Crack information detection device, method of detecting crack information, and crack information detection program
Provided are a crack information detection device, a method of detecting crack information, and a crack information detection program capable of accumulating a manual editing history with respect to an automatically detected damage to improve detection accuracy of a crack. It is assumed that a delete flag “1” of a damage vector (C 5 - 2 ) is recorded in hierarchical structure information by completion of a delete operation with respect to the damage vector (C 5 - 2 ). In the case, in a case where the number of operations or an operation time required for the completion of the delete operation is equal to or larger than a predetermined threshold value (for example, click ten times or more, drag and drop five times or more, or ten minutes or more from the start to the end of the editing operation), threshold values of an angle α 1 and an angle α 2 are reduced by a predetermined amount.
1. A crack information detection device comprising:
a processor configured to execute instructions that cause the crack information detection device to,
perform image analysis to detect crack information on a crack of a structure in a surface image of the structure;
assign a label for identifying crack information corresponding to the crack based on the crack information;
accept editing of the crack information;
register an editing result of the crack information, an operation history required for editing the crack information, and the label in association with each other in a database; and
update a model for detecting the crack information by specifying an editing history required to be learned based on the operation history of the crack information for each of the labels registered in the database and performing machine learning on a feature of a crack to be detected as a continuous crack or a discontinuous crack from the editing result corresponding to the specified editing history,
wherein the editing result of the crack information includes at least one of a position where a pixel composing crack information corresponding to the same label is deleted, a position where a pixel composing crack information corresponding to the same label is added, or a position where a pixel connecting discontinuous pieces of crack information corresponding to different labels is added.
2. The crack information detection device according to claim 1 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for deleting the pixel composing the crack information or an operation time required for deleting the pixel composing the crack information.
3. The crack information detection device according to claim 1 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for adding the pixel composing the crack information or an operation time required for adding the pixel composing the crack information.
4. The crack information detection device according to claim 1 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for adding the pixel connecting the discontinuous pieces of crack information or an operation time required for adding the pixel connecting the discontinuous pieces of crack information.
5. The crack information detection device according to claim 1 ,
wherein the feature of the crack to be detected as the continuous crack or the discontinuous crack includes vector data including a width, a length, and a direction of the crack.
6. The crack information detection device according to claim 1 ,
wherein the detecting includes detecting the crack information on the crack of the structure by at least one of Kruskal's algorithm, Prim's algorithm, neural network convergence calculation, or percolation method.
7. A method of detecting crack information that is executed by a computer, the method comprising:
executing, by a processor of the computer, instructions that cause the computer to perform,
a step of performing image analysis to detect crack information on a crack of a structure in a surface image of the structure;
a step of assigning a label for identifying crack information corresponding to the crack based on the detected crack information;
a step of accepting editing of the crack information;
a step of registering an editing result of the crack information, an operation history required for editing the crack information, and the label in association with each other in a database; and
a step of updating a model for detecting the crack information by specifying an editing history required to be learned based on the operation history of the crack information for each of the labels registered in the database and performing machine learning on a feature of a crack to be detected as a continuous crack or a discontinuous crack from the editing result corresponding to the specified editing history,
wherein the editing result of the crack information includes at least one of a position where a pixel composing crack information corresponding to the same label is deleted, a position where a pixel composing crack information corresponding to the same label is added, or a position where a pixel connecting discontinuous pieces of crack information corresponding to different labels is added.
8. A non-transitory computer-readable tangible medium having a crack information detection program for causing a computer to execute the method of detecting crack information according to claim 7 .
9. A crack information detection device comprising:
a memory storing instructions that, when executed by a processor of the crack information detection device, cause the crack information detection device to,
perform image analysis to detect crack information on a crack of a structure in a surface image of the structure;
assign a label for identifying crack information corresponding to the crack based on the crack information;
accept editing of the crack information;
register an editing result of the crack information, an operation history required for editing the crack information, and the label in association with each other in a database; and
update a model for detecting the crack information by specifying an editing history required to be learned based on the operation history of the crack information for each of the labels registered in the database and performing machine learning on a feature of a crack to be detected as a continuous crack or a discontinuous crack from the editing result corresponding to the specified editing history,
wherein the editing result of the crack information includes at least one of a position where a pixel composing crack information corresponding to the same label is deleted, a position where a pixel composing crack information corresponding to the same label is added, or a position where a pixel connecting discontinuous pieces of crack information corresponding to different labels is added.
10. The crack information detection device according to claim 9 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for deleting the pixel composing the crack information or an operation time required for deleting the pixel composing the crack information.
11. The crack information detection device according to claim 9 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for adding the pixel composing the crack information or an operation time required for adding the pixel composing the crack information.
12. The crack information detection device according to claim 9 ,
wherein the operation history required for editing the crack information includes at least one of the number of operations required for adding the pixel connecting the discontinuous pieces of crack information or an operation time required for adding the pixel connecting the discontinuous pieces of crack information.
13. The crack information detection device according to claim 9 , wherein the feature of the crack to be detected as the continuous crack or the discontinuous crack includes vector data including a width, a length, and a direction of the crack.
14. The crack information detection device according to claim 9 , wherein the detecting includes detecting the crack information on the crack of the structure by at least one of Kruskal's algorithm, Prim's algorithm, neural network convergence calculation, or percolation method.