CN116453278A - Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing - Google Patents
Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing Download PDFInfo
- Publication number
- CN116453278A CN116453278A CN202310331155.2A CN202310331155A CN116453278A CN 116453278 A CN116453278 A CN 116453278A CN 202310331155 A CN202310331155 A CN 202310331155A CN 116453278 A CN116453278 A CN 116453278A
- Authority
- CN
- China
- Prior art keywords
- intrusion
- deep learning
- image
- optical fiber
- fiber vibration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 72
- 239000013307 optical fiber Substances 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012544 monitoring process Methods 0.000 title claims abstract description 32
- 238000013135 deep learning Methods 0.000 title claims abstract description 22
- 238000013136 deep learning model Methods 0.000 claims abstract description 67
- 238000005286 illumination Methods 0.000 claims abstract description 32
- 230000006399 behavior Effects 0.000 claims abstract description 19
- 230000000007 visual effect Effects 0.000 claims abstract description 10
- 230000007613 environmental effect Effects 0.000 claims abstract description 9
- 238000002372 labelling Methods 0.000 claims description 29
- 238000012549 training Methods 0.000 claims description 28
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 230000007123 defense Effects 0.000 claims description 13
- 230000001066 destructive effect Effects 0.000 claims description 12
- 239000000835 fiber Substances 0.000 claims description 12
- 238000009412 basement excavation Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 4
- 230000035945 sensitivity Effects 0.000 claims description 3
- 238000010008 shearing Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 11
- 230000003287 optical effect Effects 0.000 description 11
- 230000009545 invasion Effects 0.000 description 7
- 238000010276 construction Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000001960 triggered effect Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000005065 mining Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000011410 subtraction method Methods 0.000 description 2
- 230000002194 synthesizing effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 230000033001 locomotion Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/16—Actuation by interference with mechanical vibrations in air or other fluid
- G08B13/1654—Actuation by interference with mechanical vibrations in air or other fluid using passive vibration detection systems
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Burglar Alarm Systems (AREA)
Abstract
The invention discloses an intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing, which comprises the following steps: (1) Receiving early warning information reported by an optical fiber vibration sensing device, and simultaneously acquiring an image of an optical fiber vibration position; (2) Distinguishing environmental scenes and application scenes of an intrusion site, and constructing a deep learning model library based on the distinguished multiple scenes; (3) And integrating the current environment scene and application scene of the optical fiber vibration sensing device, matching corresponding deep learning models from the deep learning model library according to wind power, illumination and visual angle information, and performing target detection on the image so as to determine the final intrusion alarm. The trained deep learning model library can be used for intelligently monitoring and identifying the intrusion behaviors in different scenes more accurately.
Description
Technical Field
The invention belongs to the field of security protection, and particularly relates to an intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing.
Background
In recent years, the distributed optical fiber sensing technology has the advantages of rapid development, high precision, high response speed, wide dynamic range and the like. The optical cable is used as a sensor and a signal transmission medium, on-site power supply is not needed, the characteristics of intrinsic safety, corrosion resistance, explosion resistance and the like are achieved, high-voltage cable anti-theft monitoring, submarine cable disturbance monitoring, petrochemical pipeline anti-theft monitoring, distributed optical fiber intrusion security and other distributed optical fiber perimeter security and the like can be effectively carried out, and safety accidents are eliminated in a sprouting state, and are prevented from happening in the future.
Sensing technology based on distributed optical fibers is widely applied to the fields of perimeter security, seismic exploration, rail transit, pipeline transportation and the like. The distributed optical fiber sensor mainly has two deployment application modes of buried ground and fence. The fence deployment mode is mainly used for perimeter security monitoring of factories, airports, military bases, prisons and the like. The buried deployment mode is mainly used for preventing underground laid wires, oil and gas pipelines and the like from being damaged by manual or mechanical excavation.
Meanwhile, the intelligent camera shooting technology plays an increasingly important role in the fields of national defense and important civil engineering. For example, in the electric power field, the external force of the excavator can damage an underground transmission line, and seriously threaten the power supply safety of industrial production and people life. At present, a plurality of power transmission corridor projects are monitored by adopting remote cameras, but operation and maintenance personnel are required to monitor on duty for a long time, and the efficiency is low. In the field of oil and gas transfer pipelines, pipeline accidents caused by excavator construction occur at times.
The optical fiber sensing technology is independently used for intrusion monitoring, and high false alarm rate exists, for example, the shaking of the optical cable fence in the windy day can be mistakenly considered to be intruded, and the optical cable buried close to the road can vibrate to generate false alarm when a large vehicle passes through. The intrusion monitoring by using the camera technology alone has higher false alarm rate, for example, camera shake under severe weather conditions such as strong wind and the like, and camera shooting blurring under insufficient illumination and the like can generate false alarm. How to efficiently and accurately monitor destructive intrusion behaviors all the day is a challenge to be solved.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, distinguish the environmental scenes and the application scenes of the scene, consider a plurality of environmental scenes to form a deep learning model library covering each destructive invasion behavior of different application scenes in a multi-dimensional manner, adopt the invasion detection signals of the distributed optical cable to trigger the camera to collect the images of the invasion scene, establish an image sample library to be used as a training data set, input the training data set into the deep learning model library for training, and the trained deep learning model library can more accurately intelligently monitor and identify the invasion behaviors in different scenes.
In order to achieve the above object, the present invention provides an intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing, comprising the following steps: (1) Receiving intrusion early warning information reported by an optical fiber vibration sensing device, and simultaneously acquiring an image of an optical fiber vibration position; the intrusion early warning information at least comprises a position, a wind power factor, a width, a duration and an intensity; (2) Distinguishing environmental scenes and application scenes of an intrusion site, and constructing a deep learning model library based on the distinguished multiple scenes; (3) And integrating the current environment scene and application scene of the optical fiber vibration sensing device, matching corresponding deep learning models from the deep learning model library according to wind power, illumination and visual angle information, and performing target detection on the image so as to determine the final intrusion alarm.
Further, the environmental scene in step (2) includes weather conditions, lighting conditions, and image sampling perspective positions; the application scene comprises a fence application defense area and a buried application defense area.
Further, in the step (1), the image is acquired by periodically acquiring and caching the monitoring image shot by the fixed camera device preset in the optical fiber vibration monitoring area, and calling the latest cache image at the reported intrusion alarm position when the optical fiber vibration sensing device alarms.
Further, in the step (1), the image is acquired by scheduling a movable camera device preset in the optical fiber vibration monitoring area to be aligned with an intrusion alarm position reported by the optical fiber vibration sensing device.
Further, the movable camera device comprises an unmanned aerial vehicle for patrol, a patrol robot and a rotary camera.
Further, before the target detection is performed through the deep learning model, detecting whether a moving target exists in the image by using a background difference method, and if the moving target exists, performing the target detection through the deep learning model; if there is no active target, then no target detection needs to be continued.
Further, an anti-climbing, anti-shearing and anti-knocking deep learning model library is at least respectively constructed for the fence by applying an anti-defense area.
Further, an anti-excavation and anti-jacking pipe deep learning model library is at least respectively constructed for the buried application anti-excavation area.
Further, the deep learning model library is constructed from three dimensions of weather conditions including normal weather and strong wind weather, illumination conditions including scenes including normal illumination and poor illumination, and image sampling view angle positions including a high-altitude view angle and a ground view angle; the deep learning model library comprises a plurality of deep learning models, and different value combinations of three dimensions correspond to one deep learning model.
Further, the training method of the deep learning model library comprises the following steps: (a) According to intrusion early warning information detected by the optical fiber vibration sensing device, automatically labeling the acquired intrusion site image, and accordingly establishing an image sample library to be used as a training data set; (b) Training the training data set based on a deep learning target detection algorithm; (c) And updating the corresponding deep learning model library according to the application scene, wind power, illumination and visual angle information.
Further, the image sample library in the step (a) comprises a positive sample library and a negative sample library, wherein the positive sample library is an image set with destructive invasive behaviors, and the negative sample library is an image set without destructive invasive behaviors.
Further, judging whether a moving target exists in the intrusion field image, if so, incorporating the intrusion field image into the positive sample library, and if not, incorporating the intrusion field image into the negative sample library.
Further, a background difference method is adopted to judge whether a moving target exists in the invasion field image.
Further, the intrusion site image with the moving target is also required to meet the optical fiber vibration alarm threshold value at the same time to be included in the positive sample library, otherwise, the intrusion site image with the moving target is also required to be included in the negative sample library.
Further, labeling information is added to the moving object in the intrusion field image, the labeling information comprises a labeling frame and a label, the labeling frame is a maximum circumscribed rectangle of a maximum communication area of the moving object, and the label is used for marking whether the intrusion field image to which the moving object belongs is a positive sample or a negative sample.
Further, the intrusion site image and the annotation information thereof are stored according to scene classification so as to be used for different deep learning models in the deep learning model library.
Further, in the step (b), training is performed by using a YOLOv5 target detection algorithm, the trained deep learning model is suitable for screening the annotation frame of the input intrusion field image in a weighted non-maximum mode, and finally, a classification conclusion and a frame regression of whether destructive intrusion behaviors exist in the moving target are output.
Further, the sensitivity of the optical fiber vibration sensing device in the step (1) for detecting vibration is adjustable.
Compared with the prior art, the invention has the beneficial effects that:
1. the distributed vibration optical cable is used as a triggering device, so that false alarms caused by unclear night or vision by the camera are reduced. The normal construction excavation of farmlands, roads and the like nearby the pipeline is not alarmed, and false alarms of buried pipeline application are reduced. And an intelligent shooting technology is adopted to further identify and confirm the alarm generated by fence shaking caused by strong wind and the like, so that false alarm of perimeter security is reduced.
2. From three dimensions of illumination, visual angle and weather, a deep learning model library is constructed for each destructive behavior, so that intrusion behaviors can be identified by finer division scenes. The training data of the scene are distinguished, the training model is more targeted, and the recognition effect is better.
3. By adopting the optical fiber sensing vibration detection technology, the camera is triggered to acquire the real intrusion image of the real environment, a sample library is established, the camera shooting angle, the image quality and the like are more in line with the field environment, the pertinence of the trained model is stronger, and the performance is more excellent.
4. The background differential technology is adopted to pre-determine the moving target, and automatic labeling is carried out, so that the labeling efficiency and accuracy are improved.
Drawings
FIG. 1 is a flow chart of one embodiment of the present invention.
FIG. 2 is a schematic diagram of a deep learning model library according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of detecting an excavator mining action (moving object) in an image by a background subtraction method in an embodiment of the present invention.
FIG. 4 is a diagram showing the comparison of real intrusion behavior and disturbance in an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1, one embodiment of the intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing of the present invention comprises the following steps: (1) Receiving intrusion early warning information reported by an optical fiber vibration sensing device, and simultaneously acquiring an image of an optical fiber vibration position; the intrusion early warning information at least comprises a position, a wind power factor, a width, a duration and an intensity; (2) Distinguishing environmental scenes and application scenes of an intrusion site, and constructing a deep learning model library based on the distinguished multiple scenes; (3) And integrating the current environment scene and application scene of the optical fiber vibration sensing device, matching corresponding deep learning models from the deep learning model library according to wind power, illumination and visual angle information, and performing target detection on the image so as to determine the final intrusion alarm.
In one embodiment, the environmental scene in step (2) includes weather conditions, lighting conditions, and image sampling perspective locations; the application scene comprises a fence application defense area and a buried application defense area.
In one embodiment, in the step (1), the image is acquired by periodically acquiring and caching a monitoring image captured by a fixed camera device preset in the optical fiber vibration monitoring area, and calling a latest cache image at the reported intrusion alarm position when the optical fiber vibration sensing device alarms.
In one embodiment, in the step (1), the image is acquired by scheduling a movable camera device preset in the optical fiber vibration monitoring area to be aligned with an intrusion alarm position reported by the optical fiber vibration sensing device.
In one embodiment, the movable camera device comprises an unmanned aerial vehicle for patrol, a patrol robot and a rotary camera.
In one embodiment, before target detection is performed by the deep learning model, a background difference method is used to detect whether a moving target exists in the image, if so, target detection is performed by the deep learning model; if there is no active target, then no target detection needs to be continued.
In one embodiment, a defense area is applied to the fence, and at least a deep learning model library of anti-climb, anti-shear and anti-knock impact is built respectively.
In one embodiment, a region of defense is applied to the subsurface, and at least a pool of deep learning models that are excavation-proof and jacking-proof is built separately.
In one embodiment, the deep learning model library is constructed from three dimensions of weather conditions including normal weather and strong wind weather, illumination conditions including normally illuminated and poorly illuminated scenes, and image sampling perspective locations including high altitude perspective and ground perspective; the deep learning model library comprises a plurality of deep learning models, and different value combinations of three dimensions correspond to one deep learning model.
In one embodiment, the training method of the deep learning model library comprises the following steps: (a) According to intrusion early warning information detected by the optical fiber vibration sensing device, automatically labeling the acquired intrusion site image, and accordingly establishing an image sample library to be used as a training data set; (b) Training the training data set based on a deep learning target detection algorithm; (c) And updating the corresponding deep learning model library according to the application scene, wind power, illumination and visual angle information.
In one embodiment, the image sample library in step (a) includes a positive sample library and a negative sample library, wherein the positive sample library is an image set in which destructive intrusion behavior is present, and the negative sample library is an image set in which destructive intrusion behavior is not present.
In one embodiment, whether a moving target exists in the intrusion field image is judged, if yes, the intrusion field image is included in the positive sample library, and if not, the intrusion field image is included in the negative sample library.
In one embodiment, a background subtraction method is used to determine whether a moving object is present in the image of the intrusion scene.
In one embodiment, the intrusion field image with the moving object is also included in the positive sample library while satisfying the fiber vibration alarm threshold, otherwise, included in the negative sample library.
In one embodiment, labeling information is added to a moving object in an intrusion scene image, the labeling information comprises a labeling frame and a label, the labeling frame is a maximum circumscribed rectangle of a maximum connected area of the moving object, and the label is used for marking whether the intrusion scene image to which the moving object belongs is a positive sample or a negative sample.
In one embodiment, the intrusion scene image and the annotation information thereof are saved by scene classification for different deep learning models in the deep learning model library.
In one embodiment, the training is performed in the step (b) by using a YOLOv5 target detection algorithm, the trained deep learning model is suitable for screening the annotation frame of the input intrusion scene image in a weighted non-maximum mode, and finally, a classification conclusion and a frame regression of whether the moving target has destructive intrusion behaviors are output.
In one embodiment, the fiber optic vibration sensing device of step (1) has an adjustable sensitivity for detecting vibrations.
The hardware module comprises an optical fiber sensing vibration detection device, a camera unit and an intrusion target detection and comprehensive judgment module, wherein the optical fiber sensing vibration detection device is responsible for detecting vibration signals by adopting optical fibers and transmitting the detected intrusion alarm signals to the intrusion target detection and comprehensive judgment module through a built-in recognition algorithm. The intrusion target detection and comprehensive judgment module is responsible for scheduling the camera shooting unit to shoot an on-site intrusion image, performing intrusion recognition on the intrusion image, and finally synthesizing the intrusion alarm of the optical fiber and the detection result of the image intrusion target, determining a final intrusion alarm and reporting the alarm to a user.
The basic flow of intrusion detection is as follows:
step 1: detection of intrusion signals from fiber vibration
The optical fiber perimeter intrusion monitoring method based on image recognition can be provided by patent CN201610476700.7, and is completed by an optical fiber vibration sensing system which comprises a detection optical cable, a monitoring host and an upper computer which are connected together; vibration caused by external invasion is detected by a detection optical cable and an optical signal is transmitted to a monitoring host; the monitoring host firstly converts the received optical signals into electric signals, then samples the electric signals and carries out analog-to-digital conversion to obtain discrete digital signals, and the digital signals are transmitted to the upper computer; the upper computer processes the acquired digital signals to obtain the characteristic quantities of the processed signals so as to form one or more waterfall graphs, pattern recognition of the images is carried out according to the form of the waterfall graphs, and if the intrusion event is judged, the intrusion alarm is triggered.
Step 2: and performing intrusion identification on the intrusion image, and finally synthesizing an intrusion alarm of the optical fiber and an image intrusion target detection result, determining a final intrusion alarm and reporting the alarm to a user.
From the view angle of shooting, the view angle of high-altitude inspection and the view angle of ordinary ground shooting are different, so that the training result is greatly different from the actual deployment effect.
The high-altitude unmanned aerial vehicle inspection coverage is wider, and a large number of cameras are not required to be paved in a very wide range. Therefore, from the engineering perspective, application scenes exist in both high altitude and ground common view angles.
The higher the camera is, the smaller the images of the ground object and the target object are, because the camera shoots from the high altitude. The complex image background has a certain influence on the identification of the excavator, so that the difficulty of accurately detecting the excavator object under high-altitude inspection is higher.
From the view of image quality, the images in the daytime and at night have a large influence on target recognition, sunlight in the daytime is sufficient, and the sight is clear, so that the target detection is relatively easy. At night, particularly in the peripheral and remote pipeline corridor areas, only partial lamplight is usually used for illumination, and the difficulty of target identification is increased.
When the camera device is arranged on a slightly high holder, and the wind power is relatively high, for example, the holder has a certain shake, and the road surface inspection robot or the unmanned aerial vehicle also can shake, so that the difficulty of intrusion target detection is increased.
According to the invention, a deep learning model library is constructed from three dimensions of illumination, view angle and weather, and as shown in fig. 2, intrusion behaviors can be identified by finer sub-scenes. The deep learning model library at least comprises 8 deep learning models, namely 8 small cubes in the figure 2, namely A. Illumination normal + high-altitude visual angle + normal weather respectively; B. poor illumination + high altitude viewing angle + normal weather; C. normal illumination + ground viewing angle + normal weather; D. poor illumination + ground viewing angle + normal weather; E. normal illumination + high altitude viewing angle + windy weather; F. poor illumination + high altitude viewing angle + windy weather; G. normal illumination, ground view angle and strong wind weather; H. poor illumination + ground viewing angle + windy weather.
Step 1: and receiving alarm information such as the position, wind power factor, width, duration, intensity and the like reported by the optical fiber vibration sensing device.
The wind power factor reported by the optical fiber vibration sensing device is calculated as follows:
the waterfall graph disclosed in the patent CN201610476700.7 can be used as a gray level graph between 0 and 255, and the average value of the gray level values of the image in a wide area of strong wind can be calculated and used as the representation of wind power factors.
f n The gray value d0 is the optical fiber start calculation position of the high wind area, and d1 is the optical fiber end calculation position of the high wind area.
The waterfall diagram adopts the technical scheme of patent CN201610476700.7, and each pixel can be a direct signal difference value, a signal variance, a correlation value, a power or energy characteristic of a certain frequency band after FFT conversion and a detail energy characteristic of each scale after wavelet decomposition.
Step 2: and integrating scene information of the current environment. Comprising the following steps:
determining whether the position is a fence or a buried application according to the position information of the defense area in the optical cable construction and laying stage,
judging whether the wind is a strong wind scene or not according to the strong wind factor reported by the optical fiber vibration sensor,
determining a scene with normal illumination and a scene with poor illumination according to the time of local sunlight intensity configuration in a camera debugging stage, for example, the scene with normal illumination in the daytime from 7 a.k.5 a.and the scene with poor illumination in the evening;
and determining whether the camera at the position is a high-altitude view angle or a common view angle according to the configuration of the camera installation construction stage.
The method for judging the strong wind scene is if the strong wind Factor is wind Exceeding the threshold Thr of high wind wind Then it is determined that the windy scene does not meet the threshold valueThe conditions are normal weather
Step 3: selecting different deep learning models according to application scenes
For the area where the fence is applied, the area can be selected from three deep learning model libraries of anti-climbing, anti-shearing net and anti-knocking impact, and each deep learning model library is of the structure shown in fig. 2.
For the buried region, two deep learning model libraries, namely anti-excavation and anti-jacking pipe, can be selected, and each deep learning model library is of a structure shown in fig. 2.
Step 4: and selecting a corresponding model from the deep learning model library according to the strong wind, illumination and visual angle information to perform intelligent detection. And determining a final intrusion alert.
The device type that uses the camera to monitor at present is more, has traditional fixed or rotation type camera, has the camera of ground or high altitude removal's inspection robot configuration, also has the unmanned aerial vehicle's that patrol in the special sky camera. For these non-stationary cameras, it is necessary to schedule the cameras to be aimed at the intrusion site based on the positional information of the alarm information of the fiber vibration detection.
Scheduling the camera to the intrusion area includes:
1) Scheduling rotatable cameras to align with location areas of an intrusion
2) The drone is dispatched to the field location area.
3) Dispatching inspection robots to field location areas
For a fixed camera or a camera at a preset position, images can be cached for a period of time all the time, if an alarm of vibration sensing is triggered, the cached images can be directly called for deep learning target detection, and therefore the time for dispatching the camera is greatly saved.
And detecting the moving target on the shot image by using a background difference method, so that the subsequent detection of the intrusion target is more accurate. The method can avoid false alarm when the excavator does not have the excavating action.
A schematic diagram of detecting an excavating action of an excavator with background differential techniques is shown in fig. 3.
Background difference method, which is commonly used for detecting moving objects in video images, is one of the mainstream methods of moving object detection at present. The basic principle is that the current frame in the image sequence and the background reference model (background image) which is already determined or acquired in real time are subtracted, and the area which is different from the background image pixel by a certain threshold value is calculated as a motion area, so that the characteristics of the position, the outline, the size and the like of the moving object are determined. The difficulty with the background differencing method is how to determine the background model from a certain number of images.
Step 1: and obtaining a certain number of images to perform background modeling to obtain a background image frame B.
The background modeling method adopted by the invention comprises median background modeling, mean background modeling, single Gaussian distribution model, mixed Gaussian distribution model, kalman filter model, advanced background model and the like.
Step 2: subtracting the gray values of corresponding pixel points of the current frame image and the background frame image, and taking the absolute value of the gray values to obtain a differential image
D n (x,y)＝|f n (x,y)-B(x,y)|
The current video image frame is fn, the gray values of the corresponding pixel points of the background frame and the current frame are respectively marked as B (x, y) and fn (x, y),
step 3: and setting a threshold T, and carrying out binarization processing on the pixel points one by one to obtain a binarized image R. Wherein, the point with the gray value of 255 is a foreground (moving object) point, and the point with the gray value of 0 is a background point;
step 4: and (3) carrying out connectivity analysis on the image R, and if the connected outline area is larger than a set threshold value, considering that a large-sized moving target exists, and carrying out intrusion detection judgment.
Training data set
According to the invention, the optical fiber sensing vibration detection technology is adopted, the camera is triggered to acquire the real intrusion image of the real environment, the sample library is established, the camera shooting angle, the image quality and the like are more in line with the field environment, the pertinence of the trained model is stronger, and the performance is more excellent.
According to the method, the moving target is determined in advance by adopting the background differential technology, automatic labeling is carried out, and the labeling efficiency and accuracy are improved.
Step 1: intrusion alert detection based on fiber vibration
When the image data amount is small in the initial stage of training of the deep learning model, the fiber vibration sensing is considered to be set as a relatively sensitive parameter, and the field data is collected as much as possible.
Step 2: and judging whether the movable object exists or not by using a background difference method.
Step 3: labeling the movable object, including:
step 3.1: labeling frame for labeling object
And (3) determining the maximum circumscribed rectangle of the area according to the maximum connected area of the movable target obtained by the background difference method in the step (2), wherein the maximum circumscribed rectangle is marked as a labeling frame. And (3) carrying out connectivity analysis on the image R, and if the connected outline area is larger than a set threshold value, considering that a large-sized moving target exists, and carrying out intrusion detection judgment. "
Step 3.2: and determining whether the sample is a positive sample or a negative sample according to the width and the intensity of the vibration optical fiber alarm in the step 1 and the maximum circumscribed rectangle in the step 3.1.
The warning width in step 1 is greater than the width threshold Thr Width And the intensity is greater than threshold Thr intensity And the maximum circumscribed rectangle in the step 3.1 is larger than the area threshold, the image of the invasion shot by the camera is considered to be marked as a positive sample, otherwise, the image is judged as a negative sample.
As shown in fig. 4, the width of the waterfall plot for a real tap disruption alarm is wider, while the disturbance of the on-road vehicle is a broken, relatively narrow signal. The wide signal triggers the camera to capture an image that is a positive sample intrusion image, while the narrow signal triggers the camera to capture an image that is a negative sample (vehicle) image.
Step 4: and storing the image and the labeling information according to different scenes.
Step 4.1: scene information of the current environment is determined. Comprising the following steps:
determining whether the position is a fence or a buried application according to the position information of the defense area in the optical cable construction and laying stage,
judging whether the wind is a strong wind scene or not according to the strong wind factor reported by the optical fiber vibration sensor,
determining a scene with normal illumination and a scene with poor illumination according to the time of local sunlight intensity configuration in a camera debugging stage, for example, the scene with normal illumination in the daytime from 7 a.k.5 a.and the scene with poor illumination in the evening;
and determining whether the camera at the position is a high-altitude view angle or a common view angle according to the configuration of the camera installation construction stage.
The method for judging the strong wind scene is if the strong wind Factor is wind Exceeding the threshold Thr of high wind wind Then a windy scene is determined.
Step 4.2: and (3) storing the image and the labeling information in the step (3) according to the scene classification of 4.1.
The labeling information comprises: labeling frame of target object, label information of whether target object is positive sample or negative sample.
Training
And the training data set established by the automatic labeling is subjected to manual auditing and correction on the image labeling information, so that the quality of the training data is improved.
The main stream target detection algorithm is roughly divided into one-stage and two-stage. the two-stage algorithm represents the R-CNN series and the one-stage algorithm represents the YOLO series. the two-stage algorithm firstly generates a network (such as an RPN network in a master rcnn) from an input image through a candidate frame, and then classifies the content of the candidate frame through a classification network; the one-stage algorithm only passes through one network to input images, and the generated results simultaneously contain position and category information. two-stage has higher accuracy than one-stage, but the calculation amount is larger, so that the calculation is slower.
Step 2: training using deep learning techniques
The invention adopts the YOLOV5 target detection algorithm to carry out model training.
The YOLO algorithm has been in constant update improvement, gradually evolving from the beginning YOLOV1 to YOLOV5.
The YOLOv5 model was published by Ultralytics company on 6/9/2020. The YOLOv5 model is improved based on the YOLOv3 model, and comprises four models of YOLOv5s, YOLOv5m, YOLOv5l and YOLOv5 x. The YOLOv5s network is the network with the smallest depth in the YOLOv5 series and the smallest width of the characteristic diagram, and has the fastest running speed and the lowest AP precision. On the basis of the other three networks, the network is deepened and widened continuously, the AP precision is improved continuously, and the operation amount is increased.
The YOLOv5 model mainly consists of a backbone network and a head network module.
The input end of the YOLOv5 model adopts a Mosaic data enhancement, an adaptive anchor frame calculation and an adaptive picture scaling technology to improve the performance.
The backbone network of the YOLOv5 model mainly comprises modules such as Conv convolution blocks, C3 structures, SPPF and the like. The C3 structure is applied to a Backbone network of the backhaul and is also applied to a Head network. The C3 structure used by YOLOv5 can improve network speed while ensuring accuracy.
The YOLOv5m head uses a multi-scale feature map for detection, with a large image to detect small objects and a small image to detect large objects. And (3) for the three different scale feature graphs of the neck, obtaining three feature graphs with the sizes of 80 x 255, 40 x 255 and 20 x 255 respectively through Conv, C3, upsample and Concat operation.
In the embodiment of the invention, 640 x 640 pictures and a YOLOv5m model are adopted. In the embodiment of the invention, the detection results of intrusion damage are two types, such as mining behavior and non-mining behavior, and 3 scale feature graphs are predicted by using 3 candidate frames with different sizes. And finally, screening the target frame by adopting a weighted non-maximum value mode, and outputting target classification and frame regression.
Finally, it should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (18)
1. The intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing is characterized by comprising the following steps of:
(1) Receiving intrusion early warning information reported by an optical fiber vibration sensing device, and simultaneously acquiring an image of an optical fiber vibration position;
(2) Distinguishing environmental scenes and application scenes of an intrusion site, and constructing a deep learning model library based on the distinguished multiple scenes;
(3) And integrating the current environment scene and application scene of the optical fiber vibration sensing device, matching corresponding deep learning models from the deep learning model library according to wind power, illumination and visual angle information, and performing target detection on the image so as to determine the final intrusion alarm.
2. The method of intrusion monitoring combined with fiber vibration sensing for deep learning intelligent detection of claim 1, wherein the environmental scene in step (2) includes weather conditions, lighting conditions, and image sampling perspective positions; the application scene comprises a fence application defense area and a buried application defense area.
3. The method for monitoring intrusion by combining deep learning intelligent detection and optical fiber vibration sensing according to claim 1, wherein in the step (1), the image acquisition mode is to acquire and cache monitoring images shot by a fixed camera device preset in the optical fiber vibration monitoring area at fixed time, and the latest cache image at the reported intrusion alarm position is called when the optical fiber vibration sensing device alarms.
4. The method for intrusion detection by combining deep learning intelligent detection with optical fiber vibration sensing according to claim 1, wherein in the step (1), the image is acquired by scheduling a movable camera device preset in the optical fiber vibration monitoring area to be aligned with an intrusion alarm position reported by the optical fiber vibration sensing device.
5. The method for intrusion detection combined with fiber vibration sensing according to claim 4, wherein the movable camera device comprises an unmanned aerial vehicle for patrol, a patrol robot and a rotary camera.
6. The method for intrusion detection by combining deep learning intelligent detection with optical fiber vibration sensing according to claim 3 or 4, wherein before object detection by a deep learning model, a background difference method is used to detect whether a moving object exists in an image, if so, object detection is performed by the deep learning model; if there is no active target, then no target detection needs to be continued.
7. The method for intrusion detection by combination of deep learning intelligent detection and optical fiber vibration sensing according to claim 2, wherein a defense area is applied to the fence, and at least a deep learning model library of anti-climbing, anti-shearing and anti-knock impact is constructed respectively.
8. The method for intrusion detection by combination of deep learning intelligent detection and optical fiber vibration sensing according to claim 2, wherein a deep learning model library for preventing excavation and jacking pipe is constructed for the buried application defense area at least respectively.
9. The method for intrusion monitoring combined with fiber vibration sensing according to claim 2, wherein the deep learning model library is constructed from three dimensions of weather conditions including normal weather and strong wind weather, lighting conditions including light normal and light poor scenes, and image sampling view angle positions including high altitude view angle and ground view angle; the deep learning model library comprises a plurality of deep learning models, and different value combinations of three dimensions correspond to one deep learning model.
10. The method for intrusion monitoring combined with fiber vibration sensing for deep learning intelligent detection according to any one of claims 1-9, wherein the training method for the deep learning model library comprises the following steps:
(a) According to intrusion early warning information detected by the optical fiber vibration sensing device, automatically labeling the acquired intrusion site image, and accordingly establishing an image sample library to be used as a training data set;
(b) Training the training data set based on a deep learning target detection algorithm;
(c) And updating the corresponding deep learning model library according to the application scene, wind power, illumination and visual angle information.
11. The method of claim 10, wherein the image sample library in step (a) comprises a positive sample library and a negative sample library, the positive sample library being an image set with destructive intrusion behavior and the negative sample library being an image set without destructive intrusion behavior.
12. The method of claim 10, wherein determining whether a moving object exists in the intrusion field image, if so, includes the intrusion field image into the positive sample library, and if not, includes the intrusion field image into the negative sample library.
13. The method for intrusion detection by combining deep learning intelligent detection with fiber vibration sensing according to claim 12, wherein a background difference method is used to determine whether a moving target exists in the intrusion scene image.
14. The method of claim 12, wherein the intrusion field image with moving objects is incorporated into the positive sample library only if the intrusion field image meets a fiber vibration alert threshold, otherwise the intrusion field image is incorporated into the negative sample library.
15. The method for intrusion monitoring by combining deep learning intelligent detection with optical fiber vibration sensing according to any one of claims 10 to 14, wherein labeling information is added to a moving object in an intrusion scene image, the labeling information comprises a labeling frame and a label, the labeling frame is a maximum circumscribed rectangle of a maximum communication area of the moving object, and the label is used for marking the intrusion scene image to which the moving object belongs as a positive sample or a negative sample.
16. The method of any one of claims 10-14, wherein the scene images and their annotation information are stored in scene categories for use in different deep learning models in the library of deep learning models.
17. The method for intrusion detection by combination of deep learning intelligent detection and optical fiber vibration sensing according to any one of claims 10 to 14, wherein the training is performed in the step (b) by using a YOLOv5 target detection algorithm, the trained deep learning model is suitable for screening the label frame of the input intrusion field image in a weighted non-maximum manner, and finally, the classification conclusion and the frame regression of whether the destructive intrusion behavior exists in the moving target are output.
18. The method for intrusion detection combined with optical fiber vibration sensing according to claim 1, wherein the sensitivity of the optical fiber vibration sensing device for detecting vibration in step (1) is adjustable.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331155.2A CN116453278A (en) | 2023-03-30 | 2023-03-30 | Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310331155.2A CN116453278A (en) | 2023-03-30 | 2023-03-30 | Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116453278A true CN116453278A (en) | 2023-07-18 |
Family
ID=87123176
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310331155.2A Pending CN116453278A (en) | 2023-03-30 | 2023-03-30 | Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116453278A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116704734A (en) * | 2023-08-08 | 2023-09-05 | 广东力创信息技术有限公司 | Monitoring and early warning method and system for preventing underground pipeline from being excavated based on Internet of things technology |
| CN117132601A (en) * | 2023-10-27 | 2023-11-28 | 山东飞博赛斯光电科技有限公司 | Pipeline mode identification method and system based on distributed optical fiber sensing |
-
2023
- 2023-03-30 CN CN202310331155.2A patent/CN116453278A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116704734A (en) * | 2023-08-08 | 2023-09-05 | 广东力创信息技术有限公司 | Monitoring and early warning method and system for preventing underground pipeline from being excavated based on Internet of things technology |
| CN116704734B (en) * | 2023-08-08 | 2023-11-24 | 广东力创信息技术有限公司 | Monitoring and early warning method and system for preventing underground pipeline from being excavated based on Internet of things technology |
| CN117132601A (en) * | 2023-10-27 | 2023-11-28 | 山东飞博赛斯光电科技有限公司 | Pipeline mode identification method and system based on distributed optical fiber sensing |
| CN117132601B (en) * | 2023-10-27 | 2024-01-23 | 山东飞博赛斯光电科技有限公司 | Pipeline mode identification method and system based on distributed optical fiber sensing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116453278A (en) | Intrusion monitoring method combining deep learning intelligent detection and optical fiber vibration sensing | |
| CN109164443A (en) | Rail track foreign matter detecting method and system based on radar and image analysis | |
| CN101872526B (en) | Smoke and fire intelligent identification method based on programmable photographing technology | |
| CN107483889A (en) | The tunnel monitoring system of wisdom building site control platform | |
| CN104971458B (en) | Based on the more fire source recognition methods for automatically tracking positioning jet stream extinguishing device | |
| KR101149916B1 (en) | Warning system of landslide | |
| CN101673448A (en) | Method and system for detecting forest fire | |
| KR102498352B1 (en) | Control device using artificial intelligence control device and intrusion alert systme including the same | |
| CN113203049B (en) | Intelligent monitoring and early warning system and method for pipeline safety | |
| PT1628260E (en) | Method and system for automatic forest fire recognition | |
| CN105574468B (en) | Video flame detection method, device and system | |
| CN105678730A (en) | Camera movement self-detecting method on the basis of image identification | |
| CN106651855A (en) | Image monitoring and shooting method for automatic identification and alarming of hidden troubles of power transmission line channel | |
| CN110703760B (en) | Newly-added suspicious object detection method for security inspection robot | |
| CN103152558B (en) | Based on the intrusion detection method of scene Recognition | |
| CN107360394A (en) | More preset point dynamic and intelligent monitoring methods applied to frontier defense video monitoring system | |
| CN112257683A (en) | Cross-mirror tracking method for vehicle running track monitoring | |
| CN114875877A (en) | Ship lockage safety detection method | |
| CN112349055A (en) | Target monitoring device and method based on radar and video linkage | |
| CN116434533A (en) | AI wisdom highway tunnel synthesizes monitoring platform based on 5G | |
| CN113096337B (en) | Moving target identification processing method for complex background and intelligent security system | |
| CN114677640A (en) | Intelligent construction site safety monitoring system and method based on machine vision | |
| KR20220072783A (en) | System and method for real-time flood detecting, and monitoring using CCTV image, and a recording medium recording a computer readable program for executing the method | |
| CN117351649A (en) | Falling stone identification monitoring system and method integrating YOLOv8 and frame difference method | |
| CN111063148A (en) | Remote night vision target video detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |