CN-117475342-B - Intelligent analysis and alarm system and method for video of gas station

CN117475342BCN 117475342 BCN117475342 BCN 117475342BCN-117475342-B

Abstract

The invention provides a gas station video intelligent analysis and alarm system and method, comprising a guardian identification module, a non-staff identification module, a crane tube identification module, an oil discharge vehicle identification module, a staff tracking module, a high risk area setting module and an off-duty time setting module; the monitoring personnel identification module is used for identifying monitoring personnel, the non-working personnel identification module is used for identifying the identity of the non-working personnel, the oil filling riser identification module is used for identifying oil filling risers working in an oil filling area and an oil discharging area, the oil discharging vehicle identification module is used for identifying oil discharging vehicles working in the oil filling area and the oil discharging area, the personnel tracking module is used for tracking working personnel and pedestrians working in a gas station and walking, the high risk area setting module is used for processing, analyzing and detecting targets existing in a concerned area only, and the CNN network is used for identifying multiple tasks and carrying out joint discrimination on identification results of an isolated model, so that the accuracy of identification is greatly improved.

Inventors

CUI JINGWEN
CHENG SIJIA
LI JIANXIANG
LI QIANDENG

Assignees

中国石油化工股份有限公司
中石化安全工程研究院有限公司

Dates

Publication Date: 20260508
Application Date: 20220719

Claims (6)

1. The intelligent video analysis and alarm system for the gas station is characterized by comprising a guardian identification module, a non-staff identification module, a crane tube identification module, an oil discharge vehicle identification module, a personnel tracking module, a high risk area setting module and an off-duty time setting module; The monitoring personnel identification module is used for identifying the identity of a monitoring personnel, wearing working clothes and judging whether the personnel leave the post, the non-working personnel identification module is used for identifying the identity of the non-working personnel and judging whether the personnel break into a high risk area, the loading arm identification module is used for monitoring the oil filling and discharging tasks of all points of the gas station by identifying loading arms working in the oil filling area and the oil discharging area, the oil discharging vehicle identification module is used for identifying oil discharging vehicles working in the oil filling area and the oil discharging area and monitoring the oil discharging tasks of all points of the gas station, the personnel tracking module is used for tracking the working personnel and pedestrians working in the gas station and walking to acquire the moving track and the identity ID of the personnel in the area, the high risk area setting module is used for distributing the concerned areas operated by different algorithms, and only carrying out processing analysis and detection on the targets existing in the concerned areas, and the off-post time setting module is used for controlling the alarm threshold; the working process of the oil filling riser identification module comprises the following steps: a. Collecting video image data containing a crane tube in an actual scene of a gas station, wherein the video image data comprises a refueling process video in a refueling area and a discharging process video in a discharging area; b. Decomposing the video into static frames, extracting pictures containing objects to be identified, dividing the pictures into a training set and a verification set, marking detection frames of identification targets on the training set and the verification set by LabelImg marking software, and marking 'oil filling riser'; c. data enhancement operation is carried out on the training set, contrast and brightness are adjusted, scaling is carried out to the same size, and normalization processing is carried out D. adopting a classical target detection network YOLOV as a reference network of a crane tube identification module to train the preprocessed picture so as to obtain a crane tube identification model; e. inputting real-time video of a oiling area or an oil unloading area to be identified into a crane pipe identification model to obtain an identification result; The working process of the oil discharge vehicle identification module comprises the following steps: a. collecting video image data containing an oil discharging vehicle, wherein the video image data comprises an oil charging process video in an oil charging area and an oil discharging process video in the oil discharging area; b. Decomposing the video into static frames, extracting pictures containing objects to be identified, dividing the pictures into a training set and a verification set, marking detection frames of identification targets on the training set and the verification set by LabelImg marking software, and marking an oil discharge vehicle; c. data enhancement operation is carried out on the training set, contrast and brightness are adjusted, scaling is carried out to the same size, and normalization processing is carried out D. Adopting a classical target detection network YOLOV as a reference network of a crane tube identification module to train the preprocessed picture so as to obtain an unloading vehicle identification model; e. inputting a real-time video of a refueling area or a discharging area to be identified into an unloading vehicle identification model to obtain an identification result; the working process of the personnel tracking module comprises the following steps: a. Using deep sort as a person tracker; b. Dividing a personnel detection frame into a high-resolution detection frame and a low-resolution detection frame by setting a high-resolution threshold and a low-resolution threshold, matching tracks predicted by the high-resolution detection frame and the personnel tracker, and matching the tracking tracks by using the low-resolution detection frame if the high-resolution detection frame is not matched with the tracking tracks; c. and inputting the real-time video stream of the oiling area to be tracked and the real-time video stream of the oil unloading area to be tracked into a personnel tracking detection model to obtain a tracking result.
2. The intelligent analysis and alarm system for a gas station video according to claim 1, wherein the operation of the guardian identification module comprises the steps of: a. collecting video image data containing guardianship personnel in an actual scene of a gas station; b. Decomposing the video into static frames, extracting pictures containing objects to be identified, dividing the pictures into a training set and a verification set, marking detection frames of identification targets on the training set and the verification set by LabelImg marking software, and marking guardians; c. performing data enhancement operation on the training set, adjusting contrast and brightness, scaling to the same size and performing normalization treatment; d. a classical target detection network YOLOV is adopted as a reference network of a guardian identification module to train the preprocessed picture, so as to obtain a guardian identification model; e. and inputting the real-time video of the oiling area or the oil discharging area to be identified into a guardian identification model to obtain an identification result.
3. A gas station video intelligent analysis and alarm system according to claim 2, wherein said video image data comprises a fueling process video in a fueling area and a fueling process video in a fueling area.
4. The intelligent analysis and alarm system for a gas station video according to claim 1, wherein the working process of the non-staff identification module comprises the steps of: a. Collecting video image data containing non-staff in an actual scene of a gas station, wherein the video image data comprises a refueling process video in a refueling area and an oil discharging process video in an oil discharging area; b. Decomposing the video into static frames, extracting pictures containing objects to be identified, dividing the pictures into a training set and a verification set, marking detection frames of identification targets on the training set and the verification set by LabelImg marking software, and marking non-staff; c. performing data enhancement operation on the training set, adjusting contrast and brightness, scaling to the same size and performing normalization treatment; d. adopting a classical target detection network YOLOV as a reference network of a non-staff identification module to train the preprocessed picture so as to obtain a non-staff identification model; e. And inputting the real-time video of the oiling area or the oil unloading area to be identified into a non-staff identification model to obtain an identification result.
5. The intelligent analysis and alarm system for gas station video according to claim 1, wherein the off-duty time setting module counts the time of the first video frame in which no guardian is detected until the frame in which no guardian is detected is last time, and the time difference value is compared with the set off-duty time threshold to determine whether an alarm is formed.
6. A method for intelligent analysis and alarm of a gas station video, characterized in that a system for intelligent analysis and alarm of a gas station video according to any one of claims 1-5 is adopted, comprising the following steps: S1, acquiring video streams from cameras of a refueling area and an oil discharging area of a gas station, and inputting the video streams into a gas station video intelligent analysis and alarm system carried on a video intelligent analysis box; S2, when the oil discharging vehicle is identified by the oil discharging vehicle identification module, and after the oil filling pipe is connected by the oil filling pipe identification module, starting the personnel tracking module and the guardian identification module, timing the departure of guardianship, and alarming the guardian when the time exceeds the time threshold required by the management standard of the gas station; When the oil unloading vehicle identification module does not identify an oil unloading vehicle, the high-risk area module and the non-staff identification module are started to perform area calibration on a high-risk oil tank area of a gas station and the oil gas recovery device, and when the non-staff is identified to enter, staff intrusion alarm is performed in a non-oil unloading period; And S3, pushing the alarm result to the intelligent terminal, and updating the data in real time.

Description

Intelligent analysis and alarm system and method for video of gas station Technical Field The invention relates to the technical field of safety control of gas stations, in particular to a system and a method for intelligent analysis and alarm of a video of a gas station. Background Petroleum is industrial blood, the status of the world energy industry is important, with the vigorous development of China industry and economy, the demand of petroleum is increased year by year, the number of gas stations is increased, due to the fact that the number of gas stations is large, scenes are complex, the safety requirements are increased, traffic is increased, pedestrians and vehicles in the gas stations are more complex, management personnel are tedious, traditional video monitoring needs manual real-time supervision, massive image information is inevitably missed, a hard disk video recorder is needed to store video recordings for 24 hours, most of the content is invalid, therefore, a large amount of hardware resources are wasted, video storage equipment needs huge manpower to maintain, and the video recordings in a hard disk are called up to have a plurality of defects, accidents cannot be timely handled, and the best case processing time is often obtained. Therefore, the conventional video monitoring method cannot well meet the requirement of multi-task monitoring. In order to ensure that the operation of effectively tracking vehicles and personnel in a refueling area and a refueling area in a gas station, monitoring refueling, discharging oil and the like with certain dangers and prevent accidents in daily production of the gas station, the gas station has started to apply an intelligent video analysis technology to tasks of monitoring safe loading and unloading and operation of the personnel in the refueling area and the refueling area, vehicle identification, pedestrian tracking and the like in real time. At present, related researches develop on an intelligent video recognition monitoring alarm system. CN 200972689Y discloses an intelligent video recognition monitoring alarm system, which is characterized by analyzing a motion video discrete frame image, combining image brightness change information and region change information, judging whether the system belongs to a person, an animal or a fire condition, and finally judging whether the result is alarm or not. The prior art does not consider that tasks such as guardian identification, crane tube identification, oil discharge vehicle identification and the like are executed by using an isolated model and the results are judged in a combined mode, personnel tracking cannot be optimized, a repeated alarm elimination mechanism cannot be realized, the time interval allowed by personnel leaving can not be adjusted, the workload of manual confirmation is reduced, and the effectiveness of later data statistics is improved. The system design cannot be performed in combination with the background, and the partition is performed, so that the recognition effect is not good. Disclosure of Invention Aiming at the problems in the prior art, the invention provides a system and a method for intelligent analysis and alarm of a gas station video, which are reasonable in design, overcome the defects in the prior art and have good effects. In order to achieve the purpose 1, the invention adopts the following technical scheme: A video intelligent analysis and alarm system of a gas station comprises a guardian identification module, a non-staff identification module, a crane tube identification module, an oil discharge vehicle identification module, a personnel tracking module, a high risk area setting module and an off-duty time setting module; the monitoring personnel identification module is used for identifying the identity of a monitoring personnel, wearing working clothes and judging whether the personnel leave the post, the non-working personnel identification module is used for identifying the identity of the non-working personnel and judging whether the personnel break into a high risk area, the oil filling riser identification module is used for monitoring the oil filling and discharging tasks of all points of the gas station by identifying the oil filling riser working in the oil filling area and the oil discharging area, the oil discharging vehicle identification module is used for identifying the oil discharging vehicles working in the oil filling area and the oil discharging area and monitoring the oil discharging tasks of all points of the gas station, the personnel tracking module is used for tracking the working personnel and the pedestrians in the gas station and obtaining the moving track and the identity ID of the personnel in the area, the high risk area setting module is used for distributing the concerned areas operated by different algorithms, only carrying out processing analysis and detection on the targets existing in the concerned areas, and the off-p