Construction progress recognition method and system based on image recognition

Abstract

The invention relates to the technical field of construction management, and discloses a construction progress recognition method and system based on image recognition. The method comprises the steps that real-time image data of a construction scene are obtained, a construction progress feature monitoring database mapped with the real-time image data is established, and database monitoring indexes comprise image definition, feature matching degree and the like; configuring a directional recognition target, including key node recognition, construction stage division and the like, and then determining a target classification identifier; key identification parameters are selected through the main identification target and the auxiliary identification target, directional identification target influence analysis is carried out, parameter directional identification is executed after parameter identification constraints are established, and construction progress identification is completed. When parameter directional recognition is executed, a time efficiency and accuracy evaluation function is also established, and a balance adaptation function is constructed to screen recognition schemes. The system comprises a data acquisition module, a target configuration module and the like and can implement the method. Through multi-dimensional analysis and optimization, efficient and accurate recognition of the construction progress is achieved, and the construction progress management requirement is met.

Description

Technical Field

[0001] The present invention relates to the technical field of construction management, and in particular to a construction progress recognition method and system based on image recognition. Background Art

[0002] In the construction industry, accurate identification and monitoring of construction progress is crucial for overall project management. Traditional methods for identifying construction progress often rely on manual inspections and subjective judgment, which presents numerous drawbacks. Manual inspections are labor-intensive and time-consuming, resulting in low efficiency and difficulty in achieving real-time monitoring of construction progress. Subjective judgments are easily influenced by human factors, resulting in low accuracy and reliability of identification results, making them ineffective in providing precise decision-making for project management.

[0003] While some information technology-based construction progress management methods have been proposed with the advancement of technology, these methods still face numerous challenges in practical application. For example, some methods lack adaptability to construction scenarios and struggle to cope with the complex and ever-changing construction site environment. Construction sites often present various interfering factors, such as varying lighting conditions and the storage of construction equipment and materials, which can affect the accuracy of progress identification.

[0004] Existing construction progress recognition methods also have shortcomings in data processing and analysis. They often fail to fully utilize the large amount of image data generated at the construction site, making it difficult to extract effective construction progress features. Furthermore, these methods are limited in their ability to identify key construction nodes, divide construction phases, and identify progress deviations, failing to meet the demands of modern construction projects for refined construction progress management.

[0005] Existing construction progress recognition systems lack effective management and optimization of recognition parameters. In practice, different recognition parameters have varying impacts on construction progress recognition results. Existing methods are unable to accurately analyze the impact of these parameters, nor can they dynamically adjust and optimize the parameters based on actual conditions, thus hindering the efficiency and accuracy of construction progress recognition.

[0006] Existing construction progress identification methods also have shortcomings in process matching. They have difficulty effectively matching construction progress with specific construction processes, and cannot provide strong support for the optimization and improvement of construction processes, which in turn affects the construction quality and efficiency of the entire construction project. Summary of the Invention

[0007] The purpose of the present invention is to provide a construction progress recognition method and system based on image recognition to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides a construction progress recognition method based on image recognition, the method comprising:

[0009] Acquire real-time image data of the construction scene and establish a construction progress feature monitoring database mapped with the real-time image data;

[0010] Configuring directional identification targets based on the construction progress feature monitoring database, wherein the directional identification targets include key node identification, construction phase division, progress deviation identification, and process matching identification;

[0011] Determining target classification identifiers in the directional recognition target, wherein the target classification identifiers include primary recognition targets, auxiliary recognition targets, and retained targets;

[0012] After selecting key recognition parameters using the primary recognition target and the auxiliary recognition target, performing a directional recognition target impact analysis of the key recognition parameters;

[0013] After establishing parameter identification constraints based on the results of directional target impact analysis, parameter directional identification is performed, and the construction progress identification is completed using the parameter directional identification results.

[0014] Preferably, after selecting key recognition parameters using the primary recognition target and the auxiliary recognition target, performing directional recognition target impact analysis of the key recognition parameters includes:

[0015] Get a set of adjustable parameters for image recognition;

[0016] Performing quantitative mapping of the influence of the adjustable parameter set on the main recognition target and the auxiliary recognition target;

[0017] Constructing a target influence matrix of all recognition targets in the directional recognition target, wherein the target influence matrix represents the mutual relationships between different recognition targets, and the mutual relationships include positive promotion relationships and negative conflict relationships;

[0018] Calculate the sensitivity coefficients of the adjustable parameter set based on the impact degree quantitative mapping and the target impact matrix;

[0019] The directional target impact analysis results are established based on the sensitivity coefficient calculation results.

[0020] Preferably, the execution parameter directional identification includes:

[0021] After establishing the control intervals of the parameters, an initial solution set is created based on the real-time image data;

[0022] After performing the solution fitness evaluation in the initial solution set, the optimization direction and optimization step size are established through parameter identification constraints and fitness evaluation results;

[0023] Iteratively updating the initial solution set using the optimization direction and the optimization step size;

[0024] The parameter orientation identification is completed according to the iterative update results.

[0025] Preferably, the iterative updating of the initial solution set using the optimization direction and the optimization step size includes:

[0026] Establish an iterative trajectory for each solution, and identify the iterative trajectory through the solution fitness value of each iteration;

[0027] Configuring an iterative evaluation interval, performing update status identification of the iterative trajectory in the iterative evaluation interval, and generating evaluation categories, wherein the evaluation categories include an optimal solution evaluation category, an exploration evaluation category, and an inferior solution evaluation category;

[0028] Search self-optimization management is performed with iterative updates based on the evaluation classification.

[0029] Preferably, the search self-optimization management that is iteratively updated according to the evaluation classification includes:

[0030] configuring a local proxy model in the optimal solution evaluation classification, using the local proxy model to predict improvement trends, and generating a first reference optimization direction;

[0031] A penalty optimization recognition layer is configured in the inferior solution evaluation classification, and the penalty optimization recognition layer is used to identify the wrong improvement direction and establish a window improvement taboo;

[0032] The first reference optimization direction and window improvement taboo are used to perform iterative fine-tuning updates on the solutions within the superior solution evaluation category. The first reference optimization direction and window improvement taboo are used to perform mixed exploration iterative updates on the exploration evaluation category. Random factors are configured to perform iterative updates on the solutions within the inferior solution evaluation category.

[0033] Preferably, the execution parameter directional identification further includes:

[0034] Establish a time efficiency evaluation function for parameters;

[0035] Establish a balanced adaptation function based on the time efficiency evaluation function and the accuracy evaluation function;

[0036] Identification scheme screening for parameter-oriented identification is performed based on the balance adaptation function, and the identification scheme screening result is output as the parameter-oriented identification result.

[0037] Preferably, the monitoring indicators of the construction progress feature monitoring database include image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error.

[0038] Preferably, the present invention further includes a construction progress recognition system based on image recognition, which is used to implement the above-mentioned construction progress recognition method based on image recognition, and the system includes:

[0039] A data acquisition module is used to acquire real-time image data of the construction scene and establish a construction progress feature monitoring database mapped with the real-time image data;

[0040] A target configuration module is used to configure directional identification targets according to the construction progress feature monitoring database, wherein the directional identification targets include key node identification, construction phase division, progress deviation identification, and process matching identification;

[0041] A classification identification module is used to determine the target classification identification of the directional identification target, wherein the target classification identification includes the main identification target, the auxiliary identification target and the retention target;

[0042] An impact analysis module is used to perform a directional recognition target impact analysis of the key recognition parameters after selecting the key recognition parameters using the main recognition target and the auxiliary recognition target;

[0043] The directional identification module is used to establish parameter identification constraints based on the results of directional target impact analysis, perform parameter directional identification, and use the parameter directional identification results to complete construction progress identification.

[0044] Preferably, when the target configuration module configures the directional recognition target, it includes performing a multi-dimensional correlation analysis on the image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error in the construction progress feature monitoring database, and determining the specific recognition range and accuracy requirements for key node identification, construction stage division, progress deviation identification, and process matching identification based on the correlation analysis results.

[0045] Preferably, when the impact analysis module performs directional recognition target impact analysis of key recognition parameters, it includes calculating the influence weights of the main recognition target and the auxiliary recognition target for the image resolution parameters, feature extraction algorithm parameters, and threshold setting parameters in the adjustable parameter set, constructing a target impact matrix based on the impact weight calculation results, and completing the sensitivity coefficient calculation.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The image-recognition-based construction progress identification method and system provided by this invention offer numerous advantages. By acquiring real-time image data of the construction scene and establishing a database for monitoring construction progress characteristics, it enables real-time monitoring of construction progress, overcoming the inefficiency of traditional manual inspections. This enables timely acquisition of construction progress information, providing more timely data for project management.

[0048] When configuring directional recognition targets, it covers multiple aspects, including key node identification, construction phase division, progress deviation identification, and process matching identification, making construction progress identification more comprehensive and detailed. By defining target classification identifiers, including primary recognition targets, auxiliary recognition targets, and maintenance targets, it can handle different targets in a targeted manner, improving the accuracy and effectiveness of recognition.

[0049] In the process of directional recognition target impact analysis of key recognition parameters, the adjustable parameter set of image recognition is obtained, the influence degree of the main recognition target and auxiliary recognition target is quantitatively mapped, the target impact matrix is ​​constructed and the sensitivity coefficient is calculated, making the analysis of parameter impact more scientific and accurate, and providing a reliable basis for subsequent parameter directional recognition.

[0050] When performing parameter-directed identification, parameter control intervals are established, an initial solution set is created, solution fitness is evaluated, an optimization direction and step size are established, and the initial solution set is iteratively updated, ensuring the accuracy and efficiency of parameter identification. During the iterative update process, an iteration trajectory is established and identified for each solution, and an iterative evaluation interval is configured to generate evaluation categories. Search and self-optimization management are performed based on the evaluation categories, further improving the efficiency and quality of parameter identification.

[0051] At the same time, a time efficiency evaluation function and an accuracy evaluation function of the parameters are established, and a balanced adaptation function is constructed. Based on this, identification scheme screening is carried out, which can achieve a balance between time efficiency and accuracy and make the parameter-oriented identification results more reasonable.

[0052] The monitoring indicators of the construction progress feature monitoring database include multiple dimensions such as image clarity, feature matching, node recognition rate, stage division consistency, deviation calculation error, etc., which ensure the comprehensiveness and accuracy of the database and provide reliable data support for construction progress identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a diagram showing the working principle of the construction progress recognition method based on image recognition described in the present invention.

[0054] Figure 2 Flowchart performed for parameter-directed identification.

[0055] Figure 3 Flowchart for iterative updates of the initial solution set.

[0056] Figure 4 Flowchart for search self-optimization management.

[0057] Figure 5 This is the main flow chart of the construction progress identification system. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] See also Figure 1-Figure 5 The present invention provides a construction progress recognition method based on image recognition, and the specific implementation steps are as follows:

[0060] Acquire real-time image data of the construction scene and establish a construction progress feature monitoring database mapped to the real-time image data. The monitoring indicators of the construction progress feature monitoring database include image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error.

[0061] Targeted recognition targets are configured based on the construction progress feature monitoring database. These targets include key node identification, construction phase division, progress deviation identification, and process matching identification. During configuration, a multi-dimensional correlation analysis is performed on the database's image clarity, feature matching, node recognition rate, phase division consistency, and deviation calculation error. Based on this correlation analysis, the specific recognition scope and accuracy requirements for each target are determined.

[0062] Determine the target classification identification in the directional recognition target, which includes the main recognition target, auxiliary recognition target and retention target.

[0063] After selecting the key recognition parameters using the main recognition target and the auxiliary recognition target, perform a directional recognition target impact analysis of the key recognition parameters. Specifically, obtain the adjustable parameter set for image recognition, perform a quantitative mapping of the influence of the adjustable parameter set on the main recognition target and the auxiliary recognition target, and construct a target influence matrix for all recognition targets in the directional recognition target. The target influence matrix characterizes the positive promotion relationship and negative conflict relationship between different recognition targets, calculates the sensitivity coefficient of the adjustable parameter set based on the quantitative mapping of the influence degree and the target influence matrix, and establishes the directional target impact analysis result based on the sensitivity coefficient calculation result. In this process, the image resolution parameters, feature extraction algorithm parameters, and threshold setting parameters in the adjustable parameter set need to be calculated for the influence weights of the main recognition target and the auxiliary recognition target respectively. Based on the influence weight calculation result, the target influence matrix is ​​constructed and the sensitivity coefficient calculation is completed.

[0064] After establishing parameter identification constraints based on the results of the directional target impact analysis, parameter directional identification is performed, and the results are used to complete construction progress identification. When performing parameter directional identification, a time efficiency evaluation function for the parameters is first established. Based on the time efficiency evaluation function and the accuracy evaluation function, a balanced adaptation function is established. Identification solutions for parameter directional identification are screened based on the balanced adaptation function, and the results of the screening are output as the parameter directional identification results. Furthermore, after establishing the parameter control range, an initial solution set is created based on real-time image data. After performing a fitness evaluation of the solutions within the initial solution set, the optimization direction and step size are established using the parameter identification constraints and the fitness evaluation results. The initial solution set is iteratively updated using the optimization direction and step size, and the parameter directional identification is completed based on the iterative update results.

[0065] Example 1:

[0066] When performing a targeted recognition target impact analysis of key recognition parameters, a set of adjustable image recognition parameters must be obtained. This set includes various parameters that affect the image recognition process, such as image resolution, whose different values ​​directly affect the clarity and detail of the image; feature extraction algorithm parameters, where different algorithms and their parameter settings affect the extraction of key features in the image, such as the threshold of the edge detection algorithm and the dimension of the feature descriptor; and threshold setting parameters. In image segmentation and target recognition, the threshold value determines the range and accuracy of the recognition results.

[0067] The degree of influence of the adjustable parameter set is quantitatively mapped for the main recognition target and the auxiliary recognition target respectively. The main recognition target is the target that plays a leading role in the current recognition task. For example, when the main recognition target is key node recognition, it is necessary to analyze the influence of the image resolution parameter on key node recognition. Specifically, a higher image resolution may make the details of the node clearer, which is conducive to the accurate recognition of the node, but it will also increase the amount of data and processing time; the choice of feature extraction algorithm parameters will affect whether the node features can be accurately extracted. For example, a certain algorithm may be better for extracting node features of a specific type. For auxiliary recognition targets, such as construction stage division, the influence of each adjustable parameter should also be analyzed. For example, the threshold setting parameter may affect the distinction between different construction states when dividing the stages.

[0068] Construct a target impact matrix for all identified targets within the targeted identification objective. This matrix characterizes the interrelationships between different identification targets, including positive promoting relationships and negative conflicting relationships. For example, improved accuracy in identifying key nodes can provide more accurate baseline data for schedule deviation identification, making the results more reliable. This is a positive promoting relationship. On the other hand, inappropriate division of construction phases can conflict with process matching identification, leading to deviations in the results. This is a negative conflicting relationship. When constructing the matrix, it is necessary to analyze and quantify the relationship between each two identification targets to determine the degree and direction of their influence.

[0069] After completing the quantitative impact mapping and constructing the target impact matrix, the sensitivity coefficients of the adjustable parameter set are calculated. These coefficients measure the sensitivity of each adjustable parameter to different recognition targets. Using specific calculation methods, combined with the results of the quantitative impact mapping and the relationships between the recognition targets in the target impact matrix, the impact of each parameter's changes under different recognition targets on the overall recognition results is analyzed. For example, for the image resolution parameter, its sensitivity coefficient is calculated for different recognition targets, such as key node identification and construction stage division, to determine the impact of adjusting this parameter on each target.

[0070] Based on the sensitivity coefficient calculation results, a directional target impact analysis is established. This analysis integrates the impact of various adjustable parameters on different recognition targets and the interactions between them, providing an important basis for the subsequent establishment of parameter identification constraints and parameter-oriented identification. By analyzing the sensitivity coefficients, it is possible to determine which parameters are critical and require specific attention and adjustment, and which parameters have relatively minor impacts and can be flexibly adjusted within a certain range. Furthermore, combined with the target impact matrix, the mutual influence between different recognition targets can be taken into account. When adjusting parameters, it is possible to avoid adversely affecting other targets due to optimizing one target, thereby achieving comprehensive optimization of multiple recognition targets.

[0071] Example 2:

[0072] When performing parameter-oriented recognition, after establishing the parameter control range, an initial solution set is created based on real-time image data. This parameter control range is determined based on the actual needs of the construction scenario and the technical requirements of image recognition. For example, the value range of the image resolution parameter may be limited by the performance of the camera, and the value of the feature extraction algorithm parameters must consider the computing resource capacity while ensuring recognition effectiveness. The creation of the initial solution set requires comprehensive consideration of multiple possible parameter value combinations. These combinations are representative sample points selected from within the parameter control range to cover as much of the potential effective solution space as possible.

[0073] After the initial solution set is created, the solution fitness evaluation within the initial solution set is performed. Solution fitness evaluation is achieved through a balanced fitness function that combines a time efficiency evaluation function and an accuracy evaluation function. The time efficiency evaluation function is used to measure the time consumed by each parameter combination when performing the image recognition task, such as the time required to process a frame of image, the total time required to complete a construction progress identification, etc. The accuracy evaluation function focuses on the accuracy of the recognition results, including the accuracy of key node identification, the correctness of the construction phase division, and the error size of the progress deviation calculation. When performing the evaluation, each parameter combination needs to be substituted into the image recognition process, the real-time image data is processed, the corresponding recognition results are obtained, and then the solution is scored based on the time efficiency and accuracy indicators.

[0074] The optimization direction and optimization step are established through parameter identification constraints and fitness evaluation results. Parameter identification constraints are established based on the results of directional target impact analysis. They are used to limit the range of parameter values ​​to ensure that parameter adjustments do not lead to serious conflicts between identification targets or violate the requirements of actual construction scenarios. For example, when there is a negative conflict between key node identification and process matching identification, parameter identification constraints will limit the adjustment range of certain parameters to avoid excessive damage to the effect of process matching identification due to optimization of key node identification. The determination of the optimization direction needs to be combined with the fitness evaluation results. For solutions with higher fitness, local optimization may be required in the vicinity to find a better solution; for solutions with lower fitness, adjustments need to be made in the direction of improving fitness. The setting of the optimization step needs to take into account the complexity of the solution space and the limitations of computing resources. Too large a step may result in skipping the optimal solution, while too small a step will increase the number of iterations and computing costs.

[0075] The initial solution set is iteratively updated using the optimization direction and step size. During the iterative update process, an iteration trajectory is established for each solution, and the trajectory is identified by the solution fitness value at each iteration. The iteration trajectory records the parameter values ​​and fitness changes of the solution in each iteration. By analyzing the iteration trajectory, we can understand the evolutionary trend and convergence of the solution. For example, if the fitness of a solution continues to improve over multiple iterations, it indicates that the solution is evolving towards the optimal solution. If the fitness fluctuates or stagnates, it is necessary to consider adjusting the optimization direction or step size.

[0076] During the iterative update process, the optimization direction and step size must be continuously adjusted based on the new solution fitness evaluation results and parameter identification constraints to ensure that the iterative process efficiently converges to the optimal solution or a near-optimal solution. After each iterative update, the new solution set must be evaluated to determine whether it meets termination criteria, such as when the number of iterations reaches a preset value or when the solution fitness change is less than a certain threshold. When these criteria are met, the iterative process ends, and the resulting solution set becomes the result of parameter-directed identification. The entire process must maximize time efficiency while ensuring recognition accuracy. By continuously optimizing parameter combinations, efficient and accurate construction progress identification is achieved. When processing real-time image data, the real-time and continuity of the data must be considered to ensure that the iterative update process can promptly respond to changes in the construction scene and provide reliable support for real-time monitoring and management of construction progress. At the same time, the size and quality of the initial solution set must be carefully considered. An excessively large solution set increases the computational burden, while an excessively small solution set may not cover the optimal solution. Therefore, careful design and planning are essential when creating the initial solution set.

[0077] Example 3:

[0078] When iteratively updating the initial solution set using the optimization direction and step size, it is necessary to configure an iterative evaluation interval. This interval identifies the update status of the iterative trajectory and generates corresponding evaluation categories, including optimal solution evaluation categories, exploration evaluation categories, and inferior solution evaluation categories. The iterative evaluation interval is a time period or range of iterations set based on the progress of the iterative process and computing resources. By analyzing the iterative trajectories of each solution within this interval, the quality of the solution and its evolutionary trend can be more accurately judged.

[0079] For optimal solution evaluation and classification, a local surrogate model needs to be configured and used to predict improvement trends and generate a first-reference optimization direction. The local surrogate model is an approximate model constructed based on the solution space near the optimal solution. It can predict possible improvement directions for the optimal solution by learning the characteristics and fitness values ​​of the optimal solution and its surrounding solutions. For example, when a solution is classified as an optimal solution, the local surrogate model analyzes the solution's parameter changes and fitness improvements during the iteration trajectory, identifying which parameter adjustments contribute most to the fitness improvement. This predicts possible improvement trends and generates a first-reference optimization direction.

[0080] During the evaluation and classification of inferior solutions, a penalty optimization recognition layer is configured to identify incorrect improvement directions and establish a window of improvement taboos. The penalty optimization recognition layer analyzes the parameter adjustment directions of inferior solutions during the iteration process and identifies those that lead to a decrease in the solution's fitness. For example, if a solution's fitness decreases significantly after adjusting a certain parameter in a certain iteration, the penalty optimization recognition layer will record this parameter adjustment direction as an incorrect improvement direction and establish an improvement taboo within a certain window range. This prohibits adjusting the parameter in that direction in subsequent iterations to avoid making similar mistakes again.

[0081] Based on the generated first reference optimization direction and the established window improvement taboos, fine-tune and iteratively update the solutions within the optimal solution evaluation category. The purpose of fine-tuning and iterative updating is to conduct a detailed search near the optimal solution to find a better solution. During the fine-tuning process, strictly follow the first reference optimization direction while avoiding violating the window improvement taboos. For example, the general direction of parameter adjustment is determined based on the first reference optimization direction, and then small step adjustments are made in this direction. After each adjustment, the fitness of the solution is evaluated. If the fitness improves, the adjustment is retained, otherwise it is rolled back to the previous state.

[0082] For solutions classified by exploration and evaluation, hybrid exploration and iterative updates are also performed based on the first reference optimization direction and window improvement taboos. Solutions classified by exploration and evaluation fall between optimal and inferior solutions and possess a certain degree of exploration value. Hybrid exploration and iterative updates combine multiple exploration strategies, considering the possible improvement directions indicated by the first reference optimization direction while avoiding the incorrect directions prohibited by the window improvement taboos. At the same time, a certain degree of randomness is introduced to expand the search scope. For example, random perturbations can be added to the first reference optimization direction to generate multiple candidate solutions. These candidate solutions are then evaluated and the solutions with higher fitness are selected for subsequent iterations.

[0083] For solutions within the inferior solution evaluation category, a random factor is configured to perform iterative updates. Because inferior solutions have low fitness and may be trapped in local optima, the introduction of random factors is necessary to increase solution diversity and attempt to escape local optima. The range and function of the random factor can be adjusted based on actual conditions. For example, certain parameters of the inferior solution can be randomly adjusted, or random step sizes can be added to the optimization direction to generate new solutions. After the new solution is generated, its fitness is evaluated. If its fitness improves, it is included in the new solution set; otherwise, other random adjustments are tried.

[0084] Throughout the iterative and self-optimizing search management process, solutions from each evaluation category must be continuously analyzed and processed. Based on changes in the iterative trajectory and fitness evaluation results, the local surrogate model, the parameters of the penalty optimization recognition layer, and the value of the random factor are dynamically adjusted. For example, when the improvement trend of a superior solution becomes less obvious, the local surrogate model can be rebuilt to more accurately predict the direction of improvement. When a suboptimal solution fails to improve its fitness after repeated random adjustments, the range of the random factor can be increased to increase the randomness of the search.

[0085] At the same time, it is important to carefully consider the appropriate setting of the iterative evaluation interval. A too small interval can lead to inaccurate judgments about the solution's evolutionary trend, while a too large interval can increase computational costs. Furthermore, the window range for the window improvement taboo needs to be adjusted based on the actual conditions of the iterative process. A too small range may not effectively prevent incorrect improvement directions, while a too large range may limit search flexibility.

[0086] When processing real-time image data, the iterative update search and self-optimization management process must be real-time and able to promptly respond to changes in the image data within the construction scene. For example, if the construction scene changes, resulting in changes in the image data characteristics, the solution needs to be re-evaluated and reclassified, and the optimization strategy needs to be adjusted to ensure that the results of the parameter-oriented identification are suitable for the new scenario.

[0087] Example 4:

[0088] When performing parameter-directed recognition, in addition to optimizing parameter combinations through iterative updates, a time-efficiency evaluation function must be established. This function measures the time consumption of different parameter combinations when performing image recognition tasks. Its core principle is to map parameter values ​​to processing time. For example, increasing image resolution may increase the amount of data, which in turn increases the time required for feature extraction and object recognition. The time-efficiency evaluation function must quantify this relationship.

[0089] After establishing the time efficiency evaluation function, it is necessary to combine it with the accuracy evaluation function to construct a balanced adaptation function. The accuracy evaluation function focuses on the accuracy of the recognition results, covering dimensions such as the recognition rate of key nodes, the consistency of construction phase division, and the error in the calculation of progress deviations. The balanced adaptation function must comprehensively consider time efficiency and accuracy to avoid excessive pursuit of one metric at the expense of the other.

[0090] The specific form of the equilibrium adaptation function can be expressed as:

[0091] F=w1·f time +w2·f acc

[0092] Among them, F is the output value of the balance adaptation function, which is used to characterize the comprehensive performance of the parameter combination; ftime is the calculation result of the time efficiency evaluation function, and its value reflects the time consumption level of the parameter combination. Generally, the lower the time consumption, the larger the value; f acc is the calculation result of the accuracy evaluation function, and its value reflects the recognition accuracy of the parameter combination. The higher the accuracy, the larger the value. w1 and w2 are weight coefficients, and they satisfy w1+w2=1. These two coefficients are used to adjust the importance of time efficiency and accuracy in the comprehensive evaluation. They can be set according to the actual needs of construction progress identification. For example, when the construction scene has high real-time requirements, the value of w1 can be appropriately increased. When the recognition accuracy requirements are more stringent, the weight of w2 can be increased.

[0093] When screening identification solutions for parameter-oriented identification based on a balanced fitness function, multiple candidate parameter identification solutions must first be generated. Each solution corresponds to a different set of parameter combinations, which are derived from the parameter control interval. These parameter combinations can be generated through an iterative update process of the initial solution set or by selecting representative samples from the parameter control interval through other methods.

[0094] For each candidate solution, it is necessary to substitute it into the image recognition process and process it using real-time image data. Specifically, the parameter combination is used to perform operations such as key node identification, construction phase division, and progress deviation calculation on the real-time image of the construction scene to obtain the corresponding recognition results. Then, the time consumption of the solution in processing the image data is calculated based on the time efficiency evaluation function to obtain f time At the same time, the recognition results are evaluated according to the accuracy evaluation function to obtain f acc The value of .

[0095] f time 、f acc The pre-set weight coefficients w1 and w2 are substituted into the equilibrium fitness function formula to calculate the F value of each candidate solution. The higher the F value, the better the overall performance of the parameter combination in terms of time efficiency and accuracy.

[0096] After calculating the F-values ​​of all candidate solutions, these values ​​are compared and ranked, and the solution with the highest F-value is selected as the output of the parameter-oriented identification result. If multiple solutions have similar F-values, a secondary screening can be performed based on the specific needs of the construction scenario, such as special requirements for certain identification targets.

[0097] In practice, the construction of a time efficiency evaluation function requires consideration of the image data processing flow, including image acquisition, preprocessing, feature extraction, target recognition, and progress analysis. The time consumption of each step is related to the corresponding parameter values. For example, the choice of feature extraction algorithm parameters affects the speed of feature extraction and, in turn, the overall processing time, which needs to be reflected in the function.

[0098] Constructing an accuracy evaluation function requires clarifying the evaluation metrics and calculation methods for each recognition objective. For example, by comparing the recognition results with the actual key node locations, the ratio of correctly identified nodes to the total number of nodes can be calculated. The progress deviation error can be calculated by taking the difference between the recognized progress data and the actual progress data, yielding the absolute or relative error.

[0099] The weighting factors w1 and w2 need to be set based on the project's characteristics, management requirements, and available computing resources. For example, for projects with tight deadlines, time efficiency may be more important, so a relatively large w1 value may be used. For complex projects with extremely high precision requirements, a higher w2 value may be used.

[0100] During the screening process, parameter identification constraints must also be considered. These constraints are established based on the results of the directional objective impact analysis to ensure that parameter combinations do not conflict with other identification objectives. For example, when there is a negative conflict between key node identification and process matching, parameter identification constraints will restrict the range of values ​​for certain parameters. Candidate solutions must meet these constraints to qualify for screening.

[0101] Example 5:

[0102] When establishing a database for monitoring construction progress characteristics, the first step is to acquire real-time image data of the construction scene. This can be achieved by deploying multiple image acquisition devices at the construction site. For example, high-definition cameras installed in locations such as tower cranes and scaffolding can capture the construction area from different angles and frequencies to ensure that image data covering the entire construction process and its details is captured. This image data must include various elements of the construction process, such as building structure, construction machinery, and material stacking, providing raw data support for subsequent characteristic monitoring.

[0103] After acquiring real-time image data, we begin building a construction progress feature monitoring database mapped to this data. Monitoring metrics include image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error. Taking image clarity as an example, the database records parameters such as resolution and noise level for each image. For example, an image captured in the morning with a resolution of 4096×2160 pixels may have a noise density of 5% detected through image preprocessing algorithms. This data is then recorded in real time in the database for subsequent image quality assessment and analysis.

[0104] Monitoring feature matching involves comparing identified construction features with pre-set standard features in a database. For example, the database contains standard features for a certain type of concrete component, including parameters such as shape, size, and texture. When the system processes a real-time image, it extracts the component's actual feature data, such as a length of 5.2 meters (standard value: 5 meters) and a width of 2.1 meters (standard value: 2 meters). Texture features are statistically analyzed using histograms to obtain a specific numerical distribution. The system then quantitatively compares these actual features with the standard features, calculates the specific numerical value of the feature matching, and stores it in the database.

[0105] Node recognition rate monitoring targets key nodes during the construction process. For example, in a construction project, key nodes include the completion of foundation pouring and the topping out of the main structure. When real-time image data is transmitted to the system, the system analyzes the images and identifies the completion status of these key nodes. For example, if, at a certain construction stage, eight of ten key nodes have been completed on site, and the system correctly identifies seven of them through image recognition, the node recognition rate is 70%. This data is recorded in the database to reflect the system's ability to identify key nodes.

[0106] Monitoring the consistency of stage divisions is used to verify the rationality and coherence of construction stage divisions. For example, a construction project may be divided into stages such as foundation construction, main structure construction, and interior decoration construction. The system determines the current construction stage based on the characteristics of the construction scene in the real-time image data, such as the type of construction machinery on site, the material stacking situation, and the completed building structure. Assume that over a period of time, the system divides 100 consecutive images into stages. Of these, 90 images are correctly divided into the same stage, while the other 10 images show stage deviations. In this case, the stage division consistency is 90%. This indicator data will be included in the database to evaluate the stability of the system's stage division.

[0107] Monitoring deviation calculation errors involves calculating progress deviations. The system calculates the construction progress based on image recognition results and compares it with the planned progress. For example, if a construction task is scheduled to complete the construction of a certain structure within 10 days, with a planned completion rate of 100%, and the system calculates the actual completion rate as 95% through image recognition, then the progress deviation is -5%. At the same time, this deviation is compared with the actual manually calculated progress deviation (assuming the actual manually calculated deviation is -4%), and the calculated deviation error is 1%. This error value is recorded in the database to measure the accuracy of the system's progress deviation calculation.

[0108] During the database construction process, these monitoring indicators need to be monitored and updated in real time. As construction progresses, new real-time image data is continuously acquired. The system continuously analyzes each image, calculates various monitoring indicators, and adds the results to the database. For example, every morning the system automatically batch processes all image data acquired the previous day, calculating the average image clarity, the distribution of feature matching, and the daily trend of node recognition rate. These statistical results are then stored in the database to form time series data.

[0109] The establishment of a construction progress feature monitoring database provides data support for subsequent steps such as directional recognition target configuration and key recognition parameter analysis. For example, when configuring key node identification for directional recognition targets, historical node recognition rate data in the database can be used to analyze which key nodes are more difficult to identify, allowing for targeted adjustments to recognition parameters. When analyzing the impact of key recognition parameters, the database's correlation data on image clarity and feature matching can be used to analyze the impact of image resolution parameters on feature matching.

[0110] In practical applications, the database storage structure needs to be rationally designed to ensure efficient data query and analysis. A relational database can be used to associate image data with various monitoring indicators and establish indexes to accelerate data retrieval. Furthermore, to cope with large-scale image and monitoring indicator data, a distributed storage architecture can be considered to improve database storage capacity and access performance.

[0111] Furthermore, a data quality control mechanism needs to be established to ensure that the data entered into the database is accurate and reliable. For example, image acquisition equipment should be regularly calibrated to ensure the accuracy of parameters such as image resolution. Monitoring indicators calculated by the system should be manually inspected. If a significant deviation between the calculated results and the actual situation is found, the system should be debugged and optimized in a timely manner to ensure that the data in the database truly reflects the actual construction progress.

[0112] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0113] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A construction progress recognition method based on image recognition, characterized in that: The method comprises: Acquire real-time image data of the construction scene and establish a construction progress feature monitoring database mapped with the real-time image data; Configuring directional identification targets based on the construction progress feature monitoring database, wherein the directional identification targets include key node identification, construction phase division, progress deviation identification, and process matching identification; Determining target classification identifiers in the directional recognition targets, wherein the target classification identifiers include primary recognition targets, auxiliary recognition targets, and retained targets; After selecting key recognition parameters using the primary recognition target and the auxiliary recognition target, performing a directional recognition target impact analysis of the key recognition parameters; After establishing parameter identification constraints based on the results of directional target impact analysis, parameter directional identification is performed, and the construction progress identification is completed using the parameter directional identification results.

2. The construction progress recognition method based on image recognition according to claim 1, characterized in that: After selecting key recognition parameters using the primary recognition target and the auxiliary recognition target, performing directional recognition target impact analysis of the key recognition parameters includes: Get a set of adjustable parameters for image recognition; Performing quantitative mapping of the influence of the adjustable parameter set on the main recognition target and the auxiliary recognition target; Constructing a target influence matrix of all recognition targets in the directional recognition target, wherein the target influence matrix represents the mutual relationships between different recognition targets, and the mutual relationships include positive promotion relationships and negative conflict relationships; Calculate the sensitivity coefficients of the adjustable parameter set based on the impact degree quantitative mapping and the target impact matrix; The directional target impact analysis results are established based on the sensitivity coefficient calculation results.

3. The construction progress recognition method based on image recognition according to claim 2, characterized in that: The execution parameter directional identification includes: After establishing the control intervals of the parameters, an initial solution set is created based on the real-time image data; After performing the solution fitness evaluation in the initial solution set, the optimization direction and optimization step size are established through parameter identification constraints and fitness evaluation results; Iteratively updating the initial solution set using the optimization direction and the optimization step size; The parameter orientation identification is completed according to the iterative update results.

4. The construction progress recognition method based on image recognition according to claim 3, characterized in that: The iterative updating of the initial solution set using the optimization direction and the optimization step size includes: Establish an iterative trajectory for each solution, and identify the iterative trajectory through the solution fitness value of each iteration; Configuring an iterative evaluation interval, identifying the update state of the iterative trajectory in the iterative evaluation interval, and generating an evaluation classification, wherein the evaluation classification includes an optimal solution evaluation classification, an exploration evaluation classification, and an inferior solution evaluation classification; Search self-optimization management is performed with iterative updates based on the evaluation classification.

5. The construction progress recognition method based on image recognition according to claim 4, characterized in that: The search self-optimization management that is iteratively updated according to the evaluation classification includes: configuring a local proxy model in the optimal solution evaluation classification, using the local proxy model to predict improvement trends, and generating a first reference optimization direction; A penalty optimization recognition layer is configured in the inferior solution evaluation classification, and the penalty optimization recognition layer is used to identify the wrong improvement direction and establish a window improvement taboo; The first reference optimization direction and window improvement taboo are used to perform iterative fine-tuning updates on the solutions within the superior solution evaluation category. The first reference optimization direction and window improvement taboo are used to perform mixed exploration iterative updates on the exploration evaluation category. Random factors are configured to perform iterative updates on the solutions within the inferior solution evaluation category.

6. The construction progress recognition method based on image recognition according to claim 1, characterized in that: The execution parameter directional identification further includes: Establish a time efficiency evaluation function for parameters; Establish a balanced adaptation function based on the time efficiency evaluation function and the accuracy evaluation function; Identification scheme screening for parameter-oriented identification is performed based on the balance adaptation function, and the identification scheme screening result is output as the parameter-oriented identification result.

7. The construction progress recognition method based on image recognition according to claim 1, characterized in that: The monitoring indicators of the construction progress feature monitoring database include image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error.

8. A construction progress recognition system based on image recognition, characterized in that: A system for implementing the construction progress recognition method based on image recognition according to any one of claims 1 to 7, comprising: A data acquisition module is used to acquire real-time image data of the construction scene and establish a construction progress feature monitoring database mapped with the real-time image data; A target configuration module is used to configure directional identification targets according to the construction progress feature monitoring database, wherein the directional identification targets include key node identification, construction phase division, progress deviation identification, and process matching identification; A classification identification module is used to determine the target classification identification of the directional identification target, wherein the target classification identification includes the main identification target, the auxiliary identification target and the retention target; An impact analysis module is used to perform a directional recognition target impact analysis of the key recognition parameters after selecting the key recognition parameters using the primary recognition target and the auxiliary recognition target; The directional identification module is used to establish parameter identification constraints based on the results of directional target impact analysis, perform parameter directional identification, and use the parameter directional identification results to complete construction progress identification.

9. A construction progress recognition system based on image recognition as claimed in claim 8, characterized in that: When configuring the directional recognition target, the target configuration module performs a multi-dimensional correlation analysis on the image clarity, feature matching, node recognition rate, stage division consistency, and deviation calculation error in the construction progress feature monitoring database, and determines the specific recognition range and accuracy requirements for key node identification, construction stage division, progress deviation identification, and process matching identification based on the correlation analysis results.

10. The construction progress recognition system based on image recognition according to claim 8, characterized in that: When the impact analysis module performs directional recognition target impact analysis of key recognition parameters, it includes calculating the impact weights of the main recognition target and the auxiliary recognition target for the image resolution parameters, feature extraction algorithm parameters, and threshold setting parameters in the adjustable parameter set, constructing a target impact matrix based on the impact weight calculation results, and completing the sensitivity coefficient calculation.

Images