CN-121998431-A - Piping dangerous case investigation method and system based on image recognition

CN121998431ACN 121998431 ACN121998431 ACN 121998431ACN-121998431-A

Abstract

The invention relates to the technical field of image recognition, and particularly discloses a piping dangerous case investigation method and system based on image recognition, wherein the method comprises the steps of acquiring and analyzing dam hidden danger data and historical piping data, determining initial type hidden danger data, a risk distance threshold value and a type potential risk area of each hidden danger type label, determining and optimizing the initial potential risk area of a dam to be investigated; determining isolated piping data, a plurality of investigation type hidden danger data, a plurality of merging type data and a dam potential risk area, constructing a plurality of piping dangerous situation models, collecting dam image monitoring data and dam environment monitoring data, and determining dam dangerous situation data. Can the whole coverage show hidden danger, special risk and isolated piping, avoid the risk to miss, accurate adaptation is single with compound hidden danger scene, promote dangerous situation discernment pertinence, strengthen relevance and the reliability of dangerous situation judgement, promote investigation efficiency and precision, realize piping dangerous situation all-round no dead angle prevention and control.

Inventors

WANG XIAOBO
ZHANG LIANG
ZHAO HENG
FENG ZHI
Yi Dujingzi
HE LINQING
XU FUXING
SHI GANG
LUAN YUESHENG
Wang Zhoue
LI LIANG
ZHANG YANIAN

Assignees

水利部长江勘测技术研究所

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (8)

1. A piping dangerous case investigation method based on image recognition is characterized by comprising the following steps: S1, acquiring and analyzing dam hidden danger data and historical piping data of a dam to be inspected, determining initial type hidden danger data, a risk distance threshold value and type potential risk areas of each hidden danger type label, and determining initial potential risk areas of the dam to be inspected; S2, optimizing an initial potential risk area based on historical piping data of the dam to be inspected, determining isolated piping data, inspection type hidden danger data of a plurality of hidden danger type labels and merging type data of a plurality of merging type labels, and determining the dam potential risk area of the dam to be inspected; S3, constructing a piping dangerous situation model of each hidden danger type label based on the investigation type hidden danger data of each hidden danger type label of the dam to be investigated, the merging type data of each merging type label and the historical piping data, and constructing the piping dangerous situation model of each merging type label; S4, acquiring dam image monitoring data of the dam to be inspected based on the image sensor, and acquiring dam environment monitoring data of the dam to be inspected based on the environment sensor group; S5, conducting dangerous case investigation on the dam to be inspected based on the piping dangerous case model of all hidden danger type labels of the dam to be inspected, the piping dangerous case model of all merging type labels, the dam image monitoring data and the dam environment monitoring data, and determining dam dangerous case data.
2. The piping risk investigation method based on image recognition according to claim 1, wherein obtaining and analyzing dam hidden danger data and historical piping data of a dam to be investigated, determining initial type hidden danger data, risk distance threshold value and type potential risk area of each hidden danger type tag, and determining initial potential risk area of the dam to be investigated, comprises: Acquiring dam hidden danger data and historical piping data of a dam to be inspected, wherein the dam hidden danger data comprises a plurality of dam hidden danger positions, hidden danger type labels and hidden danger degree labels of each dam hidden danger position, the historical piping data comprises historical piping sub-data of multiple piping, the historical piping sub-data comprises piping positions, dam body parts, dam pile numbers, historical piping time, piping type labels, dangerous case grade labels, historical environment vectors and historical image recognition vectors, and the dangerous case grade labels comprise primary, secondary, tertiary and quaternary; Dividing all the potential dyke body hidden danger positions based on hidden danger type labels of all the potential dyke body hidden danger positions in the dyke hidden danger data, and determining initial type hidden danger data of each hidden danger type label, wherein the initial type hidden danger data comprises a plurality of the potential dyke body hidden danger positions and hidden danger degree labels of each potential dyke body hidden danger position; Calculating a risk distance threshold value of each hidden danger type label of the dam to be inspected based on the initial type hidden danger data of each hidden danger type label of the dam to be inspected and piping positions, historic piping time and dangerous case grade labels in all historic piping sub-data in the historic piping data; the method comprises the steps of taking the hidden danger position of each hidden danger type label of a dam to be checked as a circle center in initial type hidden danger data of each hidden danger type label of the dam to be checked as a radius, and determining potential risk subareas of the hidden danger position of each hidden danger body in the initial type hidden danger data of each hidden danger type label of the dam to be checked; Determining the type potential risk area of each hidden danger type label of the dam to be checked based on the potential risk subareas of all hidden danger positions of the dam body in the initial type hidden danger data of each hidden danger type label of the dam to be checked, and determining the initial potential risk area of the dam to be checked based on the type potential risk areas of all hidden danger type labels of the dam to be checked.
3. The piping risk troubleshooting method based on image recognition of claim 2, wherein optimizing the initial potential risk area based on the historical piping data of the dam to be troubleshooted comprises: Judging whether piping positions do not belong to the initial potential risk area or not based on piping positions of all historical piping sub-data in the historical piping data and the initial potential risk area of the dam to be inspected, and determining initial isolated data based on the historical piping sub-data of all piping positions which do not belong to the initial potential risk area; Judging whether the number of the historical piping sub data in the initial isolated data is smaller than a preset duty ratio threshold value of the number of the historical piping sub data in the historical piping data, if yes, determining that the initial isolated data is the isolated piping data, otherwise, optimizing a risk distance threshold value of each hidden danger type label of the dam to be inspected, determining an optimized initial potential risk area based on the risk distance threshold value of all hidden danger type labels after optimization and the hidden danger position of each hidden danger body in the initial type hidden danger data of each hidden danger type label, and determining the isolated piping data.
4. The piping risk troubleshooting method based on image recognition of claim 3, wherein determining isolated piping data, troubleshooting type hidden danger data of a plurality of hidden danger type tags, merging type data of a plurality of merging type tags, and determining a dam potential risk area of a dam to be troubleshooted comprises: Selecting a potential risk subarea with the largest area in the initial potential risk area as a recessive judgment area of the isolated piping data; Judging whether a preset number and more than one piping positions exist in one recessive judging area based on the recessive judging areas and the piping positions of all historical piping sub-data in the isolated piping data, if so, determining that each recessive judging area with the preset number and more than one piping positions is a recessive checking area, and determining that the historical piping sub-data of all piping positions which do not belong to any recessive judging area is the adjusted isolated piping data, and if not, not adjusting the isolated piping data; Performing hidden trouble shooting on each hidden trouble shooting area, if hidden trouble shooting is found, determining the hidden trouble position, hidden trouble type label and hidden trouble degree label of the hidden trouble shooting area, and if hidden trouble shooting is not found, determining the hidden trouble shooting area as a special risk subarea; Judging whether all potential risk subareas in the initial potential risk area and all hidden investigation areas for finding hidden danger are overlapped in area, merging two or more potential risk subareas and hidden investigation areas which are overlapped in each existing area, determining each potential merging subarea, and determining the merging type label of each potential merging subarea based on hidden danger type labels corresponding to the two or more potential risk subareas and the hidden investigation areas which are overlapped in area; determining investigation type hidden danger data of each hidden danger type label of a dam to be investigated based on all potential risk subareas without area overlapping in the initial type hidden danger data of each hidden danger type label and hidden investigation area hidden danger positions without area overlapping of all found hidden dangers, hidden danger type labels and hidden danger degree labels, wherein the investigation type hidden danger data comprises a plurality of hidden danger positions of the hidden danger bodies and hidden danger degree labels of the hidden danger positions of the hidden dangers of each hidden danger body; dividing all the potential merging subareas based on the merging type labels of each potential merging subarea, and determining merging type data of each merging type label, wherein the merging type data comprises a plurality of potential merging subareas, a plurality of dyke hidden danger positions of each potential merging subarea and hidden danger degree labels of each dyke hidden danger position; and determining the dam potential risk area of the dam to be inspected based on all the potential risk subareas without area overlapping, all the hidden inspection areas without area overlapping, all the special risk subareas without hidden danger, and all the potential merging subareas in the initial potential risk area.
5. The piping risk inspection method based on image recognition according to claim 4, wherein constructing a piping risk model of each hidden danger type tag and constructing a piping risk model of each merging type tag based on the inspection type hidden danger data of each hidden danger type tag, the merging type data of each merging type tag, and the history piping data of the dam to be inspected, comprises: determining historical hidden danger piping data of each hidden danger position of the embankment body in the hidden danger data of each hidden danger type label of the embankment to be checked based on potential risk subregions or hidden checking regions of the hidden danger position of each embankment body in the hidden danger data of each hidden danger type label of the embankment to be checked and piping positions of all historical piping sub-data in the historical piping data; Taking the hidden danger position of each embankment body in the troubleshooting type hidden danger data of each hidden danger type label of the embankment to be troubleshooted, the potential risk subarea of the hidden danger position of all embankment bodies or the history environment vector and the history image identification vector of all history piping sub-data in the history hidden danger piping data of the hidden troubleshooting area as the input of the piping dangerous case model of each hidden danger type label, taking the potential risk subarea of all hidden danger type labels in the troubleshooting type hidden danger data of each hidden danger type label of the embankment to be troubleshooted or the piping position of all history piping sub-data in the history hidden danger piping data of the hidden troubleshooting area, the piping type label and the dangerous case grade label as the output of the piping dangerous case model of each hidden danger type label, and constructing the piping dangerous case model of each hidden danger type label; determining historical merged piping data of each potential merged sub-region in the merged type data of each merged type label of the dam to be checked based on the potential risk sub-region or the hidden checking region of all the hidden dyke body hidden danger positions of each potential merged sub-region in the merged type data of each merged type label of the dam to be checked and the piping positions of all the historical piping sub-data in the historical piping data; And taking all potential merging sub-areas in the merging type data of each merging type label of the dam to be checked as the input of a piping dangerous situation model of each hidden danger type label, taking all hidden dyke body hidden danger positions of all potential merging sub-areas in the merging type data of each merging type label of the dam to be checked, the historical environment vectors and the historical image identification vectors of all historical piping sub-data in the historical merging piping data of all potential merging sub-areas as the output of the piping dangerous situation model of each hidden danger type label, and constructing the piping dangerous situation model of each merging type label.
6. The piping risk investigation method based on image recognition according to claim 1, wherein the image sensor-based collection of dam image monitoring data of the dam to be investigated and the environmental sensor-based collection of dam environmental monitoring data of the dam to be investigated comprise: Acquiring area image monitoring data of each sub-area in a dam potential risk area of a dam to be inspected based on an image sensor, wherein the area image monitoring data comprises a plurality of monitoring images, and monitoring positions and shooting angles of each monitoring image, and the sub-areas are potential risk sub-areas, hidden inspection areas for finding hidden danger, special risk sub-areas and potential merging sub-areas respectively; Collecting regional environment monitoring data of each sub-region in a potential risk region of a dam to be inspected based on an environment sensor group; Acquiring isolated image monitoring data of each piping position in isolated piping data of a dam to be inspected based on an image sensor, wherein the isolated image monitoring data comprises a plurality of monitoring images, and monitoring positions and shooting angles of each monitoring image; acquiring isolated environment monitoring data of each piping position in isolated piping data of a dam to be inspected based on an environment sensor group; Determining dam image monitoring data of the dam to be inspected based on area image monitoring data of all subareas in the dam potential risk area of the dam to be inspected and isolated image monitoring data of all piping positions in the isolated piping data; and determining the dam environment monitoring data of the dam to be inspected based on the regional environment monitoring data of all sub-regions in the dam potential risk region of the dam to be inspected and the isolated environment monitoring data of all piping positions in the isolated piping data.
7. The piping risk inspection method based on image recognition according to claim 6, wherein the determining of the dam risk data based on the piping risk models of all hidden danger type labels of the dam to be inspected, the piping risk models of all merging type labels, the dam image monitoring data, the dam environment monitoring data, the dangerous case inspection of the dam to be inspected, comprises: Preprocessing all monitoring images in the area image monitoring data of each sub-area in the dam potential risk area of the dam to be inspected, performing image alignment on all preprocessed monitoring images with the same dam pile number in the area image monitoring data of each sub-area, and determining a monitoring pile number image set of each dam pile number of each sub-area in the dam potential risk area of the dam to be inspected; Extracting features of all monitoring images in a monitoring pile number image set of each dam pile number of each sub-area in a dam potential risk area of the dam to be inspected, and determining pile number image identification vectors of each dam pile number of each sub-area in the dam potential risk area of the dam to be inspected; extracting characteristics of environment monitoring data of each sub-area in a potential risk area of the dam to be inspected, and determining an area environment vector of each sub-area in the potential risk area of the dam to be inspected; If the subarea is a potential risk subarea or a hidden investigation area for finding hidden trouble, the area environment vector of the subarea and the monitoring pile number image set of all dykes and piles are input into a piping dangerous situation model of a hidden trouble type label of the potential risk subarea corresponding to the subarea or the hidden investigation area for finding hidden trouble, and the area dangerous situation data of the subarea is determined based on the piping dangerous situation model, wherein the area dangerous situation data comprises a dangerous situation prediction label, a plurality of predicted piping positions when the dangerous situation prediction label exists, a predicted piping type of each predicted piping position and a predicted dangerous situation grade; If the subarea is a potential merging subarea, the area environment vector of the subarea and the monitoring pile number image set of all dyke pile numbers are input into a piping dangerous situation model of a merging type label of the potential merging subarea corresponding to the subarea, and area dangerous situation data of the subarea is determined based on the piping dangerous situation model; If the subarea is a special risk subarea, carrying out artificial dangerous case identification based on the area environment vector of the subarea and the monitoring pile number image set of all dykes and dams pile numbers, and determining area dangerous case data of the subarea; Determining a prediction dangerous case label and isolated dangerous case data of each isolated piping based on a dyke body part of each piping position in the isolated piping data, isolated image monitoring data, isolated environment monitoring data and dyke body parts, piping type labels, dangerous case grade labels and historical environment vectors of all historical piping sub data corresponding to all piping positions in all the isolated piping data, wherein the prediction dangerous case label comprises dangerous cases and non-dangerous cases, and if the prediction dangerous case label is the existence of dangerous cases, the isolated dangerous case data comprises the prediction piping type and the prediction dangerous case grade, otherwise, the isolated dangerous case data is empty; And determining the dam dangerous situation data of the dam to be inspected based on the regional dangerous situation data of all subareas in the dam potential risk region of the dam to be inspected and the prediction dangerous situation labels and the isolation dangerous situation data of all piping positions in the isolated piping data.
8. A piping risk investigation method system based on image recognition, which is characterized by being used for executing the piping risk investigation method based on image recognition as claimed in any one of claims 1 to 7, comprising: The acquisition module is used for acquiring and analyzing dam hidden danger data and historical piping data of the dam to be inspected, determining initial type hidden danger data, a risk distance threshold value and type potential risk areas of each hidden danger type label, and determining initial potential risk areas of the dam to be inspected; The determining module is used for optimizing the initial potential risk area based on the historical piping data of the dam to be inspected, determining isolated piping data, the inspection type hidden danger data of the hidden danger type labels and the merging type data of the merging type labels, and determining the dam potential risk area of the dam to be inspected; The construction module is used for constructing a piping dangerous situation model of each hidden danger type label based on the investigation type hidden danger data of each hidden danger type label of the dam to be investigated, the merging type data of each merging type label and the historical piping data, and constructing the piping dangerous situation model of each merging type label; the system comprises an acquisition module, a control module and a control module, wherein the acquisition module acquires dam image monitoring data of a dam to be inspected based on an image sensor and acquires dam environment monitoring data of the dam to be inspected based on an environment sensor group; the dangerous case module is used for conducting dangerous case investigation on the dam to be inspected based on the piping dangerous case model of all hidden danger type labels of the dam to be inspected, the piping dangerous case model of all combined type labels, the dam image monitoring data and the dam environment monitoring data, and determining dam dangerous case data.

Description

Piping dangerous case investigation method and system based on image recognition Technical Field The invention relates to the technical field of image recognition, in particular to a piping dangerous case investigation method and system based on image recognition. Background The traditional piping dangerous case investigation relies on manual inspection and single sensor monitoring, the manual inspection efficiency is low, the labor intensity is high, the inspection is subject to experience level and severe environment, the inspection is easy to miss and misjudge, the single sensor can only acquire local data, the global risk is difficult to cover, and the hidden danger and early piping sign capturing capability are insufficient. Although image recognition, multiple sensors and intelligent algorithms are introduced in current piping dangerous case investigation, the problems of low multi-source data fusion efficiency, unadapted composite hidden danger scenes of models, dynamic optimization of risk areas without combination of historical data and the like exist, isolated piping and hidden danger are easy to miss, the false alarm rate is high under complex working conditions, and the accurate and efficient piping investigation requirements are difficult to meet. Therefore, the invention provides a piping dangerous case investigation method and system based on image recognition. Disclosure of Invention The invention provides a piping dangerous situation investigation method and system based on image recognition, which are characterized by comprising the steps of analyzing acquired dam hidden danger data and historical piping data of a dam to be investigated, determining initial type hidden danger data, risk distance threshold values and type potential risk areas of each hidden danger type label, determining and optimizing initial potential risk areas of the dam to be investigated, determining isolated piping data, investigation type hidden danger data of a plurality of hidden danger type labels and merging type data of a plurality of merging type labels, determining dam potential risk areas of the dam to be investigated, constructing a piping dangerous situation model of each hidden danger type label and a piping dangerous situation model of each merging type label, collecting dam image monitoring data and dam environment monitoring data of the dam to be investigated, carrying out dangerous situation investigation on the dam to be investigated, and determining dam dangerous situation data. The method can comprehensively cover the conventional hidden danger, special risks and isolated piping, avoid risk omission, accurately adapt to single and composite hidden danger scenes, improve dangerous case identification pertinence, strengthen relevance and reliability of dangerous case judgment, improve investigation efficiency and accuracy, and realize no dead angle prevention and control of the piping dangerous case universe. The invention provides a piping dangerous case investigation method based on image identification, which comprises the following steps: S1, acquiring and analyzing dam hidden danger data and historical piping data of a dam to be inspected, determining initial type hidden danger data, a risk distance threshold value and type potential risk areas of each hidden danger type label, and determining initial potential risk areas of the dam to be inspected; S2, optimizing an initial potential risk area based on historical piping data of the dam to be inspected, determining isolated piping data, inspection type hidden danger data of a plurality of hidden danger type labels and merging type data of a plurality of merging type labels, and determining the dam potential risk area of the dam to be inspected; S3, constructing a piping dangerous situation model of each hidden danger type label based on the investigation type hidden danger data of each hidden danger type label of the dam to be investigated, the merging type data of each merging type label and the historical piping data, and constructing the piping dangerous situation model of each merging type label; S4, acquiring dam image monitoring data of the dam to be inspected based on the image sensor, and acquiring dam environment monitoring data of the dam to be inspected based on the environment sensor group; S5, conducting dangerous case investigation on the dam to be inspected based on the piping dangerous case model of all hidden danger type labels of the dam to be inspected, the piping dangerous case model of all merging type labels, the dam image monitoring data and the dam environment monitoring data, and determining dam dangerous case data. Preferably, a piping dangerous case investigation method based on image recognition obtains and analyzes dam hidden danger data and history piping data of a dam to be investigated, determines initial type hidden danger data, a risk distance threshold value and a type potential risk area of each hidden danger