CN-121990433-A - Automatic elevator detection violation judging method based on machine vision

CN121990433ACN 121990433 ACN121990433 ACN 121990433ACN-121990433-A

Abstract

The automatic elevator detection violation judging method based on machine vision includes the steps of obtaining video frames of elevator operation in all areas, and performing time synchronization on the video frames to construct a unified time sequence visual sequence set. And extracting abnormal events of different areas from the visual sequence set, and carrying out structural characterization on the abnormal events to generate a local abnormal event set. And carrying out time sequence analysis on the local abnormal event set to determine the time sequence of the abnormal events in different areas, and constructing a candidate causal chain among the cross-area abnormal events according to the time sequence. And carrying out consistency check on the candidate causal links according to preset elevator operation constraint so as to screen out effective causal links conforming to the violation mechanism. And carrying out violation fact judgment on the effective causal chain so as to output the violation type, the violation time and the violation area corresponding to the effective causal chain, realizing automatic closing judgment of the cross-regional violation facts, and accurately reflecting the complete formation mechanism of various high-risk violations.

Inventors

WANG JUN
LUO XIANGZHI
HU BO
LI YANYAN
ZHOU JUNQIANG
WANG MENGYING
TAN ZHENGKANG

Assignees

中博信息技术研究院有限公司

Dates

Publication Date: 20260508
Application Date: 20260112

Claims (10)

1. An automatic elevator detection violation judging method based on machine vision, which is characterized by comprising the following steps: acquiring video frames of elevator operation in each area, and performing time synchronization on the video frames to construct a visual sequence set with unified time sequence; Extracting abnormal events of different areas from the visual sequence set, and carrying out structural characterization on the abnormal events to generate a local abnormal event set; performing time sequence analysis on the local abnormal event set to determine the time sequence of the abnormal event in different areas, and constructing a candidate causal chain among the cross-area abnormal events according to the time sequence; Carrying out consistency check on the candidate causal links according to preset elevator operation constraint so as to screen effective causal links conforming to an illegal mechanism; And carrying out violation fact judgment on the effective causal link so as to output the violation type, the violation time and the violation area corresponding to the effective causal link.
2. The machine vision-based automatic determination of elevator detection violations method of claim 1, wherein the acquiring video frames of elevator operation for each zone and time synchronizing the video frames to construct a unified time-ordered set of visual sequences comprises: Collecting the video frames at a fixed sampling frame rate through monitoring each path of cameras operated by the elevator in each area, and writing local time stamps corresponding to each path of cameras into the video frames to output a video frame sequence carrying the local time stamps; converting local time stamps corresponding to all paths of cameras into relative time so as to align the starting moments of the video frame sequences output by different paths of cameras and obtain an initial synchronous video frame sequence set; Based on the initial synchronous video frame sequence set, mapping the relative time corresponding to each path of cameras to the same global time axis through a linear model, and mapping the video frames to the global time axis according to a time sequence to construct the visual sequence set.
3. The machine vision-based automatic determination of elevator detection violations method of claim 1, characterized in that the extracting abnormal events of different regions from the set of vision sequences and structurally characterizing the abnormal events to generate a set of local abnormal events comprises: setting corresponding state observables for each region, converting the image frames corresponding to each region into state fragments of the corresponding region according to the state observables, wherein each region type corresponds to one type of state fragments; calculating the state stability of each type of state segment in a set time window, and dividing an abnormal candidate interval in each region according to the state stability, a state scalar and a comparison result of a set threshold; the state stability is used for measuring the time continuity of the state segment, the state scalar is used for digitizing the state segment, and the state stability and the state scalar in the abnormal candidate interval are not lower than the corresponding set threshold value.
4. The machine vision-based automatic determination of elevator detection violations method of claim 3, characterized in that the extracting abnormal events of different regions from the set of vision sequences and structurally characterizing the abnormal events to generate a set of local abnormal events further comprises: Calculating an event intensity score within the anomaly candidate interval and classifying the event intensity score into different local anomaly event categories, the event intensity score being proportional to the state stability and a state scalar; Judging whether the time interval between adjacent abnormal events is lower than the set time interval or not under the same abnormal event type of the same area, and if so, merging the adjacent abnormal events into the same local abnormal event.
5. The machine vision-based automatic determination of elevator detection violations method of claim 1, wherein the performing a time sequence analysis on the local set of abnormal events to determine a time sequence of the abnormal events in different areas and constructing a candidate causal chain between cross-area abnormal events according to the time sequence comprises: Calculating the central moment and the event duration of each local abnormal event in the local abnormal event set, and setting a unique number for each local abnormal event to construct an event time sequence standard segment; and sequencing the event time sequence standard fragments according to the sequence of the central moments, and establishing cross-region candidate association pairs among different region abnormal events when the central moment intervals of the different region abnormal events are positioned in a set time window.
6. The machine vision based automatic determination of elevator detection violations method of claim 5, wherein the performing a time sequence analysis on the local set of abnormal events to determine a time sequence of the abnormal events in different areas and constructing a candidate causal chain between cross-area abnormal events according to the time sequence further comprises: Splicing the abnormal events of the different areas into candidate causal chains with set lengths through the edges to be connected by taking the cross-area candidate association pairs as the edges to be connected; And carrying out causal distribution on the local abnormal events in the candidate causal chain according to a time sequence to obtain role marks respectively corresponding to a head event, a tail event and an intermediate event of the candidate causal chain, and carrying out consistency check on the role marks according to a preset regional priority so as to output candidate causal chains which pass the consistency check and carry the role marks.
7. The machine vision based automatic determination of elevator detection violations method of claim 6, wherein said performing consistency check on said candidate causal links according to preset elevator operation constraints to screen out valid causal links that meet a violation mechanism comprises: Based on role marks and cross-region candidate association pairs of each abnormal event in the abnormal event sequence corresponding to the candidate causal chain, generating a region sequence corresponding to the candidate causal chain, and carrying out consistency check on the region sequence according to preset region combination constraint to determine a region violation penalty value; Extracting an abnormal event type sequence from candidate causal chains meeting the regional combination constraint, and comparing the abnormal event type sequence with a preset event type sequence template to output candidate causal chains matched with the event type sequence template and type sequence violation penalty values; and determining a time beat violation punishment value based on a comparison result of a time interval between adjacent abnormal events in the chain and a set time interval, and outputting an effective causal chain and a validity score corresponding to the effective causal chain according to the time beat violation punishment value, the regional violation punishment value and the type sequence violation punishment value.
8. The machine vision based automatic determination of elevator detection violations method of claim 7, wherein said performing a violation facts determination on said active causal link to output a violation type, a violation time, and a violation area corresponding to said active causal link comprises: Checking an event area and an event start-stop time interval in the effective causal chain to determine a chain-level ground fact corresponding to the effective causal chain, and generating an violation type corresponding to the effective causal chain based on the event type in the chain corresponding to the chain-level ground fact and a preset mapping table; And merging the chain-level ground facts of the same violation type and with the event start-stop time intervals overlapped or with the event start-stop time intervals lower than the set time intervals to output the violation facts comprising the violation type, the violation time and the violation area corresponding to the effective causal chain.
9. A terminal comprises a processor and a storage medium, and is characterized in that: The storage medium is used for storing instructions; the processor being operative according to the instructions to perform the steps of the method according to any one of claims 1-8.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1-8.

Description

Automatic elevator detection violation judging method based on machine vision Technical Field The invention relates to the technical field of artificial intelligence, in particular to an automatic elevator detection violation judging method based on machine vision. Background In engineering scenarios involving cross-regional violation causal links, elevator violations are often not triggered directly by some isolated action, but rather are composed together of a plurality of operational or state anomalies distributed in different physical areas, occurring sequentially in time, to complete violations. At present, in the prior art, the machine vision-based violation determination scheme generally takes the visual information of a single area such as the inside of a car, a landing door area or a hoistway entrance and the like as an analysis unit to respectively detect and determine abnormal behaviors in each area, but a unified expression mechanism for association of causal relationships and engineering logics among different areas is lacking, so that a system can only identify a plurality of local violation fragments and can not integrate the local violation fragments into a complete violation event with definite cause, process and result. Taking "abnormal opening of landing door-mistaken entry of person into hoistway-out-of-position running of car" as a typical example, in this process, abnormal opening of landing door itself may only be determined as abnormal state of door zone, entry of person into door zone is identified as out-of-range behavior of person, and out-of-position running of car belongs to independent running abnormality. Because the behaviors respectively occur in different areas and have intervals in time, the three are difficult to be closed into the same violation facts in engineering logic in the prior art, so that the complete formation mechanism and the severity of the high-risk violation are difficult to accurately reflect. Disclosure of Invention In view of the foregoing, it is necessary to provide an automatic elevator detection violation determination method based on machine vision. The invention adopts the following technical scheme. The invention discloses a machine vision-based automatic elevator detection violation judging method, which comprises the following steps: acquiring video frames of elevator operation in each area, and performing time synchronization on the video frames to construct a visual sequence set with unified time sequence; Extracting abnormal events of different areas from the visual sequence set, and carrying out structural characterization on the abnormal events to generate a local abnormal event set; performing time sequence analysis on the local abnormal event set to determine the time sequence of the abnormal event in different areas, and constructing a candidate causal chain among the cross-area abnormal events according to the time sequence; Carrying out consistency check on the candidate causal links according to preset elevator operation constraint so as to screen effective causal links conforming to an illegal mechanism; And carrying out violation fact judgment on the effective causal link so as to output the violation type, the violation time and the violation area corresponding to the effective causal link. Further, the obtaining the video frame of elevator operation in each area and performing time synchronization on the video frame to construct a visual sequence set with uniform time sequence includes: Collecting the video frames at a fixed sampling frame rate through monitoring each path of cameras operated by the elevator in each area, and writing local time stamps corresponding to each path of cameras into the video frames to output a video frame sequence carrying the local time stamps; converting local time stamps corresponding to all paths of cameras into relative time so as to align the starting moments of the video frame sequences output by different paths of cameras and obtain an initial synchronous video frame sequence set; Based on the initial synchronous video frame sequence set, mapping the relative time corresponding to each path of cameras to the same global time axis through a linear model, and mapping the video frames to the global time axis according to a time sequence to construct the visual sequence set. Further, the extracting the abnormal events of different areas from the visual sequence set and performing structural characterization on the abnormal events to generate a local abnormal event set includes: setting corresponding state observables for each region, converting the image frames corresponding to each region into state fragments of the corresponding region according to the state observables, wherein each region type corresponds to one type of state fragments; calculating the state stability of each type of state segment in a set time window, and dividing an abnormal candidate interval in each region according to the