A steel bar skeleton binding quality detection method and system based on image recognition

Abstract

This invention belongs to the field of image processing technology, specifically relating to a method and system for detecting the quality of rebar cage binding based on image recognition. The method includes: acquiring images of the rebar cage, performing grayscale conversion and Gaussian filtering for noise reduction; identifying straight segments of the rebar and determining intersection nodes through edge detection and Hough line transform; delineating a local analysis region centered on the nodes, extracting the coordinates and strength of binding feature points; calculating positional consistency based on a weighted sum of strength and proximity, calculating morphological consistency based on the strength distribution differences of sectors at multiple angles, and fusing the two to obtain a binding consistency index; determining the actual binding action center by weighting the coordinates of the binding feature points by their strength, and generating a quality deviation index by combining its normalized positional offset with the intersection node and the consistency index; determining whether the quality of each node is qualified based on a preset threshold, and outputting the results. This invention improves the accuracy of rebar binding quality detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method and system for detecting the quality of rebar cage binding based on image recognition. Background Technology

[0002] During the construction process, the steel reinforcement cage is usually fixed by binding to ensure the stability of the steel reinforcement position during concrete pouring and vibration. The effectiveness of the steel reinforcement binding is directly related to the structural stress performance and construction quality. Therefore, checking the quality of steel reinforcement binding after construction is an important part of project quality control.

[0003] In current engineering practice, the inspection of rebar binding quality is gradually shifting from relying entirely on manual inspection to introducing image acquisition and automatic recognition methods. Some technical solutions identify the presence of binding wires in the images of the rebar skeleton by taking pictures, thus using this as a basis for judging whether the binding is complete. These methods reduce the workload of manual inspection to some extent, but have gradually revealed obvious limitations in actual engineering applications.

[0004] The most prominent problem is that existing image recognition-based detection methods have difficulty accurately distinguishing between the presence of tying and whether the tying is effective.

[0005] In actual construction sites, even if there are binding wires at the intersection of reinforcing bars, the binding position may deviate due to non-standard construction operations, limited procedures, or complex site conditions. Such bindings may still show binding marks in appearance, but their ability to restrain the intersection of reinforcing bars is obviously insufficient. During the subsequent concrete pouring process, the reinforcing bars may still shift, change angles, or loosen locally, posing potential structural quality risks.

[0006] However, most existing image detection methods mainly focus on identifying whether there are textures, brightness, or edge features in the image that are similar to the binding wire. Once the relevant features are detected, the binding is determined to be completed at that location. This judgment logic is essentially a binary judgment of the existence of binding, rather than an analysis of the effectiveness of the binding process. When the binding wire is off-center from the intersection of the reinforcing bars, only forms local contact, or is loosely wrapped, it may still be identified as qualified binding, which leads to the detection results being inconsistent with the actual engineering quality, and makes it impossible to determine whether the binding is effective. Summary of the Invention

[0007] To address the aforementioned technical problem of the inability to effectively distinguish whether the binding is effective, the present invention provides solutions in the following aspects.

[0008] In a first aspect, the present invention provides a method for detecting the quality of rebar cage binding based on image recognition, comprising: acquiring an image of the rebar cage, performing grayscale conversion and Gaussian filtering for noise reduction; identifying rebar straight segments and determining rebar intersection nodes by edge detection and Hough line transform on the processed rebar cage image; delineating a local analysis region centered on the rebar intersection node, and extracting the coordinates and strengths of binding feature points in the local analysis region; calculating positional consistency by weighted summation of the strength of each binding feature point relative to its proximity; calculating morphological consistency by dividing the local analysis region into sectors at multiple angles and analyzing the strength differences of binding feature points in each sector relative to the local analysis region; combining positional consistency with morphological consistency to obtain a binding consistency index; obtaining the actual binding action center by weighted summation of the coordinates of all binding feature points; calculating the positional offset between the actual binding action center and the rebar intersection node, and generating a quality deviation index by combining the binding consistency index; setting a quality deviation threshold, determining whether the quality of each rebar intersection node is qualified based on the generated quality deviation index, and outputting the determination result and quality deviation index.

[0009] This invention constructs a complete technical chain from rebar geometric identification and tying feature extraction to multi-dimensional fusion evaluation, establishing a detection framework driven by spatial concentration, circumferential uniformity, and center alignment accuracy. It not only accurately extracts rebar intersections as evaluation benchmarks but also effectively distinguishes between genuine tight tying and hidden defects such as eccentricity and looseness through strength-weighted proximity and sector distribution analysis. Furthermore, it integrates positional consistency, morphological consistency, and normalized positional offset to generate a unified quality deviation index, using continuous values ​​to support the determination of engineering quality deviation thresholds. This significantly improves detection accuracy and reliability, avoids human error and missed detections by traditional methods, and reduces structural safety risks and re-inspection costs.

[0010] Preferably, in the weighted summation, the weight of each binding feature point is equal to the strength of each binding feature point divided by the sum of the strengths of all binding feature points.

[0011] Preferably, the proximity is equal to , For the first Euclidean distance from each binding feature point to the rebar intersection node; This represents the maximum permissible distance for the local analysis region.

[0012] This invention introduces a proximity-weighted summation mechanism with the strength of the binding feature points as the weight, which makes the binding feature points that are close to the intersection node and significant contribute more to the positional consistency, effectively suppresses edge noise interference, assesses the spatial concentration of binding, and avoids misjudgment caused by isolation.

[0013] Preferably, the formula for calculating the morphological consistency is: In the formula, For morphological consistency; The number of sectors divided at multiple angles; For the first The sum of the strength of all binding feature points within a sector at each angle; This is the arithmetic mean of the strength of all binding feature points within the local analysis area; This indicates taking the absolute value.

[0014] This invention effectively assesses the uniformity of tying distribution in the circumferential direction by dividing the local analysis area into sectors with multiple angles and calculating the absolute deviation of the strength of each sector from the mean. It can sensitively identify incomplete tying forms such as unilateral wrapping and partial binding, and significantly improve the detection capability of functional defects.

[0015] Preferably, the binding consistency index is obtained by multiplying positional consistency by morphological consistency.

[0016] This invention achieves dual constraints by coupling the two dimensions of positional concentration and circumferential uniformity through a product form. The index is high only when the binding is both concentrated and uniform. Defects in either dimension will lead to a significant decrease in the index, thereby avoiding one-sided judgment based on a single indicator and enhancing the overall reliability of the evaluation results.

[0017] Preferably, the position offset is the Euclidean distance between the actual binding action center and the intersection node of the reinforcing bars, and is normalized by dividing by the half-diagonal length of the local analysis region.

[0018] Preferably, the formula for calculating the quality deviation index is: In the formula, The deviation index for tying quality; This is the position offset; For binding consistency index; This represents the risk sensitivity weight, with a value range of [0.2, 0.8].

[0019] This invention constructs a continuously adjustable quality deviation index by fusing normalized position offset and binding consistency deficiency through weighted geometric mean. This index can reflect both positioning deviation risk and structural integrity deficiency, and supports flexible adaptation to different engineering requirements for position accuracy or binding form through risk sensitivity weights.

[0020] Preferably, the local analysis area is a square area of ​​fixed size defined with the rebar intersection node as the center. When the area exceeds the image boundary, it is automatically cropped to the effective range.

[0021] Preferably, the value range of the mass deviation threshold is 0.20 to 0.45, and it is adjusted according to the importance of the component: for critical stress parts, the value range is 0.25 to 0.30, and for general parts, the value range is 0.35 to 0.40.

[0022] In a second aspect, the present invention provides a rebar cage binding quality inspection system based on image recognition, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned rebar cage binding quality inspection method based on image recognition is implemented.

[0023] By adopting the above technical solution, a computer program for detecting the quality of steel reinforcement cage binding based on image recognition is generated and stored in a memory so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and processor for convenient use.

[0024] The beneficial effects of this invention are as follows:

[0025] This invention constructs a unified quality deviation index by integrating positional consistency, morphological consistency, and normalized positional offset, forming a comprehensive evaluation mechanism from spatial concentration and circumferential uniformity to center alignment accuracy. This index can effectively identify hidden defects such as eccentricity and looseness, adapt to different rebar sizes and complex site environments, and support the customization of quality deviation thresholds and alignment with engineering standards with continuous values, achieving accurate judgment of binding quality, significantly reducing misjudgments and omissions, and lowering safety risks and manual re-inspection costs. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating an image recognition-based method for detecting the quality of steel reinforcement cage binding according to the present invention.

[0027] Figure 2 It is a grayscale image of the original image after grayscale conversion and Gaussian filtering for noise reduction;

[0028] Figure 3 It is a full-map gradient magnitude map;

[0029] Figure 4 This is a comparison chart of key indicators for rebar tying quality.

[0030] Figure 5 It is a graph showing the matching results of the detection nodes in spatial coordinates. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] This invention discloses a method for detecting the quality of rebar cage binding based on image recognition, referring to... Figure 1 This includes steps S1-S4:

[0034] S1: Acquire images of the rebar skeleton, perform grayscale conversion and Gaussian filtering for noise reduction, and identify rebar straight segments and rebar intersection nodes through edge detection and Hough linear transform.

[0035] At construction sites, the original images of steel reinforcement cages often contain complex environmental backgrounds, lighting interference, and construction noise. If quality assessment is performed directly, it will lead to problems such as inaccurate positioning of intersection nodes and blurred extraction of binding features, which will seriously affect the accuracy of automated detection. Therefore, the system must first remove irrelevant information through image preprocessing and accurately establish geometric spatial benchmarks to provide a reliable analysis area for subsequent quality analysis.

[0036] To address the aforementioned issues, the acquired color images of the rebar skeleton were first converted to grayscale images using a weighted average method, and then denoised using Gaussian filtering to improve image quality. Subsequently, the Canny edge detection algorithm was used to extract significant edges from the image, and Hough line transform was combined to identify continuous, relatively long rebar line segments. Based on the extracted rebar line segments, their pairwise geometric intersections were calculated as potential rebar crossing nodes. Intersections that were too close in space were merged to avoid duplicate detection. A fixed-size local analysis region was defined centered on each finally determined rebar crossing node. The size of the local analysis region was set based on the typical rebar diameter multiplied by an empirical factor. The typical rebar diameter is the product of the rebar's nominal diameter and the imaging resolution, representing the rebar diameter in the image coordinate system. The empirical multiple is used to ensure that the analysis area can reasonably cover the core functional range of the tying node. If the empirical multiple is set too small, the analysis area may not be able to completely contain the tying intersection or local offset features, affecting the accuracy of the evaluation. If the empirical multiple is set too large, it may introduce adjacent irrelevant structures, dilute key tying information, and reduce local sensitivity. Therefore, the reasonable range of values ​​for the empirical multiple is [2,5]. In this embodiment, it is set to 3. Three times the diameter of the rebar is sufficient to cover a single tying knot and its adjacent area in most high-resolution engineering images, focusing on the core quality features while avoiding redundant background interference. In other embodiments, the implementer can adjust the empirical multiple according to the actual rebar spacing. If the generated local analysis area exceeds the image boundary, it will be automatically cropped to the effective range.

[0037] For example, Figure 2 It is a grayscale image of the original image after grayscale conversion and Gaussian filtering for noise reduction. This effectively suppresses background noise, makes the steel structure clearer, and provides a stable foundation for subsequent processing.

[0038] For example, Figure 3 It is a full-image gradient magnitude map. By evaluating the local contrast intensity of each pixel, it highlights areas such as binding wires as bright spots, and the brightness value is directly used as the basis for the strength of binding feature points in subsequent calculations.

[0039] Simultaneously, within each local analysis region, image binding feature points are extracted to characterize the presence of binding: by performing high-frequency texture detection on the local analysis region, which can be implemented using either gradient magnitude or local binary mode, a set of binding feature points is obtained. Each binding feature point contains the coordinates of its binding feature point and the intensity of the corresponding binding feature point, reflecting its significance as a binding feature.

[0040] The strength of the binding feature points is represented by the gradient magnitude. Specifically, Sobel edge detection is performed on the local analysis region, the gradient magnitude of each pixel is calculated, and the local maxima points that are higher than the preset quality deviation threshold are taken as binding feature points, and their gradient magnitude is the corresponding strength.

[0041] Results storage: Saves the rebar intersection nodes, the corresponding local analysis area, and the coordinates and strengths of all tying feature points extracted within that local analysis area.

[0042] S2: Delineate the local analysis area with the rebar intersection as the center, extract the coordinates and strength of the binding feature points, calculate the positional consistency based on the weighted sum of strength and proximity, calculate the morphological consistency through the strength distribution differences of sectors from multiple angles, and combine the two to obtain the binding consistency index.

[0043] In the inspection of rebar cage binding, simply identifying the rebar intersections is insufficient to determine the effectiveness of the binding. This is because in actual engineering projects, issues such as binding position deviation, loose binding, or background interference often arise, leading to some unqualified bindings being judged as qualified. Therefore, this step rigorously evaluates the binding process effectiveness from two dimensions: first, positional consistency, i.e., whether the binding features are closely distributed near the rebar intersections. This is achieved by calculating the distance from each binding feature point to the rebar intersection and weighting it in conjunction with the strength of the binding feature points; second, morphological consistency, i.e., whether the binding is evenly distributed in all directions around the rebar intersections. This is achieved by dividing the local analysis area into sectors with multiple angles and analyzing the consistency of the binding feature point density in each direction. Finally, combining positional concentration and morphological uniformity forms a binding consistency index, used to comprehensively judge the binding consistency.

[0044] Since ties are typically represented as multiple discrete points in an image, positional consistency is constructed by first calculating the proximity for each local analysis region. This is calculated by subtracting the distance from the ties feature point to the center of the rebar intersection from the half-diagonal length of the local analysis region, and then dividing by the half-diagonal length of the local analysis region. This value reflects the degree to which the ties feature point is close to the center of the rebar intersection; the larger the value, the closer the ties feature point is to the center. Subsequently, the weight of each feature point is calculated based on its strength. The following expression for positional consistency is obtained by weighted summation of the proximity scores:

[0045] ;

[0046] In the formula, Positional consistency is measured by the distance from the binding feature point to the center of the intersection. The larger the value, the more concentrated the binding feature points are at the center of the rebar intersection node, and the better the spatial aggregation. The total number of identified binding feature points; For the first The weights of each bound feature point, and , The strength of the binding feature points, The total strength of the binding feature points is used to normalize the molecules. The larger, the better The greater the contribution; For the first The Euclidean distance from each binding feature point to the center of the rebar intersection node The smaller the value, the closer the point is to the center, and the higher the proximity. The larger; The maximum permissible distance for the local analysis region is equal to half the diagonal length of the local analysis region, used for distance... Normalization; For the first The proximity of each binding feature point ensures that the closer the distance, the higher the proximity, thus ensuring that the positional consistency increases monotonically with the increase of binding concentration.

[0047] To assess whether the binding forms a complete enclosure around the rebar intersection, its angular uniformity needs to be measured. If the binding is concentrated only in a local analysis area, it cannot provide effective constraint. Therefore, morphological consistency is also constructed by dividing the local analysis area into sectors with multiple angles. Since a circumferential angle is 360°, to ensure that the angles of each sector are equal and without overlap or gaps, the number of sectors must be divisible by 360°. For example, dividing into 8 or 12 sectors are reasonable choices. In this embodiment, dividing into 8 sectors is preferred. The strength of the binding feature points in each sector is statistically analyzed, and the sum of their absolute deviations from the average strength is calculated. By subtracting the normalized value of this deviation from 1, the morphological consistency is obtained. The smaller the deviation, the more uniform the distribution, and the closer the morphological consistency is to 1. The expression for morphological consistency is as follows:

[0048] ;

[0049] In the formula, For morphological consistency, it is measured in Within a sector at each angle, the strength of the bound feature points Relative to the average Uniformity, The larger the size, the more complete and uniform the binding shape. The number of sectors divided at multiple angles; For the first The sum of the strength of all binding feature points within a sector at a given angle reflects the density of binding features in that angular direction. For all Total intensity of each sector The arithmetic mean of the values ​​is used as a benchmark reference for uniformity analysis. The total absolute deviation is the largest possible value. A larger total absolute deviation indicates a more uneven distribution. The smaller.

[0050] Finally, positional consistency and morphological consistency are combined to construct the binding consistency index, which is expressed as follows:

[0051] ;

[0052] In the formula, The C-value is the binding consistency index, which characterizes the degree to which the binding conforms to the qualified process in terms of positional concentration and morphological uniformity. The larger the C-value, the closer the binding is to the qualified process. For positional consistency, The larger the size, the more concentrated the binding should be in the center; For morphological consistency, The larger the size, the more complete and uniform the binding shape.

[0053] It should be noted that this index achieves the coupling of the dual characteristics of the binding structure through a product: the index value only increases significantly when the binding position is highly consistent and the shape is highly consistent; if there is a defect in either dimension, the index value will decrease significantly.

[0054] After calculating the positional consistency and morphological consistency, the actual binding effect center is calculated based on all binding feature points within the same local analysis area. Specifically, the actual binding effect center is determined as follows: for all binding feature points within the local analysis area, the horizontal and vertical coordinates are weighted and averaged according to the weight of each binding feature point. The resulting coordinates are the actual binding effect center. The weight of each binding feature point is equal to the strength of each binding feature point divided by the sum of the strengths of all binding feature points.

[0055] S3: The actual center of action of the binding is obtained by using the strength-weighted coordinates, and the quality deviation index is generated by combining the normalized position offset of the intersection with the rebar and the binding consistency index.

[0056] Based on the completion of the consistency characterization of the tying structure, in order to further achieve a refined evaluation of the quality of rebar tying, the focus needs to shift from whether the tying has effective characteristics to how much its actual state differs from the ideal process; a qualified tying should not only be concentrated and uniformly wrapped, but its actual center of action must also fall accurately on the rebar intersection node and form a complete tying structure.

[0057] Therefore, this step uses the binding consistency index and binding feature points from step S2 to calculate the relative offset between the actual binding action center and the rebar intersection node. Combined with its structural consistency level, a continuous quality deviation index is constructed to comprehensively reflect the degree of deviation of the binding from the ideal process in terms of both positioning accuracy and structural integrity, providing a unified basis for subsequent high-precision judgment.

[0058] First, based on the rebar intersection determined in step S1 and the coordinates of the actual binding action center obtained from the strength-weighted calculation based on the binding feature points, the Euclidean distance between them is calculated. This distance reflects the absolute offset of the actual binding action center relative to the ideal rebar intersection. Next, the normalized position offset expression is calculated as follows:

[0059] ;

[0060] In the formula, This is the normalized position offset, representing the degree of position deviation. The larger the value, the more pronounced the positional deviation. This is the Euclidean distance between the rebar intersection and the actual center of action of the tying. The larger the value, the greater the positional deviation. This represents the maximum permissible distance for the local analysis region, used for normalization to eliminate scale effects.

[0061] It should be noted that converting the Euclidean distance between the rebar intersection and the actual center of action of the binding into a normalized position offset enhances the model's adaptability to different rebar sizes.

[0062] The normalized positional offset is organically integrated with the binding consistency index to further achieve a comprehensive assessment of binding quality. The expression for the binding quality deviation index is as follows:

[0063] ;

[0064] In the formula, The tying quality deviation index indicates the degree of deviation in the final tying quality. The larger the value, the more severe the quality deviation. This is the normalized position offset, reflecting the degree of misalignment in the binding position. The larger the size, the higher the location risk; For binding consistency index; The ligation consistency index reflects the degree of inconsistency in the ligation structure. The larger the size, the higher the structural risk. As a risk sensitivity weight, and structural risks Deviation index of final result tying quality The risk sensitivity weight is adjusted according to the project focus due to the influence of the project: if the project has high requirements for the accuracy of the binding position, α is preferably 0.6 to 0.8; if the project has high requirements for the uniformity of the binding shape, α is preferably 0.2 to 0.4; if the project has similar requirements for the accuracy of the binding position and the uniformity of the binding shape, α is preferably 0.4 to 0.6; therefore, the reasonable range of α is [0.2, 0.8]; in this embodiment, it is recommended that α be 0.7. In other embodiments, the implementer can set the risk sensitivity weight according to the project focus; since P and C have been normalized, the theoretical range of the binding quality deviation index Q is [0, 1].

[0065] It should be noted that this weighted geometric mean coupled model jointly assesses location risk and structural risk. If either the normalized location deviation or the degree of deficiency in the binding consistency index is too large, it will lead to… Increase; through Factors and models can be flexibly adapted to engineering needs, for example, by increasing... Prioritize punishment for inaccurate positioning, or reduce it. Prioritize punishment for loosely structured structures.

[0066] S4: Determine whether the quality of each node is qualified by setting a preset threshold, and output the result.

[0067] After obtaining the continuous quality deviation index of the local analysis area of ​​each rebar intersection, it needs to be converted into a clear judgment conclusion that conforms to the engineering acceptance standard to support the practical application of the automated quality inspection system. Since construction specifications usually set clear qualified and unqualified boundaries for the binding process, this step maps the continuous evaluation results to the judgment results by introducing a preset maximum allowable quality deviation threshold, thereby realizing the effective connection between the detection system and the engineering acceptance standard.

[0068] Set a maximum allowable quality deviation threshold for a project. For each detection point, if its calculated value is... If it is qualified, then it is deemed qualified; if If it does not meet the requirements, it will be deemed unqualified.

[0069] The preset maximum mass deviation threshold is used to define the reliability boundary of whether the rebar tying has effective binding capacity. Since the theoretical range of the tying quality deviation index is [0,1], if the maximum mass deviation threshold is set too small, it may lead to misjudging slight deviations or uneven distribution within the normal construction tolerance as unqualified, increasing unnecessary rework costs; if it is set too large, it may miss defects in tying with obvious positional deviations or loose structures, weakening the node constraint performance and bringing structural safety hazards; therefore, the reasonable range of this threshold is [0.25]. [0.40] In this embodiment, the value is set to 0.35. The setting of 0.35 is based on balancing detection sensitivity and actual construction tolerance in engineering. It avoids excessive stringency that may lead to misjudgment, while ensuring effective identification of functional defects, thereby achieving a balance between safety and operability. In other embodiments, the implementer can adjust the threshold within the range according to the importance of the component: a lower value is taken for key stress parts, with a value range of 0.25 to 0.30, and a higher value is taken for general parts, with a value range of 0.35 to 0.40.

[0070] Figure 4 This chart compares key indicators of rebar tying quality, using parallel bar charts to display the degree of quality deviation and tying consistency index level at each rebar intersection. A lower quality deviation index indicates that the tying meets the acceptable process requirements; if the value exceeds the preset maximum quality deviation threshold, it is considered unacceptable, indicating problems such as positional misalignment or loose structure. A higher consistency index indicates that the tying points are more concentrated and evenly distributed around the intersection; a low value reflects incomplete, sparse, or biased tying. This chart visually distinguishes between acceptable and unacceptable nodes and helps determine the specific causes of quality problems, providing a clear basis for project acceptance and construction rectification.

[0071] For example, Figure 5 The system is designed to detect the matching results of nodes in spatial coordinates. It accurately locates the intersection area of ​​the reinforcing bars, calculates the actual center of action of each binding point, and compares it with the ideal center point in real time. The system distinguishes whether the installation position meets the standards by different frame states, and intuitively reflects whether the binding of each node meets the process requirements.

[0072] This invention also discloses an image recognition-based rebar cage binding quality inspection system, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an image recognition-based rebar cage binding quality inspection method according to this invention.

[0073] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for detecting the quality of steel reinforcement skeleton binding based on image recognition, characterized in that, include: Acquire images of the steel reinforcement skeleton, perform grayscale conversion and Gaussian filtering for noise reduction; For the processed rebar skeleton image, the straight segments of the rebars are identified by edge detection and Hough line transform, and the rebar intersection nodes are determined; A local analysis region is defined with the rebar intersection as the center, and the coordinates and strengths of the binding feature points in the local analysis region are extracted. The positional consistency is calculated by weighted summation of the strength of each binding feature point and its proximity. The morphological consistency is calculated by dividing the local analysis region into sectors with multiple angles and analyzing the strength difference of the binding feature points in each sector relative to the local analysis region. The binding consistency index is obtained by combining the positional consistency and morphological consistency. The actual center of action of the binding is obtained by summing the coordinates of all binding feature points by weighted summation of their strengths. Calculate the positional offset between the actual center of action of the binding and the intersection node of the rebar, and generate a quality deviation index by combining the binding consistency index; The system presets a quality deviation threshold, determines whether the quality of each rebar intersection node is qualified based on the generated quality deviation index, and outputs the determination result and the quality deviation index. The expression for positional consistency is as follows: In the formula, For positional consistency; The total number of identified binding feature points; For the first The weights of each bound feature point, and , The strength of the binding feature points, This represents the total strength of the binding feature points; For the first The Euclidean distance from each binding feature point to the center of the rebar intersection node; This represents the maximum permissible distance within the local analysis region. For the first The proximity of each binding feature point; The formula for calculating morphological consistency is: In the formula, For morphological consistency; The number of sectors divided at multiple angles; For the first The sum of the strength of all binding feature points within a sector at each angle; This is the arithmetic mean of the strength of all binding feature points within the local analysis area; Indicates taking the absolute value; The binding consistency index is obtained by multiplying positional consistency by morphological consistency; The formula for calculating the quality deviation index is: In the formula, The deviation index for tying quality; This is the position offset; For binding consistency index; This represents the risk sensitivity weight, with a value range of [0.2, 0.8].

2. The method for detecting the quality of rebar cage binding based on image recognition according to claim 1, characterized in that, The position offset is the Euclidean distance between the actual center of action of the binding and the intersection node of the rebar, and is normalized by dividing by the half diagonal length of the local analysis area.

3. The method for detecting the quality of rebar cage binding based on image recognition according to claim 1, characterized in that, The local analysis area is a square area of ​​fixed size defined by the intersection of the reinforcing bars. When the area exceeds the image boundary, it is automatically cropped to the effective range.

4. The method for detecting the quality of rebar cage binding based on image recognition according to claim 1, characterized in that, The value range of the quality deviation threshold is 0.20 to 0.45, and it is adjusted according to the importance of the component: for critical stress parts, the value range is 0.25 to 0.30, and for general parts, the value range is 0.35 to 0.

40.

5. A rebar cage binding quality inspection system based on image recognition, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for detecting the quality of steel reinforcement cage binding based on image recognition, as described in any one of claims 1-4.

Images