CN103716618A - Abnormity detecting method of camera - Google Patents
Abnormity detecting method of camera Download PDFInfo
- Publication number
- CN103716618A CN103716618A CN201210380791.6A CN201210380791A CN103716618A CN 103716618 A CN103716618 A CN 103716618A CN 201210380791 A CN201210380791 A CN 201210380791A CN 103716618 A CN103716618 A CN 103716618A
- Authority
- CN
- China
- Prior art keywords
- video camera
- acquisition image
- scene
- model
- image
- 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
Images
Landscapes
- Closed-Circuit Television Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an abnormity detecting method of a camera. The method comprises the following steps: obtaining an acquisition image of the camera; detecting the color saturation of the acquisition image; when the color saturation of the acquisition image is lower than a first threshold value, determining that the camera enters a night shooting mode, otherwise, the camera maintains a daylight shooting mode; selecting proper background models according to the determination about the daylight mode or a night mode; and determining whether the acquisition image is abnormal according to a scene structure formed by comparing the acquisition image to the image edges of the background models.
Description
Technical field
The present invention is relevant with supervision camera chain, particularly a kind of exception detecting method of video camera.
Background technology
Current supervision camera chain on the market, fechtable monitors the image of scene, and is synchronously shown in display, for personnel, immediately by display, is watched the image of one or more scenes.Monitor that the image that camera chain captures can further be recorded in the Storage Medias such as video tape or computer hard disc, after occurring in particular event (as deathwatch's event), replay this image to confirm event generating process.
Yet, for fear of the thieves or robbers who act under cover of night for criminal offence front, first video camera is imposed break/turn to/out of focus/spray paint/with gimmicks such as thing cover, cause the lower invalid image of surveillance record.One of existing strick precaution mode is that video camera is installed additional to detecting loop, with whether detect video camera and surveillance between keep line.But the warning that the image transmitting line that video camera and video recorder can only be reached in this detecting loop is cut off.
Whether in order to detect video camera, be diverted, another kind of strick precaution mode is that displacement sensor (as three-axis gyroscope or three axles acceleration rule) is installed in to video camera, by displacement sensor, detects the situation whether video camera has displacement.Yet this kind of method can only be supported video camera and be turned to, if video camera is covered, the situation such as out of focus, the method cannot detect this abnormality.
Therefore, also having a kind of mode of taking precautions against is that image and the background base figure capturing for video camera compares, and whether detecting video camera has and the anomalous event such as cover, out of focus and occur.Yet, when environment light or light have great change (as moment turns off the light or turns on light), in order to avoid, detecting is wrong to be reported by mistake, need be by the manual new background model sample of setting of personnel, the flow process of inconvenience will produce extra burden to personnel like this, makes image detecting without real benefit.
Summary of the invention
In view of above problem, the invention provides a kind of exception detecting method of video camera, use the existing problem that needs manually to set new background model sample for light variation personnel of prior art that solves.
One of the present invention embodiment provides a kind of exception detecting method of video camera.Wherein, video camera has between one day screening-mode and screening-mode in one night.Exception detecting method comprises: obtain the captured acquisition image of this video camera; Detect the color saturation of this acquisition image; When the color saturation of this acquisition image is during lower than first threshold value, judge that this video camera enters shooting at night pattern or screening-mode in the daytime; According to the screening-mode of this video camera, select a corresponding background model; And judge that according to this background model whether this acquisition image is abnormal.
According to the screening-mode of this video camera, select this corresponding background model to comprise:
According to this video camera, respectively at this screening-mode or the captured background image of this shooting at night pattern in the daytime, set up this background model;
When this video camera enters in the daytime screening-mode, select this background model of screening-mode in the daytime; And
When this video camera enters shooting at night pattern, select this background model that should shooting at night pattern.
Also comprise:
In a background image, on average choose a plurality of sampling points, with the edge strength of those sampling points, set up an edge characteristic model;
This background image is cut into a plurality of scene blocks of Two dimensional Distribution and forms a scene structure model, and wherein respectively this scene block partly overlaps with this contiguous scene block; And
Merge this edge feature model and this scene structure model for this background model.
Respectively this scene block partly overlaps with this contiguous scene block.
According to this background model, judge that whether this acquisition image is abnormal, comprise:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points does not all exist, judge the event out of focus that occurs.
According to this background model, judge that whether this acquisition image is abnormal, comprise:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points of part does not exist, judge the event of covering.
If the edge feature of those sampling points of part does not exist, judge and cover event, comprise:
Whether similarly compare in a plurality of successive frames of this acquisition image those corresponding respectively scene blocks; And
When the correspondence position of dissimilar those scene blocks in each successive frame is to change continuously, judge the event of covering.
According to this background model, judge that whether this acquisition image is abnormal, comprise:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points all exists, judge the switch lamp event that occurs.
According to this background model, judge that whether this acquisition image is abnormal, comprise:
Whether the scene structure of judging this acquisition image according to this scene structure model changes;
If the scene structure of this acquisition image changes, judge the event that turns to.
Also comprise:
If this acquisition image is without extremely, the edge feature of those sampling points of this background image in this background model and those scene blocks are updated to edge feature and those scene blocks of those sampling points in this current acquisition image.
According to the exception detecting method of video camera of the present invention, can jointing edge feature and scene structure, as the background model (being edge feature model and scene structure model) of detecting anomalous event.Because edge feature has robustness under Different Light, even under low lighting environment, by infrared view, still can possess edge feature, therefore the exception detecting method of video camera of the present invention is applicable to the environment of any luminous intensity, and can resist violent light variation, avoid reporting by mistake situation.Moreover the exception detecting method of video camera of the present invention is further according to the type of edge feature model and scene structure model analysis anomalous event, and can distinguishes event out of focus, cover event, turn to event and shock event (turning to a little) etc.
Accompanying drawing explanation
Fig. 1 is for monitoring according to an embodiment of the invention the configuration diagram of camera chain.
Fig. 2 is the exception detecting method flow chart of video camera according to an embodiment of the invention.
Fig. 3 is the initialization flowchart of the exception detecting method of video camera according to an embodiment of the invention.
Fig. 4 A is the schematic diagram of edge feature model according to an embodiment of the invention.
Fig. 4 B is the schematic diagram of scene structure model according to an embodiment of the invention.
Fig. 5 A is the configuration diagram of image analysis software according to an embodiment of the invention.
Fig. 5 B is the configuration diagram of scene structure model detecting unit according to an embodiment of the invention.
Fig. 5 C is the configuration diagram of edge sampling pattern detecting unit according to an embodiment of the invention.
Fig. 6 is the thin portion flow chart of step S250 in the flow chart shown in Fig. 2.
Fig. 7 is another exception detecting method flow chart of video camera according to an embodiment of the invention.
Wherein, Reference numeral:
100 monitor that camera chain 110 monitors main frame
130 video camera 150 video recorders
170 display 400 background images
410 sampling point 420 scene blocks
500 image analysis software 510 image capturing modules
512 color conversion cell 514 calculation unit, edge
516 binarization unit 530 light change detecting module
532 color saturation detecting unit 534 amount of edge detecting units
The continuous block of 536 scene structure model detecting unit 538 covers detecting unit
550 video camera abnormity detecting module 552 scene structure model detecting units
The 5521 background block cutting fuzzy comparer of subelement 5523 unit
5524 scene structure comparer unit 554 edge sampling pattern detecting units
5541 prospect detecting subelements
5543 unstable sampling points are analyzed subelement 5545 Gauss models and are built subelement
570 module 572 abnormal deciding means of extremely transmitting messages
The unit of extremely transmitting messages, 574 anomalous counts unit 576
Embodiment
Fig. 1 is for monitoring according to an embodiment of the invention the configuration diagram of camera chain 100.
As shown in Figure 1, monitor that camera chain 100 can comprise supervision main frame 110, video camera 130, video recorder 150 and display 170.Monitor that main frame 110 essence can be host computer (as the computer system based on x86 framework) or embedded host (as the embedded system based on ARM, SoC or DSP framework), in order to move an image analysis software, and can receive the picture signal from video camera 130 and video recorder 150, and these image signal outputs to display 170 is shown.
Video camera 130 can be arranged at monitor area and towards specific direction photography, according to user demand, one or more video cameras 130 can be set.In this, video camera 130 can be digital camera, and can be connected with monitoring main frame 110 signals via interfaces such as monitoring the capturing images card of main frame 110 or network card, make to monitor that main frame 110 can receive the image that video camera 130 is taken, and capture images in display 170 demonstrations.Video camera 130 also can be analog camera, exports the image of acquisition to video recorder 150(as being connected between video camera 130 and video recorder 150 via coaxial wire signal with analog signaling).Video recorder 150 can be digital video Video Recorder (Digital video recorder, DVR), the acquisition image of the video camera 130 connecting in order to immediate backup, and be sent to supervision main frame 110 to show in display 170 after this acquisition image is further converted to digital signal.Display 170 can be cathode-ray tube display or liquid crystal display etc.
In this, the video camera 130 of embodiment of the present invention indication can be infrared camera, has infrared photography function.By whether enabling of infrared photography function, can obtain infrared photography image or colour phhotograpy image.And video camera 130 has optical detector, can detect ambient light intensity, with when ambient brightness is not enough, automatically enable infrared photography function, use and overcome luminance shortage and cause the not fogging clear problem of acquisition.
In other words, the video camera 130 of embodiment of the present invention indication has screening-mode and shooting at night pattern in the daytime, in screening-mode fechtable coloured image in the daytime, when entering shooting at night pattern, video camera 130 will be opened the infrared facility in it and take infrared view.
Fig. 2 is the exception detecting method flow chart of video camera 130 according to an embodiment of the invention.By monitoring image analysis software and the video camera 130 of main frame 110 operations, the exception detecting method of the video camera 130 shown in execution graph 2.
Please refer to Fig. 2.First, obtain the captured acquisition image (step S210) of video camera 130.Then, the color saturation (step S220) of detecting acquisition image.When the color saturation of acquisition image is during lower than the first threshold value, judge that video camera 130 enters shooting at night pattern or screening-mode (step S230) in the daytime.Continuous and, according to the screening-mode of video camera 130, select corresponding background model (step S240).If video camera 130, in screening-mode in the daytime, is selected the corresponding background model of screening-mode in the daytime; If video camera 130, in shooting at night pattern, is selected the background model of corresponding shooting at night pattern.Finally, according to selected background model, judge acquisition image whether extremely (step S250), can know that whether video camera 130 is abnormal.
In this, as previously mentioned, video camera 130 can be infrared camera, and can Auto-Sensing ambient light intensity.When video camera 130 is when screening-mode detects ambient light intensity lower than the second threshold value (as 10 Luxs (Lux)) in the daytime, video camera 130 automatically switches to infrared photography pattern (being shooting at night pattern), take and obtains the acquisition image as infrared view.Otherwise, from shooting at night pattern, switch to screening-mode in the daytime.
That is to say, when monitor area insufficient light while turning off the light (as), video camera 130 switches to infrared photography pattern, and therefore the acquisition image of video camera 130 be infrared view, and the color saturation that captures image also decreases.For further difference video camera in the daytime out of focus with the switching at night or camera lens, the anomalous event such as cover, except color saturation must be lower than the first threshold value, scene structure and marginal information also need to meet specified conditions, could correct distinguish video camera in the daytime out of focus with the switching at night or camera lens, the anomalous event such as cover.
In certain embodiments, abovementioned steps S240 can comprise the following step: first, according to video camera 130, respectively at screening-mode or the captured background image of shooting at night pattern in the daytime, set up background model.Then,, when video camera 130 enters in the daytime screening-mode, select the corresponding background model of screening-mode in the daytime; When video camera 130 enters shooting at night pattern, select the background model of corresponding shooting at night pattern.
Fig. 3 is the initialization flowchart of the exception detecting method of video camera 130 according to an embodiment of the invention.At just (between step S210 and the step S220) of exception detecting method that carries out the video camera 130 of the present embodiment, need first utilize marginal information to set up scene structure model and adopt hybrid Gauss model (Gaussian mixture model) to set up edge feature model, as shown in Figure 3, its step comprises:
Step S301: on average choose a plurality of sampling points 410 in a background image (being background scene image) of video camera 130 acquisitions, the hybrid Gauss model of setting up each sampling point 410 with the edge strength of these a little sampling points 410 becomes edge feature model, as shown in Figure 4 A.Fig. 4 A is the schematic diagram of edge feature model according to an embodiment of the invention.Sampling point 410 quantity shown in Fig. 4 A and distribution are only signal, and embodiments of the invention are non-as limit.Hybrid Gauss model only upgrades the sampling point 410 of on average choosing, and can reduce operand by this, and can accelerate to obtain operation result.
In this, the detecting of the edge strength of sampling point 410 can be used Sobel (Sobel) image border detection method to realize, but embodiments of the invention are non-as limit, also can be realized by other edge detection methods (as Robert operator, Prewitt operator or Laplacian operator etc.).After carries out image edge detection method, also can carry out binaryzation (as Otsu algorithm) calculation to the edge strength of each point in background image, to judge which sampling point 410 belongs to marginal point.That is to say, via binaryzation, the point that edge strength in those sampling points 410 can be greater than to particular value is considered as marginal point.
Step S302: background image 400 is cut into a plurality of scene blocks 420 of Two dimensional Distribution and forms a scene structure model, and wherein each scene block 420 partly overlaps with contiguous scene block 420, as shown in Figure 4 B.Fig. 4 B is the schematic diagram of scene structure model according to an embodiment of the invention.It is positive integer that background image may be partitioned into m * n scene block 420(m, n), scene block 420 quantity shown in Fig. 4 B are only signal, embodiments of the invention are non-as limit.In this, because adjacent scene block 420 partly overlaps each other, therefore can lower video camera 130 rocks caused wrong report.
By background image 400 be cut into a plurality of scene blocks 420 after, can further set up the Regional Characteristics of each scene block 420, Regional Characteristics is edge distribution and the quantity in scene block 420 for this reason, utilizes the Regional Characteristics of scene block 420 can form aforesaid scene structure model.
Step S303: merging edge feature model and scene structure model is background model.That is to say, background model comprises aforesaid edge feature model and scene structure model.Meaning, background model is set up and is formed to choose the edge feature of a plurality of sampling points 410 and background image 400 is cut into a plurality of scene block 420 in background image 400, uses by edge feature and scene block 420 and distinguishes that anomalous event is that video camera turns to, is masked, the situation such as out of focus.
With reference to Fig. 5 A, it is the configuration diagram of image analysis software 500 according to an embodiment of the invention.Image analysis software 500 can comprise image capturing module 510, light changes detecting module 530, video camera abnormity detecting module 550 and the module 570 of extremely transmitting messages.
Image capturing module 510 receives the acquisition image (as background image or present image) of video camera 130, and it is carried out to image processing.Image capturing module 510 comprises color conversion cell 512, calculation unit 514, edge and binarization unit 516.Color conversion cell 512 is in order to the acquisition image of primaries (RGB) color space is converted to the acquisition image of the color space (as HSV) of brightness and color-separated, to obtain the color saturation parameter of acquisition image.Calculation unit, edge 514 can utilize the gradient approximation of Sobel (Sobel) operator operation image brightness function, to detect the edge strength of sampling point 410.Binarization unit 516 can be utilized and the edge strength of sampling point 410 be carried out to binaryzation as aforementioned Otsu algorithm, to find out marginal point.
Light changes detecting module 530 and video camera abnormity detecting module 550 receives the acquisition image after image capturing module 510 is processed, and carry out respectively light variation detecting and video camera abnormity detecting.When the infrared photography function not enabled of video camera 130 (when ambient light is sufficient), directly with video camera abnormity detecting module 550, carry out abnormity detecting, when the infrared photography function of video camera 130 is enabled (when ambient light is inadequate), by light, change detecting module 530 and according to acquisition image, detect that light changes and carry out abnormity detecting.So, the exception detecting method of the video camera 130 of the embodiment of the present invention, no matter whether bright and clear all applicable.Particularly, under these two kinds of varying environment brightness conditions, can set up respectively aforesaid edge model and scene structure model.
Light variation detecting module 530 comprises color saturation detecting unit 532, amount of edge detecting unit 534, scene structure model detecting unit 536 and continuous block and covers detecting unit 538.
Color saturation detecting unit 532 is in order to the color saturation parameter obtained according to aforementioned color conversion cell 512, whether the color saturation of detecting acquisition image lower than the first threshold value, detects video camera 130 and be converted to shooting at night pattern by screening-mode in the daytime.On the other hand, color saturation detecting unit 532 also can be detected color saturation when lower than the first threshold value and significantly rise to the situation higher than the first threshold value, detects video camera 130 and is converted to screening-mode in the daytime by shooting at night pattern.
Amount of edge detecting unit 534 is in order to identify the quantity of edge feature according to resulting in of calculation unit, edge 514.Scene structure model detecting unit 536 is in order to acquisition image is cut into aforementioned a plurality of scene blocks 420, and whether detect each scene block 420 of current acquisition image similar to corresponding each scene block 420 of background image 400.
Whether continuous block covers detecting unit 538 changes along adjacent scene block 420 in order to differentiate dissimilar scene block 420 in successive frame (frame), to take a decision as to whether veil, covers the event of covering that camera lens is caused.In other words, block covers detecting unit 538 in order to detect whether there is the similitude of continuous scene block 420 lower than one the 3rd threshold value continuously.
Video camera abnormity detecting module 550 comprises scene structure model detecting unit 552 and edge sampling pattern detecting unit 554.Scene structure model detecting unit 552 and the edge sampling pattern detecting unit 554 of aforementioned scene structural model detecting unit 536 and video camera abnormity detecting module 550 will be in aftermentioneds.
At this, the specification exception module 570 of transmitting messages comprises abnormal deciding means 572, anomalous counts unit 574 and the unit 576 of extremely transmitting messages in advance.Abnormal deciding means 572 judges whether to occur anomalous event in order to change the detecting result of detecting module 530 and video camera abnormity detecting module 550 according to light, and judges that anomalous event is whichever.Anomalous counts unit 574 is judged as the result accumulative total frequency of anomalous event according to abnormal deciding means 572, when adding up to specific times, notify the unit 576 of extremely transmitting messages to give the alarm.Use and avoid of short duration deliberately non-and cause image frame that the situation that huge change causes wrong report occurs, as the head lamp of driving vehicle, lightning etc.
Fig. 5 B is the configuration diagram of scene structure model detecting unit 552/536 according to an embodiment of the invention.
As shown in Figure 5 B, scene structure model detecting unit 552/536 comprises the cutting of background block subelement 5521, fuzzy comparer unit 5523 and scene structure comparer unit 5525.
Background block cutting subelement 5521 is cut into aforementioned a plurality of scene block 420 in order to will capture image (as background image 400 or present image).In this, it is positive integer that current acquisition image is divided into m * n scene block 420(m, n equally), the quantity of scenic spot piece 420 is identical with the quantity of the scene block 420 of scene structure model on the spot.
Fuzzy comparer unit 5523 utilizes fuzzy similarity (Fuzzy-similarity) algorithm to calculate respectively at the similarity between scene structure model and each scene block 420 corresponding in acquisition image, use in the scene block 420 that 5525 identifications of confession scene structure comparer unit capture images with scene structure model in the dissimilar person of corresponding scene block 420.In this, the significance level of visible scene block 420 is set weight parameter, and with aforementioned similarity to the product of weight parameter as the whether similar foundation of identification.By this, can avoid the region (as corridor) because of regular change to cause wrong report.
Fig. 5 C is the configuration diagram of edge sampling pattern detecting unit 554 according to an embodiment of the invention.
As shown in Figure 5 C, edge sampling pattern detecting unit 554 comprises prospect detecting subelement 5541, unstable sampling point analyzes subelement 5543 and Gauss model builds subelement 5545.
Prospect is detected subelement 5541 in order to the edge different according to the edge feature resulting in identification and edge feature model of calculation unit, edge 514, and is regarded as foreground edge.Unstable sampling point is analyzed subelement 5543 and is analyzed edge degree of stability, for example: can by each sampling point 410 in the edge strength of different time analyze its entropy (Entopy) be whether minimum value (because of the entropy of sampling point 410 less, represent that this sampling point is more stable), to confirm that this marginal point belongs to stabilised edge.Gauss model builds subelement 5545 in order to carry out abovementioned steps S301 to step S303, and when initial, sets up aforesaid edge feature model.
Fig. 6 is the thin portion flow chart of step S250 in the flow chart shown in Fig. 2, in order to explanation detect color saturation lower than the first threshold value after, how further whether differentiation there is anomalous event and anomalous event is whichever.As shown in Figure 6, step S250 comprises:
Step S251: whether relatively capture the color saturation of image lower than the first threshold value.If the color saturation of acquisition image is lower than the first threshold value, represent this acquisition image may for infrared view or occur out of focus, cover, switch lamp or the anomalous event such as turn to, therefore enter step S2511, to judge whether to occur anomalous event according to the background model of shooting at night pattern, if not, represent this acquisition image may for the coloured image taken in the daytime or occur out of focus, cover, switch lamp or the anomalous event such as turn to, and enter step S2512.About how to determine whether, occur extremely, will illustrate in detail in Fig. 7.
In step S2511, if the anomalous event of detecting enters step S252, if not, enter step S260, to upgrade background model.
Similarly, in step S2512, if the anomalous event of detecting enters step S252, if not, enter step S260, to upgrade background model.If do not detect extremely and while entering step S260, the edge feature of the sampling point of background image in background model 410 and scene block 420 are updated to edge feature and the scene block 420 of the sampling point 410 in current acquisition image.Therefore, edge feature model and scene structure model (being background model) also can incremental learnings during process in detecting, to adapt to scene changes (as: furniture displacement, light slowly change etc.).
Step S252: whether the scene structure of judging acquisition image according to scene structure model changes, and is if so, judged to be the event that turns to; If not, enter step S253, continue to judge whether to occur other anomalous events (step S254).
Step S253: whether the edge feature of judging the sampling point of acquisition image according to edge sampling pattern exists.If the edge feature of sampling point does not all exist, can assert event out of focus occurs, if because of the event of turning off the light, still can there is (step S255) in all edge features of infrared view.If the edge feature of Subsampling point does not exist, can assert the event of may covering (step S256).If the edge feature of sampling point all exists, can assert the event (step S257) of may turning on light/turn off the light.Rear (being the rear of step S257) in detecting switch lamp event, enters step S260, to upgrade background model.
In step S260, if detect as the event of turning off the light, by correspondence in the daytime the background model of screening-mode be replaced by the background model of corresponding shooting at night pattern; Anti-, the background model by corresponding shooting at night pattern is replaced by the corresponding background model of screening-mode in the daytime.In detecting the event of covering, event out of focus or turning to after event (in step S254, S255 and S256 rear), will enter step S270, to give the alarm.
In this, in order to be further confirmed whether really to cover event, step S256 also can comprise the following step: first, whether similar distinguish corresponding scene block 420 in a plurality of successive frames of comparison acquisition image.Then,, when the correspondence position of dissimilar scene block 420 in each successive frame is to change continuously, judge the event of covering.
In order further to distinguish and to confirm turning to of video camera, step S254 also can comprise the following step: first, whether the scene block 420 in a frame of comparison acquisition image is similar with the corresponding scene block 420 adjacent to this scene block 420 in its successive frame.Then,, according to the similar person in these a little adjacent block, judge turning to of video camera.For example, if each scene block 420 in current acquisition image, compared to each similar and adjacent scene block 420 in former frame, for to left dislocation, can judge video camera 130 in the opposite direction (to the right) turn to.
Fig. 7 is another exception detecting method flow chart of video camera 130 according to an embodiment of the invention, how to utilize image analysis software and the video camera 130 that monitors main frame 110 operations in order to explanation, and detecting anomalous event is also upgraded the background model having built.In this, this flow chart will omit the flow process that initially builds background model and the flow process of distinguishing anomalous event, and its related procedure please refer to Fig. 3 and Fig. 6.
As shown in Figure 7, first, user monitors main frame 110(step S710 as long as the state of first whether the infrared function of video camera 130 being opened inputs to), monitor that camera chain 100 carries out full-automatic study and judgement by starting, without whether and manually changing background model sample for ambient light light and shade thereafter.In this, the input of this step, except the infrared function opening of user's input, also comprises the acquisition image of video camera 130.
After step S710, enter step S720, the color saturation of detecting acquisition image, to judge whether the initiate mode of the infrared function of video camera 130 changes, whether ambient light produces light and shade variation (step S730).By this, can whether change according to infrared ray initiate mode, carry out different detection process.That is to say, if change (the state-transition by infrared ray not enabled is the state that infrared ray is enabled, or the state-transition of being enabled by infrared ray is the state of infrared ray not enabled), enter step S740; If not, enter step S750.
In step S740, carry out respectively continuous block distinctiveness ratio and analyze (step S741), amount of edge detecting (step S743) and background structure comparison (step S745).
In step S741, carry out the analysis of continuous block distinctiveness ratio, whether meaning differentiated as the aforementioned dissimilar scene block 420 and in successive frame, along adjacent scene block 420, changed.Continuous and, in step S742, whether the judgement continuously quantity of different scene block 420 is less than one the 4th threshold value, and output logic judged result (be True(be/true) or False(no/vacation)).If logic judgment result is that vacation, representative may be covered event or be turned to event.
After step S743, enter step S744, to judge whether amount of edge is less than one the 5th threshold value, and output logic judged result.If logic judgment result is that vacation, representative may be covered event or event out of focus.
After step S745, enter step S746, to judge whether the background structure of acquisition image and the similarity between the scene structure model after light variation are greater than one the 6th threshold value, and output logic judged result.If logic judgment result is that vacation, representative may be covered event, event out of focus or be turned to event.
In this, as wanted, further judge anomalous event as the event of covering, event out of focus or turn to event, can judge by the flow process shown in earlier figures 6.
After step S742, step S744 and step S746, exported result is carried out to " AND " logical operation, if operation result is true (True), be that the result that step S742, step S744 and step S746 export is very, the infrared facility that representative detects video camera switches and no exceptions event, enters step S761; If operation result is false (False), there is anomalous event in representative, enters step S762.
In step S761, upgrade background model, meaning is about to the edge feature of the sampling point 410 of background image in background model and edge feature and the scene block 420 that scene block 420 is updated to the sampling point 410 in current acquisition image, and the edge feature of all sampling points 410 of resampling, and get back to step S710, next acquisition image is analyzed.
In step S762, accumulative total anomalous event number of times, and judge that whether anomalous event number of times surpasses pre-determined number (step S771), if so, sends alarm (step S772).Use and avoid of short duration deliberately non-and cause image frame that the situation that huge change causes wrong report occurs, as the head lamp of driving vehicle, lightning etc.
In step S750, judge whether aforementioned resampling completes, if complete, enter respectively step S781 and step S745(dotted line); If not, only enter step S745(pecked line).
In this, the imperfect flow process (pecked line) of resampling is described.After execution of step S745, enter step S747, to judge whether the similarity between the acquisition background structure of image and the scene structure model of same light line states is greater than one the 7th threshold value, and output logic judged result.If so, enter step S748, to upgrade background model, meaning is about to the edge feature of the sampling point 410 of background image in background model and edge feature and the scene block 420 that scene block 420 is updated to the sampling point 410 in current acquisition image; If not, representative may be covered event, event out of focus or be turned to event, enters step S762, accumulative total anomalous event number of times, and judge that whether anomalous event number of times surpasses pre-determined number (step S771), if surpass pre-determined number, sends alarm (step S772).
Then, the flow process (dotted line) while having resampled is described, performs step S745 and step S781 simultaneously.Step S745 and follow-up step S747 thereof please refer to aforementioned, and in this, it is no longer repeated.In step S781, according to sampling point 410, carry out foreground edge detecting, and enter step S782.In step S782, judge whether prospect ratio is less than one the 8th threshold value, meaning is whether part edge feature disappears (as the event of covering causes).
According to the logic judged result of step S747 and step S782, carry out " AND " logical operation, if operation result is true (True), the result that step S747 and step S782 export is very, represents no exceptions event, enter step S748, and upgrade background model; If operation result is false (False), there is anomalous event (step S748) in representative, enters step S762, accumulative total anomalous event number of times, and when exceeding pre-determined number, give the alarm (step S772).
In certain embodiments, can in step S772, specific exceptions event be given the alarm in conjunction with the flow process as shown in earlier figures 6 and Fig. 7, so that user learns what anomalous event (as cover event, event out of focus or turn to event), occur.
According to the exception detecting method of video camera 130 of the present invention, can automatically judge whether the infrared mode of video camera 130 opens according to input picture, and then comparison in the daytime or the edge feature at night and scene structure, as the background model (being edge feature model and scene structure model) of detecting anomalous event.Because edge feature has robustness under Different Light, even under low lighting environment, by infrared view, still can possess edge feature, therefore the exception detecting method of video camera 130 of the present invention is applicable to the environment of any luminous intensity, and can resist violent light variation, avoid reporting by mistake situation.Moreover the exception detecting method of video camera 130 of the present invention is further according to the type of edge feature model and scene structure model analysis anomalous event, and can distinguishes event out of focus, cover event, turn to event and shock event (turning to a little) etc.
Although the present invention with aforesaid embodiment openly as above; so it is not in order to limit the present invention, any one of ordinary skill in the art, without departing from the spirit and scope of the present invention; when doing a little change and modification, therefore scope of patent protection of the present invention is as the criterion with claims.
Claims (10)
1. an exception detecting method for video camera, this video camera has between one day screening-mode and screening-mode in one night, it is characterized in that, and this exception detecting method comprises:
Obtain the captured acquisition image of this video camera;
Detect the color saturation of this acquisition image;
When the color saturation of this acquisition image is during lower than first threshold value, judge that this video camera enters shooting at night pattern or screening-mode in the daytime;
According to the screening-mode of this video camera, select a corresponding background model; And
According to this background model, judge that whether this acquisition image is abnormal.
2. the exception detecting method of video camera as claimed in claim 1, is characterized in that, according to the screening-mode of this video camera, selects this corresponding background model to comprise:
According to this video camera, respectively at this screening-mode or the captured background image of this shooting at night pattern in the daytime, set up this background model;
When this video camera enters in the daytime screening-mode, select this background model of screening-mode in the daytime; And
When this video camera enters shooting at night pattern, select this background model that should shooting at night pattern.
3. the exception detecting method of video camera as claimed in claim 1, is characterized in that, also comprises:
In a background image, on average choose a plurality of sampling points, with the edge strength of those sampling points, set up an edge characteristic model;
This background image is cut into a plurality of scene blocks of Two dimensional Distribution and forms a scene structure model, and wherein respectively this scene block partly overlaps with this contiguous scene block; And
Merge this edge feature model and this scene structure model for this background model.
4. the exception detecting method of video camera as claimed in claim 3, is characterized in that, respectively this scene block partly overlaps with this contiguous scene block.
5. the exception detecting method of video camera as claimed in claim 3, is characterized in that, according to this background model, judges that whether this acquisition image is abnormal, comprises:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points does not all exist, judge the event out of focus that occurs.
6. the exception detecting method of video camera as claimed in claim 3, is characterized in that, according to this background model, judges that whether this acquisition image is abnormal, comprises:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points of part does not exist, judge the event of covering.
7. the exception detecting method of video camera as claimed in claim 6, is characterized in that, if the edge feature of those sampling points of part does not exist, judges and covers event, comprises:
Whether similarly compare in a plurality of successive frames of this acquisition image those corresponding respectively scene blocks; And
When the correspondence position of dissimilar those scene blocks in each successive frame is to change continuously, judge the event of covering.
8. the exception detecting method of video camera as claimed in claim 3, is characterized in that, according to this background model, judges that whether this acquisition image is abnormal, comprises:
Whether the edge feature of judging those sampling points of this acquisition image according to this edge feature model exists; And
If the edge feature of those sampling points all exists, judge the switch lamp event that occurs.
9. the exception detecting method of video camera as claimed in claim 8, is characterized in that, according to this background model, judges that whether this acquisition image is abnormal, comprises:
Whether the scene structure of judging this acquisition image according to this scene structure model changes;
If the scene structure of this acquisition image changes, judge the event that turns to.
10. the exception detecting method of video camera as claimed in claim 3, is characterized in that, also comprises:
If this acquisition image is without extremely, the edge feature of those sampling points of this background image in this background model and those scene blocks are updated to edge feature and those scene blocks of those sampling points in this current acquisition image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210380791.6A CN103716618A (en) | 2012-10-09 | 2012-10-09 | Abnormity detecting method of camera |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210380791.6A CN103716618A (en) | 2012-10-09 | 2012-10-09 | Abnormity detecting method of camera |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103716618A true CN103716618A (en) | 2014-04-09 |
Family
ID=50409135
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210380791.6A Pending CN103716618A (en) | 2012-10-09 | 2012-10-09 | Abnormity detecting method of camera |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103716618A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106344073A (en) * | 2015-07-15 | 2017-01-25 | 聪运国际企业有限公司 | saliva image identification method for predicting female ovulation period |
| CN108429892A (en) * | 2018-05-10 | 2018-08-21 | 杭州雄迈集成电路技术有限公司 | One kind being based on RGB sensor diurnal pattern automatic switching methods |
| CN112720073A (en) * | 2020-12-21 | 2021-04-30 | 珠海格力智能装备有限公司 | Tool setting gauge control method and device, storage medium and processor |
| CN113301324A (en) * | 2021-04-30 | 2021-08-24 | 重庆紫光华山智安科技有限公司 | Virtual focus detection method, device, equipment and medium based on camera device |
| JP2022069339A (en) * | 2020-10-23 | 2022-05-11 | 横河電機株式会社 | Device, system, method and program |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080317356A1 (en) * | 2007-06-25 | 2008-12-25 | Masaya Itoh | Image monitoring system |
| CN101783970A (en) * | 2009-12-22 | 2010-07-21 | 新太科技股份有限公司 | Methods, devices and systems for automatically detecting and managing fault of camera |
| CN101833813A (en) * | 2010-03-03 | 2010-09-15 | 杭州创导安全技术有限公司 | ATM machine foreign intrusion detection alarming device and method |
| CN201839380U (en) * | 2010-09-11 | 2011-05-18 | 莱阳市科盾通信设备有限责任公司 | Day and night type infrared integrated camera |
| CN102542673A (en) * | 2010-12-30 | 2012-07-04 | 广州市浩云安防科技股份有限公司 | Automatic teller machine (ATM) pre-warning method and system based on computer vision |
| CN102572508A (en) * | 2012-02-13 | 2012-07-11 | 北京文安科技发展有限公司 | Video fault diagnosis system and method |
| CN102682568A (en) * | 2011-03-10 | 2012-09-19 | 鸿富锦精密工业（深圳）有限公司 | System and method for detecting monitored image |
-
2012
- 2012-10-09 CN CN201210380791.6A patent/CN103716618A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080317356A1 (en) * | 2007-06-25 | 2008-12-25 | Masaya Itoh | Image monitoring system |
| CN101783970A (en) * | 2009-12-22 | 2010-07-21 | 新太科技股份有限公司 | Methods, devices and systems for automatically detecting and managing fault of camera |
| CN101833813A (en) * | 2010-03-03 | 2010-09-15 | 杭州创导安全技术有限公司 | ATM machine foreign intrusion detection alarming device and method |
| CN201839380U (en) * | 2010-09-11 | 2011-05-18 | 莱阳市科盾通信设备有限责任公司 | Day and night type infrared integrated camera |
| CN102542673A (en) * | 2010-12-30 | 2012-07-04 | 广州市浩云安防科技股份有限公司 | Automatic teller machine (ATM) pre-warning method and system based on computer vision |
| CN102682568A (en) * | 2011-03-10 | 2012-09-19 | 鸿富锦精密工业（深圳）有限公司 | System and method for detecting monitored image |
| CN102572508A (en) * | 2012-02-13 | 2012-07-11 | 北京文安科技发展有限公司 | Video fault diagnosis system and method |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106344073A (en) * | 2015-07-15 | 2017-01-25 | 聪运国际企业有限公司 | saliva image identification method for predicting female ovulation period |
| CN108429892A (en) * | 2018-05-10 | 2018-08-21 | 杭州雄迈集成电路技术有限公司 | One kind being based on RGB sensor diurnal pattern automatic switching methods |
| JP2022069339A (en) * | 2020-10-23 | 2022-05-11 | 横河電機株式会社 | Device, system, method and program |
| JP7415872B2 (en) | 2020-10-23 | 2024-01-17 | 横河電機株式会社 | Apparatus, system, method and program |
| CN112720073A (en) * | 2020-12-21 | 2021-04-30 | 珠海格力智能装备有限公司 | Tool setting gauge control method and device, storage medium and processor |
| CN113301324A (en) * | 2021-04-30 | 2021-08-24 | 重庆紫光华山智安科技有限公司 | Virtual focus detection method, device, equipment and medium based on camera device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8964030B2 (en) | Surveillance camera system having camera malfunction detection function to detect types of failure via block and entire image processing | |
| US10204275B2 (en) | Image monitoring system and surveillance camera | |
| JP4803376B2 (en) | Camera tampering detection method | |
| US6937743B2 (en) | Process and device for detecting fires based on image analysis | |
| US6985172B1 (en) | Model-based incident detection system with motion classification | |
| CN103716618A (en) | Abnormity detecting method of camera | |
| CN101739827B (en) | Vehicle detecting and tracking method and device | |
| CN103873825A (en) | ATM (automatic teller machine) intelligent monitoring system and method | |
| JPH0337354B2 (en) | ||
| KR101625538B1 (en) | Car Number Recognition system | |
| US20100302397A1 (en) | Motion image data generator, system using motion image data, and methods thereof | |
| CN110210427B (en) | Corridor bridge working state detection system and method based on image processing technology | |
| CN201726494U (en) | Device and system which utilize image color information to conduct image comparison | |
| TWI479432B (en) | Abnormal detection method for a video camera | |
| JP4025007B2 (en) | Railroad crossing obstacle detection device | |
| CN201435128Y (en) | Visible perimeter alarm invasion detector | |
| KR20190136515A (en) | Vehicle recognition device | |
| JPH07160891A (en) | Picture monitoring device | |
| CN111460949A (en) | Real-time monitoring method and system for preventing external damage of power transmission line | |
| CN110866462A (en) | Behavior recognition system and method integrated in intelligent police car | |
| TWI476735B (en) | Abnormal classification detection method for a video camera and a monitering host with video image abnormal detection | |
| KR101706221B1 (en) | Security camera and Method of controlling thereof | |
| CN112507993B (en) | Abnormal driving-away detection method and system, electronic equipment and storage medium | |
| JPH0514891A (en) | Image monitor device | |
| JP3613301B2 (en) | Video surveillance and tracking device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140409 |