Intelligent detection guidance method based on road infrastructure domain knowledge

Abstract

The present invention discloses an intelligent detection guidance method based on knowledge in the field of road infrastructure. The method includes the following steps: first, a large amount of historical disease inspection data and disease knowledge graph data corresponding to road infrastructure can be obtained from the existing basic database. Then, a portion of data is randomly selected from all the acquired historical disease inspection data and disease knowledge graph data as sample data to construct a training set. After that, the water wave model can be iteratively trained using the sample data in the training set until the training results meet the requirements, and then the types of diseases that may exist in each subcomponent of the target facility are determined. Finally, the trained water wave model is used to calculate the probability of occurrence of each disease based on the type of disease. The probability of occurrence of each disease is used to guide staff to inspect and maintain the corresponding subcomponents of the target facility.

Description

Technical Field

[0001] The present invention relates to the technical field of road infrastructure detection and maintenance, and in particular to an intelligent detection guidance method based on road infrastructure field knowledge. Background Art

[0002] At present, the scale of existing historical inspection data in the field of road infrastructure (bridges, tunnels, slopes, culverts) is huge, and the inspection data is increasing year by year. With the rapid development of modern information technologies such as "Internet +", cloud computing, and big data, and their large-scale integrated application in the field of highway traffic detection, the country has put forward clear requirements for road inspection collection, decision-making, etc.: In December 2019, the Ministry of Transport issued the "Action Outline for Promoting the Development of Comprehensive Transportation Big Data (2020-2025)", which clearly requires the construction of a comprehensive big data analysis technology model, and the research and establishment of a comprehensive and global big data analysis model with strong application value, to form a series of easy-to-understand, conclusive, practical and authoritative big data analysis results to effectively support comprehensive transportation decision-making management and services.

[0003] Currently, scholars at home and abroad have developed a comprehensive understanding of the types, causes, and treatment methods of damage in bridges, tunnels, slopes, culverts, and other areas, which can be used to guide the inspection and maintenance of road infrastructure. However, this research is based on the field experience of industry technicians, making it difficult to adapt to the information development needs of today's rapidly developing big data era. Furthermore, it does not fully utilize existing big data resources to provide intelligent road inspection and maintenance services. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this paper proposes an intelligent inspection guidance method based on knowledge of road infrastructure. This method can assess the probability of occurrence of each type of disease corresponding to each subcomponent of road infrastructure and use this probability to guide personnel in road foundation inspection and maintenance. The specific technical solution is as follows:

[0005] In a first aspect, a detection intelligent guidance method based on road infrastructure domain knowledge is provided, comprising:

[0006] Obtain historical inspection data and disease knowledge graph data corresponding to road infrastructure;

[0007] Sampling the acquired disease historical inspection data and disease knowledge graph data, and constructing a training set based on the sample data obtained;

[0008] Training the water wave model using the constructed training set;

[0009] Obtain the disease types corresponding to each subcomponent of the target facility;

[0010] Based on the trained water wave model, the occurrence probability of each disease corresponding to the target facility is calculated according to the disease types corresponding to each subcomponent of the target facility.

[0011] In conjunction with the first aspect, in a first possible implementation of the first aspect, sampling the acquired historical disease inspection data and disease knowledge graph data includes:

[0012] Preprocess the acquired disease historical inspection data and disease knowledge graph data;

[0013] The diseases that have occurred in subcomponents are used as positive samples, and the diseases that have not occurred in subcomponents are used as negative samples. In the knowledge graph, the diseases that have occurred in subcomponents are used as the head entity, and all subcomponents that have experienced the disease are used as the tail entity. The association between subcomponents and diseases is used as the relationship. Positive and negative sample sampling is performed on the preprocessed disease history inspection data, and triple sampling is performed on the disease knowledge graph data.

[0014] Construct a sample set based on the data samples obtained from positive and negative sample sampling and triple sampling;

[0015] The training set is constructed by randomly extracting sample data from the constructed sample set.

[0016] In combination with the first implementable manner of the first aspect, in the second implementable manner of the first aspect, the preprocessing of the acquired disease historical inspection data and disease knowledge graph data includes:

[0017] Delete the spaces and irrelevant characters at the beginning and end of the disease history inspection data and the disease knowledge graph data;

[0018] Delete duplicate data and missing data in the disease historical inspection data and the disease knowledge graph data.

[0019] In combination with the first aspect, in a third possible implementation of the first aspect, the water wave model is measured using the following loss function:

[0020]

[0021]

[0022] Among them, u and v are subcomponents and diseases respectively, Y is the interaction matrix between subcomponents and diseases, and y uv is the label value of the actual disease of the subcomponent in the interaction matrix, is the sigmoid function, σ(u Tv) is the probability of occurrence calculated by the water wave model, V and E are the embedding matrices of all subcomponents and diseases, R is the embedding matrix of the relationship between subcomponents and diseases, and I r is a slice of the tensor I of the relationship between sub-components and defects in the knowledge graph, λ1 and λ2 are the loss weights of the regularization term and the loss weight of the knowledge graph, respectively.

[0023] In conjunction with the first aspect, in a fourth possible implementation of the first aspect, obtaining the disease type corresponding to each subcomponent of the target facility includes:

[0024] Obtain characteristic data of target facilities;

[0025] Matching the characteristic data of the target facility with the characteristic data of all similar infrastructures in the basic database to determine the similar infrastructure with the highest matching degree;

[0026] According to the types of damage corresponding to the subcomponents of the similar infrastructure with the highest matching degree, the types of damage corresponding to the subcomponents of the target facility are determined.

[0027] In conjunction with the fourth implementation manner of the first aspect, in a fifth implementation manner of the first aspect, determining the similar infrastructure with the highest matching degree includes:

[0028] Calculating the Jaccard similarity coefficient between the target facility and different similar infrastructures based on the corresponding characteristic data;

[0029] The Jaccard similarity coefficients between the target facility and different similar infrastructures are compared, and the similar infrastructure with the largest Jaccard similarity coefficient is the one with the highest matching degree.

[0030] In combination with the first aspect, a sixth possible implementation of the first aspect further includes:

[0031] Regularly obtain disease inspection data of road infrastructure;

[0032] Updating the historical scheduled inspection data according to the acquired disease scheduled inspection data;

[0033] Retrain the water wave model based on the updated historical inspection data and disease knowledge graph data;

[0034] The trained water wave model replaces the original water wave model.

[0035] In combination with the first aspect, in a seventh implementable manner of the first aspect, it further includes: sorting all disease types in descending order according to the occurrence probability to obtain a disease probability sequence.

[0036] In a second aspect, a storage medium is provided, storing a computer program, characterized in that when the computer program is executed, it implements the intelligent detection guidance method based on road infrastructure domain knowledge as described in the first aspect and any one of the first to sixth implementable methods of the first aspect.

[0037] Beneficial effect: By adopting the intelligent detection guidance method based on road infrastructure field knowledge of the present invention, a large amount of historical inspection data and knowledge graph data of road infrastructure can be used to train the water wave algorithm model, and then the trained water wave algorithm model can be used to calculate the probability of occurrence of various diseases that may exist in each sub-component of the target facility. In this way, the existing big data resources can be fully utilized to guide staff to inspect and maintain the road foundation. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the specific embodiments of the present invention, the following briefly introduces the drawings required for use in the specific embodiments. In all the drawings, each element or part is not necessarily drawn according to the actual scale.

[0039] Figure 1 A flowchart of an intelligent detection guidance method based on road infrastructure domain knowledge provided by one embodiment of the present invention;

[0040] Figure 2 A flowchart of constructing a training set provided by one embodiment of the present invention;

[0041] Figure 3 A flowchart for determining the types of diseases that may occur in a target facility according to an embodiment of the present invention;

[0042] Figure 4 A flow chart of updating a water wave model provided by an embodiment of the present invention;

[0043] Figure 5 A calculation flow chart of a water wave model provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following embodiments of the technical solution of the present invention will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and are therefore only examples and are not intended to limit the scope of protection of the present invention.

[0045] It should be understood that in this application, "historical inspection data" refers to historical inspection data for bridges, tunnels, slopes, or culverts. This historical inspection data includes defects that have occurred in these facilities. For example, the hinged joint of a sub-component on the left side of a passage bridge may have a history of defects such as spalling, corner chipping, and water seepage and efflorescence.

[0046] It should also be understood that, in this application, "defect knowledge graph data" refers to the corresponding subcomponents of bridges, tunnels, slopes, or culverts, as well as the historical inspection data for these subcomponents. For example, the right span of a bridge contains subcomponents such as T-beams, diaphragms, wet joints, cap beams, piers, abutment foundations, and drainage holes. This subcomponent's T-beam history includes defects such as concrete crushing and cracks in continuous beam bridges, continuous rigid frame bridges, cantilever beam bridges, and T-shaped rigid frame bridges.

[0047] like Figure 1 The flowchart of the detection intelligent guidance method based on road infrastructure domain knowledge is shown. The evaluation method includes:

[0048] Step 1: Obtain historical inspection data and disease knowledge graph data corresponding to road infrastructure;

[0049] Step 2: Sampling the acquired disease history inspection data and disease knowledge graph data, and constructing a training set based on the sampled data;

[0050] Step 3: training the water wave model using the constructed training set;

[0051] Step 4: Obtain the disease types corresponding to each subcomponent of the target facility;

[0052] Step 5: Based on the trained water wave model, the probability of occurrence of each disease corresponding to the target facility is calculated according to the disease types corresponding to each subcomponent of the target facility.

[0053] Specifically, first, a large amount of historical inspection data and knowledge graph data for road infrastructure defects can be obtained from existing databases. Then, a portion of this data is randomly selected as sample data to construct a training set. The water wave model is then iteratively trained using the sample data in the training set until the training results meet the requirements. The possible types of defects in each subcomponent of the target facility are then determined. Finally, the trained water wave model is used to calculate the probability of each defect. The probability of each defect can be used to guide personnel in inspecting and maintaining the corresponding subcomponents of the target facility.

[0054] The following will be combined Figure 2 The step 2 is described in detail.

[0055] In this embodiment, optionally, sampling the acquired disease history inspection data and disease knowledge graph data includes:

[0056] Step 2-1: Preprocess the acquired disease history inspection data and disease knowledge graph data;

[0057] Step 2-2: Use the diseases that have occurred in the subcomponent as positive samples and the diseases that have not occurred in the subcomponent as negative samples. In the knowledge graph, the disease that has occurred in the subcomponent is the head entity, and all subcomponents that have experienced the disease are the tail entities. The association between the subcomponent and the disease is used as the relationship. Perform positive and negative sample sampling on the preprocessed disease history inspection data and triple sampling on the disease knowledge graph data;

[0058] Step 2-3: construct a sample set based on the data samples obtained from positive and negative sample sampling and triple sampling;

[0059] Step 2-4: randomly extract sample data from the constructed sample set to construct the training set.

[0060] Specifically, before sampling, the acquired data can be preprocessed, including removing leading and trailing spaces and irrelevant characters from the historical disease inspection data and the disease knowledge graph data. Duplicate and missing data in the historical disease inspection data and the disease knowledge graph data can also be removed. For example, data with no value, such as historical inspection data that has features but no subcomponents, or data that has subcomponents but no historical inspection data, can be removed.

[0061] When sampling the preprocessed historical inspection data, positive and negative samples can be sampled using diseases that occurred in subcomponents as positive samples and diseases that did not occur in subcomponents as negative samples. Furthermore, triple sampling can be performed on the preprocessed knowledge graph data, using the disease that occurred in the subcomponent as the head entity, all subcomponents that experienced the disease as the tail entity, and the association between subcomponents and diseases as the relationship. Afterwards, a sample set can be constructed using the sampled historical inspection data and triple data. Samples from this constructed sample set can then be randomly extracted as sample data for the training set. This training set can then be used to train the water wave model.

[0062] In this embodiment, the water wave model is as follows:

[0063]

[0064]

[0065]

[0066] Among them, u represents all subcomponents of the infrastructure, represents the probability of a subcomponent u suffering from a disease v, t represents a subcomponent that has suffered a disease, and t iThe embedding matrix is obtained by decomposing the subcomponents in the historical inspection data into tensors according to random tensor decomposition. v represents the possible types of diseases of all subcomponents, which are the possible types of diseases of all subcomponents of road infrastructure given in the existing road infrastructure inspection specifications. h represents the type of disease that has occurred in the subcomponent, which can be determined through the historical inspection data of road infrastructure. i It represents the embedding matrix obtained by decomposing the relationship between defects and subcomponents in historical inspection data into tensors according to random tensor decomposition.

[0067] During the training process, sample data can be randomly extracted from the training set, and the types of diseases that have occurred in the subcomponents in the extracted sample data are input into the water wave model. The probability of occurrence of the disease type is calculated by the water wave model, and then the loss function is used to measure the calculated results. This process is repeated iteratively until the requirements are met to obtain a trained water wave model. The loss function is as follows:

[0068]

[0069] Among them, Y is the interaction matrix between subcomponents and diseases, y uv is the label value of the actual disease of the subcomponent in the interaction matrix, is the sigmoid function, σ(u T v) is the probability of occurrence calculated by the water wave model, V and E are the embedding matrices of all subcomponents and diseases, R is the embedding matrix of the relationship between subcomponents and diseases, and I r is a slice of the tensor I of the relationship between sub-components and defects in the knowledge graph, λ1 and λ2 are the loss weights of the regularization term and the loss weight of the knowledge graph, respectively.

[0070] Among them, through ∑ (u，v)∈Y -(y uv logσ(u T v)+(1-y uv )log(1-σ(u T v))) can adjust the distance between the predicted data distribution and the real data distribution, by Ability to adjust tensor decomposition parameters and results, It can prevent overfitting. The entire loss function can reduce the distance between the predicted data distribution and the actual data distribution, update the tensor decomposition parameters and results and prevent overfitting, so that the results tend to be more accurate.

[0071] The following will be combined Figure 3 Step 4 is explained in detail.

[0072] In step 4, the disease types corresponding to the subcomponents of the target facility are obtained, including:

[0073] Step 4-1, obtaining characteristic data of the target facility;

[0074] Step 4-2: matching the characteristic data of the target facility with the characteristic data of all similar infrastructures in the basic database to determine the similar infrastructure with the highest matching degree;

[0075] Step 4-3: Determine the type of damage corresponding to each subcomponent of the target facility based on the type of damage corresponding to each subcomponent of the similar infrastructure with the highest matching degree.

[0076] Specifically, since the target facility may be a new facility with no scheduled inspection data in actual application, the process of disease assessment for such target facility may encounter a cold start problem, resulting in the inability to conduct the assessment. To solve this problem, similar road infrastructure can be found as a close substitute. The specific steps are as follows:

[0077] First, the characteristic data of the target facility can be obtained. The characteristic data may include the structural form, structural materials, functional type, climate of the location, design load, operating years and other necessary data to describe the target facility.

[0078] Then, based on the characteristic data of the target facility, the characteristic data of all similar infrastructures in the basic database can be traversed to find the infrastructure with the highest matching degree with the characteristic data of the target facility.

[0079] Finally, the types of diseases that may occur in each sub-component of the target facility can be determined based on the found infrastructure inspection data.

[0080] In this embodiment, optionally, determining the infrastructure of the same type with the highest matching degree includes:

[0081] Calculating the Jaccard similarity coefficient between the target facility and different similar infrastructures based on the corresponding characteristic data;

[0082] The Jaccard similarity coefficients between the target facility and different similar infrastructures are compared, and the similar infrastructure with the largest Jaccard similarity coefficient is the one with the highest matching degree.

[0083] Specifically, in this embodiment, the Jaccard similarity coefficient formula can be used to calculate the Jaccard similarity coefficient between the target facility and each infrastructure in the basic database based on the characteristic data of the target facility and the characteristic data of the infrastructure, thereby obtaining a coefficient set. The infrastructure with the greatest similarity is then selected from the obtained coefficient geometry as the infrastructure of the same type with the highest matching degree. The specific calculation formula for the Jaccard similarity coefficient is as follows:

[0084]

[0085] Among them, N(u) represents the characteristic data of the target facility, and N(v) represents the characteristic data of the infrastructure.

[0086] The number of dimensions included in the target facility's feature data directly affects the accuracy of finding similar infrastructure. Therefore, when obtaining the target facility's feature data, it is recommended to obtain feature data from multiple dimensions to fully characterize the target facility and thus find the best alternative infrastructure.

[0087] In this embodiment, optionally, the step of updating the trained water wave model is also included, which will be discussed below in conjunction with Figure 5 The updating steps of the water wave model are explained in detail.

[0088] In this embodiment, the step of updating the water wave model includes:

[0089] Step 5-1: Regularly obtain road infrastructure disease inspection data;

[0090] Step 5-2: updating the historical scheduled inspection data based on the acquired disease scheduled inspection data;

[0091] Step 5-3: Retrain the water wave model based on the updated historical inspection data and disease knowledge graph data;

[0092] Step 5-4: Replace the original water wave model with the trained water wave model.

[0093] Specifically, in the field of road infrastructure inspection, road infrastructure is regularly inspected. Common knowledge in the field, such as road infrastructure inspection and maintenance regulations, indicates that all types of damage that may occur to subcomponents of road infrastructure are clearly defined in the regulations.

[0094] If the latest inspection data for a subcomponent of a road infrastructure contains defects that have not appeared in previous inspection data but are specified in the specifications, the trained water wave model cannot assess the defects using the latest inspection data.

[0095] To this end, we can periodically merge newly acquired scheduled inspection data with historical data, then use this merged data to train the water wave model. This new water wave model is then used to update the existing one. By continuously iteratively learning from new scheduled inspection data and replacing the existing one with the trained new water wave model, the water wave model becomes comprehensive, representative, updatable, and iterative.

[0096] In this embodiment, optionally, it further includes: sorting all disease types in descending order according to the occurrence probability to obtain a disease probability sequence, which can be used to guide staff to inspect and maintain target facilities.

[0097] A storage medium stores a computer program, which, when executed, implements the above-mentioned intelligent detection guidance method based on road infrastructure domain knowledge.

[0098] To verify the technical effect of the above-mentioned evaluation method, the evaluation method of this embodiment and the similar algorithms DKN (sequential training and learning) and CKE (joint training and learning) were used to evaluate the probability of occurrence of a bridge disease. The obtained evaluation results were analyzed. The obtained analysis results are shown in the following table:

[0099] Table 1 Algorithm model indicators (AUC, ACC)

[0100] Algorithm Model AUC ACC Guide algorithm training 1 time 0.78 0.76 Guide the algorithm to train 5 times 0.88 0.84 Guide the algorithm to train 10 times 0.90 0.85 CKE 0.79 0.73 DKN 0.65 0.59

[0101] As can be seen from Table 1, the algorithm evaluation indicators AUC (the number of correct answers to the total number of answers) and ACC (accuracy) values in the evaluation method used in the present invention tend to be stable after the algorithm is trained for 5 times, and both can reach above 0.8. In addition, compared with the similar algorithms DKN (sequential training and learning) and CKE (joint training and learning) algorithms, the guidance algorithm of the present invention performs better.

[0102] In order to further verify the evaluation method of the present invention, five selections were made in the database, and 10 bridges were randomly selected as target facilities each time. The average MAP (mean average precision) value and NDCG (normalized discounted cumulative gain) value of the results obtained by sorting using the evaluation method of the present invention are shown in the following table.

[0103] Table 2 Algorithm model indicators (MAP, NDCG)

[0104]

[0105]

[0106] Table 2 shows that the average MAP and NDCG of the results obtained by the evaluation method of this embodiment can reach about 0.4 and above 0.7, respectively. Compared with similar algorithms such as TRSL (a software crowdsourcing task recommendation algorithm based on ranking learning) and MPRank (a graph-based recommendation algorithm), the evaluation method of the present invention produces superior results.

[0107] To verify the technical effectiveness of the updated model, since new disease inspection data was temporarily unavailable, the existing inspection data was randomly split. Assuming that 3 / 4 of the data set was historical inspection data and the remaining 1 / 4 was new disease inspection data, the water wave model was trained five times. The trained new water wave model was evaluated. The evaluation results were analyzed, as shown in the table below. This shows that the updated water wave model maintains the original AUC and ACC metrics, with slight improvements.

[0108] Algorithm Model AUC ACC Using 3 / 4 data 0.85 0.80 Use all data 0.88 0.84

[0109] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A detection intelligent guidance method based on road infrastructure domain knowledge, characterized in that: include: Obtain historical inspection data and disease knowledge graph data corresponding to road infrastructure; Sampling the acquired disease historical inspection data and disease knowledge graph data, and constructing a training set based on the sample data obtained; Training the water wave model using the constructed training set; Obtain the disease types corresponding to each subcomponent of the target facility; Based on the trained water wave model, the probability of occurrence of each disease corresponding to the target facility is calculated according to the disease types corresponding to each subcomponent of the target facility; The following loss function is used to measure the water wave model: ； in, 、 are subcomponents and defects, is the interaction matrix between subcomponents and diseases, is the label value of the actual disease of the subcomponent in the interaction matrix, is the sigmoid function, is the probability of occurrence calculated by the water wave model, 、 are the embedding matrices of all subcomponents and defects, is the embedding matrix of the relationship between subcomponents and defects, is the tensor of the relationship between sub-components and defects in the knowledge graph slices, 、 are the loss weights of the regularization term and the knowledge graph, respectively.

2. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 1 is characterized in that: The sampling of the acquired disease history inspection data and disease knowledge graph data includes: Preprocess the acquired disease historical inspection data and disease knowledge graph data; The diseases that have occurred in subcomponents are used as positive samples, and the diseases that have not occurred in subcomponents are used as negative samples. In the knowledge graph, the diseases that have occurred in subcomponents are used as the head entity, and all subcomponents that have experienced the disease are used as the tail entity. The association between subcomponents and diseases is used as the relationship. Positive and negative sample sampling is performed on the preprocessed disease history inspection data, and triple sampling is performed on the disease knowledge graph data. Construct a sample set based on the data samples obtained from positive and negative sample sampling and triple sampling; The training set is constructed by randomly extracting sample data from the constructed sample set.

3. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 2 is characterized in that: The preprocessing of the acquired disease history inspection data and disease knowledge graph data includes: Delete the spaces and irrelevant characters at the beginning and end of the disease history inspection data and the disease knowledge graph data; Delete duplicate data and missing data in the disease historical inspection data and the disease knowledge graph data.

4. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 1 is characterized in that: The disease types corresponding to each subcomponent of the target facility are obtained as follows: Obtain characteristic data of target facilities; Matching the characteristic data of the target facility with the characteristic data of all similar infrastructures in the basic database to determine the similar infrastructure with the highest matching degree; According to the types of damage corresponding to the subcomponents of the similar infrastructure with the highest matching degree, the types of damage corresponding to the subcomponents of the target facility are determined.

5. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 4 is characterized in that: The infrastructure of the same type that is determined to have the highest degree of matching includes: Calculating the Jaccard similarity coefficient between the target facility and various similar infrastructures based on the corresponding characteristic data; The Jaccard similarity coefficients between the target facility and different similar infrastructures are compared, and the similar infrastructure with the largest Jaccard similarity coefficient is the one with the highest matching degree.

6. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 1 is characterized in that: Also includes: Regularly obtain disease inspection data of road infrastructure; Updating the historical scheduled inspection data according to the acquired disease scheduled inspection data; Retrain the water wave model based on the updated historical inspection data and disease knowledge graph data; The trained water wave model replaces the original water wave model.

7. The intelligent guidance method for detection based on road infrastructure domain knowledge according to claim 1 is characterized in that: Also includes: A disease probability sequence is obtained by sorting all disease types in descending order according to the occurrence probability.

8. A storage medium storing a computer program, characterized in that: When the computer program is executed, the intelligent detection guidance method based on road infrastructure domain knowledge as described in any one of claims 1 to 7 is implemented.

Images