule is shown in FIG. 9. Figure 2

[0082] According to the pattern extraction rule, if the node is not located at the edge of the image, the node will be left with three basic combination types after pattern extraction, as shown in FIG. 10. The node can enter the next step of screening to obtain the expected "cross intersection" combination. If the node is located at the edge of the image, the node will be left with less than three combination types after pattern extraction. It can be determined that it is an edge node and does not enter the next step of screening, but the spatial relationship of the node can be represented by the adjacent internal nodes. Figure 3

[0083] Further, in the step S1, the expected identification result is screened out by combining the intersection angle and the average distance of the "cross intersection". The method includes the following two kinds:

[0084] (1) The three basic combinations left are sorted by intersection angle from large to small. The top two combinations are sorted by average distance. The smaller average distance is the expected identification combination, and the identification result is defined as the reference "cross intersection".​​​

[0085] (2) Set the "cross intersection" angle of the adjacent node of the node to be solved and the node that has completed the spatial relationship analysis as a, extract any "cross intersection" intersection angle as a i , the average distance is B i , let the parameter A i = abs (a-a i ), find the minimum value of A i , A min ; Then sort the remaining "cross intersection" combination according to the value of B i from small to large, and correct them one by one (A-A min )<10 If the condition is met, it means that the combination meets the A and B parameter requirements at the same time, and the combination is the expected recognition result. The recognition result of this method is defined as the positive "cross intersection".

[0086] Step S2: The pattern recognition method proposed in step S1 analyzes the spatial relationship of the whole steel mesh node through the execution flow of the growth number coordination algorithm, and establishes the row and column numbers for all steel mesh nodes.

[0087] Step S2 includes the following sub-steps:

[0088] P1: First round of growth: First, according to the pattern recognition of step S1, randomly select nodes for local spatial analysis from the numerous steel mesh nodes according to the screening method, and establish the first reference "cross intersection". The center point of the reference "cross intersection" is taken as the growth reference point. Set the row and column numbers of the steel mesh node as (x, y), then the node number is (0, 0), and the row and column numbers of the nodes around the growth reference point are (0, -1), (1, 0), (0, 1), and (-1, 0) in turn, as described in step S1. The node number of the first node is the smallest node in the pixel y coordinate.

[0089] P2: Second round of growth: The nodes around the (0, 0) reference point are taken as new reference points, and the nodes around the reference "cross intersection" are screened out by the reference "cross intersection", and the row and column numbers are numbered and local spatial analysis is performed. Update the positive "cross intersection" as the reference "cross intersection". Taking the growth along the i node direction as an example: the node local spatial analysis of the i node is performed, and the "cross intersection k" established by the i node is compared with the reference "cross intersection" of the growth reference point according to the screening method of step S1. The node sequence of "cross intersection k" is also determined by comparison.

[0090] The specific point sequence comparison method is that the surrounding four points of the "cross intersection" are sorted in a counterclockwise direction, and only the point sequence of the first node needs to be determined to determine the point sequences of the four nodes. The angle J0 between the direction of the center point of the reference "cross intersection" pointing to the first node and the positive direction of the y-axis of the pixel coordinate is calculated, and the angles J1, J2, J3 and J4 between the directions of the center points of the "cross intersection k" respectively pointing to the four points and the positive direction of the y-axis of the pixel coordinate are calculated, wherein the direction represented by the value closest to J0 is defined as the first node point sequence method. The point sequence is strictly determined to ensure the consistency of the growth direction and the correctness of the node row and column number. Figure 5 As shown in the figure, the gray color in the figure is the reference cross intersection, the black color is the recognized positive cross intersection, and the node point sequence is shown in the figure. It can be seen that the point sequences of the nodes corresponding to the two "cross intersections" are the same, and they respectively point to the four directions of the steel bar mesh horizontal and vertical directions. It can be understood that the point sequences of the "up, down, left and right" four directions are determined to ensure the consistency of the growth direction and to be able to grow in four directions to the whole image.

[0091] The row and column number rules of the four points on the "cross intersection" of the expected result of the spatial analysis of the i-th node are as follows: taking the row and column number of the i-th node as the reference, assuming that its value is (x, y), the x value of the first node is unchanged, and the y value is reduced by 1; the x value of the second node is increased by 1, and the y value is unchanged; the x value of the third node is unchanged, and the y value is increased by 1; the x value of the fourth node is reduced by 1, and the y value is unchanged; the row and column numbers of the four nodes are obtained according to this rule.

[0092] The i-th node executes the same steps as the first to fourth nodes, and the row and column numbers of all adjacent nodes around the first to fourth nodes can be obtained. The spatial analysis result after the second round of growth is shown in Figure 6 .

[0093] P3: Repeat the P2 operation steps to update the growth reference point and enter the next round of growth until all nodes are recorded or the growth reaches the image edge, and the growth algorithm ends, realizing the growth throughout the whole image to establish the row and column numbers for all nodes.

[0094] After the growth algorithm in step S2 is executed, there are still a small number of nodes without established row and column numbers. The row and column numbers can be determined by using the row and column number characteristics of the numbered nodes.

[0095] Step S3: According to the spatial analysis result of the steel bar mesh nodes and the binding condition of the nodes, it is judged that the steel bar mesh binding method is all binding or interval binding; when the binding method is interval binding, the node binding qualified condition is searched according to the spatial relationship of the nodes, the unqualified nodes are identified, and the optimal binding scheme of the unqualified nodes is calculated; when the binding method is all binding, the binding rate can be directly calculated.

[0096] Step S3 includes the following sub-steps:

[0097] J1: Calculate the binding rate

[0098] J1.1: By the identification result based on the machine vision method, first assume that the binding method is all binding, directly calculate the reinforcement mesh binding rate, according to the probability analysis, when the binding rate is greater than 75%, it can be judged that the binding method is all binding, directly calculate the binding rate and identify the unqualified binding node; otherwise, it can be considered that the binding method is interval binding.

[0099] J1.2: When the binding method is interval binding, according to the binding rate rule, set the node number of the inspected binding as (x, y), when the condition 1: the node numbered (x, y) is bound, or the condition 2: the nodes numbered (x-1, y), (x+1, y), (x, y+1), (x, y-1) are all bound, it can be judged that the node numbered (x, y) is bound qualified; otherwise, record the node as unqualified binding node.

[0100] J2: Generate the optimal binding scheme

[0101] In the case of interval binding, calculate the sum of the row coordinate and the column coordinate of the actual bound node (x, y), count the number of odd and even numbers, if the number of odd numbers is greater than the number of even numbers, mark it as class one, otherwise mark it as class two.

[0102] The optimal binding scheme of class one: calculate the sum of the row coordinate and the column coordinate of the unqualified binding node, and the binding scheme is to bind the node with odd sum;

[0103] The optimal binding scheme of class two: calculate the sum of the row coordinate and the column coordinate of the unqualified binding node, and the binding scheme is to bind the node with even sum.

[0104] In interval binding, taking a node as a reference, if the node is correctly bound, the next node to be bound should be the node corresponding to the row coordinate plus 2 or minus 2 or the column coordinate plus 2 or minus 2, so that in the ideal case, the sum of the row and column coordinates of the bound node has the same parity. In the binding process, there will be incorrect binding, only the nodes with dominant parity need to be counted to determine the optimal binding scheme. Comparing the standard binding scheme (sum of row and column coordinates is odd or even) with the unqualified binding node can determine the binding scheme of the unqualified binding node, which is the optimal binding scheme. The unqualified binding node is shown in Figure 7 The optimal binding scheme is shown in Figure 8 .

[0105] According to the pixel coordinates of the unqualified binding node, the unqualified binding node and the optimal binding scheme of the unqualified binding node can be visualized and output.

[0106] The binding rate calculation results in this embodiment are as follows: the binding method of the steel mesh nodes is intermittent binding, the binding rate is 95.65%, the number of nodes with unqualified binding is 6, and the nodes with unqualified binding are visualized.

[0107] Step S4: Based on the spatial analysis results of the steel mesh nodes and the scale variation law in photogrammetry, establish a mathematical model for optimizing the objective function of the deviation coefficient Cv to restore the length of the steel bar segments in the planar steel mesh to the orthographic projection state, analyze the degree of deviation of the length of the steel bar segments in the steel mesh, evaluate the uniformity of the steel mesh binding, and visualize the uniformity evaluation results.

[0108] Step S4 includes the following sub-steps:

[0109] K1: Based on the row and column numbering results of step S2, cluster by row number to obtain horizontal rebar segment groups; cluster by column number to obtain vertical rebar groups. For a rebar after clustering, nodes divide the rebar into segments, defined as rebar segments. Calculate the pixel spacing between adjacent nodes on the rebar to form a sequence L = {l0, l1, ..., l...} of the rebar segment pixel lengths. n}, where n represents the nth segment. One endpoint of the reinforcing bar is defined as the reference point. The distance from the midpoint of each reinforcing bar segment to the reference point is calculated, forming a sequence T = {t0, t1, ..., tn}. n The elements in sequence L and T correspond one-to-one, forming complete information about the rebar segment;

[0110] K2: To restore all segments of a steel bar to the same scale, with a restored length l. Ri It can be calculated using the following formula:

[0111]

[0112] In the formula: k l is the scaling factor, which determines the size of the scale after restoration; b is the coordination factor; i is the i-th segment.

[0113] Propose k l Using the initial value of b, calculate the restored rebar segment length sequence L according to the following formula. R ={l R0 ,l R1 ,…,l Rn The coefficient of variation C v :

[0114]

[0115] This is the average of the reduction lengths;

[0116] With coefficient of variation C vThe minimum is taken as the optimization target to solve the parameter b, and finally the reinforcement segments are reduced to the same scale size;

[0117] K3: According to the reduction result of K2, the average value of the lengths of the reinforcement segments is solved The x of each reinforcement segment is calculated i The relative deviation degree of the average value is defined as the reinforcement segment deviation index According to the deviation index, the color scale corresponding to the deviation degree of each reinforcement segment is established; the deviation index of all transverse reinforcement groups and longitudinal reinforcement groups is calculated and the color scale is established, and the deviation degree of each reinforcement segment of the reinforcement mesh is visualized and output.

[0118] According to the reinforcement segment deviation index, the evaluation result is output: the deviation index of all transverse reinforcement groups is calculated, the deviation index I is greater than 0, indicating that the length of the reinforcement segment is large, the deviation index I is less than 0, indicating that the length of the reinforcement segment is small, the transverse large index is the average value of all deviation indexes greater than 0, and the transverse small index is the average value of all deviation indexes less than 0; the longitudinal large index and the longitudinal small index are defined in the same way.

[0119] In this embodiment, the deviation index of all transverse reinforcement groups and longitudinal reinforcement groups is calculated: the transverse large index of the reinforcement mesh binding is 0.173, and the transverse small index is -0.171; the longitudinal large index of the reinforcement mesh binding is 0.139, and the longitudinal small index is -0.126, and the uniformity evaluation result is visualized and output, as shown in Figure 9 、 Figure 10 .

[0120] The above is only a preferred specific embodiment of the present application, but the scope of protection of the present application is not limited thereto, any modification, equivalent replacement and improvement made by any person skilled in the art within the technical range disclosed by the present application shall be included in the protection scope of the present application.

Claims

1. A method for evaluating the spacing and tying rate of a reinforcement mesh based on node space analysis, characterized by, The method comprises the following steps: Step S1: identifying the binding state information of the reinforcing mesh nodes in the image according to the machine vision technology, and analyzing the local spatial relationship between the reinforcing mesh nodes by using a "cross intersection" pattern recognition method; Step S2: based on the pattern recognition method proposed in step S1, analyzing the overall spatial relationship of the reinforcing mesh nodes by executing the flow process of the growth number coordination algorithm, and establishing the row and column numbers for all the reinforcing mesh nodes; Step S3: judging the reinforcing mesh binding mode to be full binding or interval binding according to the spatial analysis result of the reinforcing mesh nodes and the binding state of the nodes; when the binding mode is interval binding, searching the node binding qualified condition according to the node spatial relationship, identifying the unqualified nodes, calculating the reinforcing mesh node binding rate, and generating the optimal binding scheme of the unqualified nodes; when the binding mode is full binding, the binding rate can be directly calculated; The step S3 comprises the following sub-steps: J1: calculating the binding rate J1.1: directly calculating the reinforcing mesh binding rate by assuming the binding mode to be full binding based on the recognition result of the machine vision method; according to the probability analysis, when the binding rate is greater than 75%, it can be judged that the binding mode is full binding, and the binding rate and the unqualified nodes are directly calculated; otherwise, it is considered that the binding mode is interval binding; J1.2: When the binding mode is interval binding, according to the binding rate rule, when the inspection node or its surrounding 4 nodes have been bound, the inspection node is qualified for binding; otherwise, the node is recorded as a node that fails to bind; let the number of the inspection node be ( x , y ), when the condition 1: the nodes numbered ( x , y ) have been bound, or the condition 2: the nodes numbered ( x -1, y ), ( x +1, y ), ( x , y +1), ( x , y -1) have been bound, the node numbered ( x , y ) can be judged to be qualified for binding; otherwise, the node is recorded as a node that fails to bind; J2: generating the optimal binding scheme In the case of intermittent binding, calculate the row and column coordinates of the actual bound nodes. x , y The sum of the row and column coordinates of ) x + y Count the number of odd and even numbers. If the number of odd numbers is greater than the number of even numbers, mark them as category one; otherwise, mark them as category two. The optimal binding scheme of the first type: calculating the sum of the row and column coordinates of the unqualified nodes, and the optimal binding scheme is to bind the nodes with an odd sum; The optimal binding scheme of the second type: calculating the sum of the row and column coordinates of the unqualified nodes, and the optimal binding scheme is to bind the nodes with an even sum; According to the pixel coordinates of the unqualified nodes, the unqualified nodes and the optimal binding scheme of the unqualified nodes are visualized and outputted; Step S4: According to the spatial analysis results of the reinforcement mesh nodes and the scale change rule in photogrammetry, the deviation coefficient is established C v The objective function optimization mathematical model is used to restore the reinforcement segment length of the plane reinforcement mesh to the orthographic projection state, analyze the deviation degree of the reinforcement segment spacing in the reinforcement mesh, evaluate the uniformity of the reinforcement mesh spacing, and visually output the uniformity evaluation results.

2. The method for evaluating the spacing and tying rate of a reinforcing mesh based on node spatial analysis according to claim 1, characterized in that: The step S1 comprises the following sub-steps: A certain node to be analyzed for local spatial relationship is defined as a center node, 8 nodes closest to the center node are selected as candidate nodes, 4 nodes are randomly selected from the candidate nodes to form a "cross intersection" combination with the center node, and the expected recognition result is screened out from all combinations by using the pattern recognition method, that is, the four nodes adjacent to the center node are identified, that is, the four nodes on the "cross intersection".

3. The method for evaluating the spacing and tying ratio of a reinforcing bar mesh based on node space analysis according to claim 2, characterized in that, In the step S1, the pattern recognition method of the expected recognition result is as follows: The surrounding 4 nodes in each combination are sorted counterclockwise as No. 1-4 nodes, No. 1 node and No. 3 node, and No. 2 node and No. 4 node are connected as straight lines, the intersection of the two straight lines is the "cross intersection", and the coordinates of the intersection point are calculated; the distance between the intersection point and the center node is calculated, and the distance is less than 0.22 times the average distance between the 4 surrounding nodes of the "cross intersection" and the center node, which is the identification pass; For the combination that passes the identification, three basic combination types are obtained by excluding the combination containing the edge node, and the expected recognition result is screened out according to the intersection angle and the average distance of the "cross intersection".

4. The method for evaluating the spacing and tying ratio of a reinforcing mesh based on node space analysis according to claim 3, characterized in that, In the step S1, the method for screening out the expected recognition result according to the intersection angle and the average distance of the "cross intersection" comprises the following two methods: (1) The three types of basic combinations are sorted by cross angle from large to small, and the top two combinations are sorted by average distance. The smaller average distance is the expected identification combination, and the identification result of this method is defined as the reference "cross intersection"; (2) Set the "cross intersection" angle of the adjacent nodes of the node to be solved and the nodes that have completed spatial relationship analysis as a , extract the intersection angle of any "cross intersection" as a i , the average distance is B i , let the parameter A i =abs ( a - a i ), find the minimum value of A i as A min ; then sort the remaining "cross intersection" combinations according to the value of B i from small to large, and correct them one by one in this order ( A - A min )<10; if the condition is met, it means that the combination meets the A , B parameter requirements, and the combination is the expected recognition result, defining the recognition result of this method as a positive "cross intersection".

5. The method for evaluating the spacing and tying rate of the reinforcement mesh based on the node space analysis according to claim 4, characterized in that: The step S2 includes the following sub-steps: P1: First round of growth: Based on the screening method (1), establish a reference "cross" and use its center point as the reference point for growth. Set the row and column number of any rebar mesh node as ( x,y If the reference point node is (0,0), then the row and column numbers of the four nodes surrounding the reference point are (0,-1), (1,0), (0,1), and (-1,0), respectively, where node 1 is a pixel. y The node with the smallest coordinates; P2: Second round of growth: the surrounding nodes of the (0, 0) reference point are taken as new reference points, the surrounding nodes are screened according to the screening method (2) in order of point sequence and by referring to the "cross intersection", the node local spatial relationship analysis and row and column numbering are performed, and the positive "cross intersection" is updated as the reference "cross intersection"; P3: Repeat the P2 operation step to update the growth reference point to the next round of growth until all nodes are recorded or the growth algorithm ends when the image edge is reached, and the growth throughout the image is realized to establish row and column numbering for all nodes.

6. The method for evaluating the spacing and tying rate of a reinforcing mesh based on node spatial analysis according to claim 5, characterized in that: The determination method of the point sequence of the sub-step P2 of the step S2 is as follows: right i Perform local spatial analysis on node number 1, and... i The "cross" established by node number k "Compare with the reference 'cross' at the same growth benchmark point to ensure the 'cross' is..." k "The order of the four nodes is consistent with that of the reference point." 7. The method for evaluating the spacing and tying rate of a reinforcing mesh based on node spatial analysis according to claim 5, characterized in that: The row and column numbering method of the sub-step P2 of the step S2 is as follows: i The row and column numbering rule of the four nodes on the "cross" of the node space analysis result is as follows: taking the node No. 1 as the reference, assuming that the value of the node No. 1 is (0, 0), the value of the node No. 2 is increased by 1, the value of the node No. 3 is not changed, the value of the node No. 4 is decreased by 1, the value of the node No. 5 is increased by 1, the value of the node No. 6 is not changed, the value of the node No. 7 is not changed, the value of the node No. 8 is increased by 1, the value of the node No. 9 is decreased by 1, and the value of the node No. 10 is not changed. i x,y x y x y x y x y According to the rule, the row and column numbers of the four nodes are obtained.​​​​​​​​​ 8. The method for evaluating the spacing and tying rate of a reinforcing mesh based on node spatial analysis according to claim 1, characterized in that: The step S4 includes the following sub-steps: K1: the row and column numbers of the reinforcement mesh node obtained according to step S2 are clustered according to the row numbers to obtain a horizontal reinforcement segment group; the column numbers are clustered to obtain a longitudinal reinforcement group; for a certain reinforcement after clustering, the node divides the reinforcement into segments, which are defined as reinforcement segments; the pixel distance between adjacent nodes on the reinforcement is calculated to form a sequence L={ l 0 , l 1,…, l n}, n representing the first n segment, one end of the reinforcement is defined as a reference point, the distance from the midpoint of each reinforcement segment to the reference point is calculated to form a sequence T={ t 0 ,t 1,…, t n}The elements in the sequences L and T correspond one by one to form complete reinforcement segment information; K2: the reduced length of each bar segment to the same scale size on one bar l Ri may be calculated from the equation: In the formula: k l is a proportional coefficient, determining the size of the scale after reduction; b is a synergistic coefficient; i is the first i paragraph; drafting k l with b initial value, the length sequence L of the reduced reinforcement segments is calculated according to the following formula R ={ l R0 ,l R1 …, l Rn coefficient of variation of C v : was the mean of each reduced length; with the coefficient of variation C v minimizing as an optimization objective b and finally reducing the reinforcement segments to the same scale size; K3: According to the reduction result of K2, the average value of the length of each reinforcing bar section is calculated , the relative deviation degree of each reinforcing bar section from the average value is calculated, and the deviation index is defined as , the deviation index is used to establish the corresponding color scale for each reinforcing bar section, and the evaluation result is visualized and output.​

Images