ES2306020T3 - SMOKE DETECTION WITH A VIDEO CAMERA. - Google Patents

Abstract

The method involves examination of the moving area of at least one video image by determining the direction and the size of the moving area at the probable existence of the smoke. After positive inspection result, a part of the moving area is evaluated, depending on the information characterized for smoke regarding the existence of smoke. An independent claim is also included for the smoke detection device.

Description

Smoke detection with a video camera.

The invention relates to a method and to a device for smoke detection by analyzing al less a video image captured by a video camera that monitors An area.

US 5153722 A discloses a fire detection system with a color video camera, a frame display recorder and a computer processor for Evaluate and store images captured by the camera. Detectors ultraviolet and infrared indicate a risk in detecting a threshold value defined above. Next the processor Evaluate the images received by the camera according to various criteria, such as clear areas.

Recently, attempts have been made to use video systems, existing anyway for surveillance of security in buildings, tunnels, etc., for smoke detection. Because video images are often not of interest to an observer and also only very small modifications are generated in the video image by the smoke, a surveillance by staff on screens. In any case, the surveillance can only be done by automatic evaluation of the video images. In a known procedure for automatic examination of video images with respect to the appearance of smoke the intensity values of the Individual pixels of consecutive images. When they measure intensity values that are representative of one more image clear, caused by the presence of smoke, it is concluded that there is smoke and an alarm is triggered.

In this procedure there are among other things the problem that smoke is not recognized in front of a light background, and even a fire that only generates little smoke is not detected. further variations in clarity can trigger a false alarm, such as they are produced for example by people who move across the countryside visual of the camera. An attempt was made to solve this problem. examining in addition to the surveillance zone itself an area exterior, and when variations occur in this outer zone interrupting the observation of the surveillance zone. This procedure has the disadvantage that a fire is detected, if necessary, only after a certain delay, and that is not detect sources of smoke in the outside area provided additionally to the surveillance zone.

The objective of the present invention is based on propose an effective possibility for smoke detection by at least one video image captured by a video camera that Watch an area.

The objective is solved according to the invention in each case by means of the objects of the claims independent. Improvements of the invention are indicated in the dependent claims.

A main idea of the invention is based on that smoke is detected by analyzing at least one image of Video captured by a video camera that monitors an area. An area it may be in this respect a room, (a part of) a tunnel, a parking lot, a street or a street section, etc. Basically it check in a first stage a probable existence of smoke in the area that moves by determining the direction and of the size in an area that moves from the at least one image of video. If an area that moves has a result of positive check, then there is a certain probability of existence of smoke Next, at least a part of the area that moves based on at least one information smoke characteristic with respect to the existence of smoke. How Characteristic smoke information is considered according to the invention the speed of smoke, the number of pixels in the video image describing this movement, the variation of luminance (variation of clarity) of the at least one image of video with respect to the background, the color variation of the smoke moved and the movement of smoke.

An advantage of the process according to the invention or of the device according to the invention is based on the fact that it can Smoke detected effectively. This is especially achieved. through the evaluation divided into two parts and through the proper selection of smoke characteristic information.

The invention is explained in more detail by an exemplary embodiment represented in a figure. To this respect show

Figure 1, a block diagram according to the invention for smoke detection,

Figure 2, a simplified representation of a video image,

Figure 3, a decision diagram for the smoke detection,

Figure 4, a device according to the invention.

Figure 1 shows a block diagram according to The invention for smoke detection. From at least one video image, which was generated with a certain frequency, is get at least one image [X_ {ij} (t)] of intensity. The video image can present in this regard for example a size of 352x288 pixels. As the next stage follows the preprocessing The preprocessing has the objective of filter the areas that are relevant for smoke detection, from The video image. For this, a matrix is created first [B_ {ij} (t)] of fund accumulation. Matrix [B_ {ij} (t)] of fund accumulation is obtained from intensity [X_ {ij} (t)] images weighted with a weighting factor, indicating the weighting factor α to what extent intensity images influence the matrix [B_ {ij} (t)] of accumulation.

The accumulation matrix is determined as follow:

Bij (t) = \ alpha \ Bij (t-1) + (1- \ alpha) \ Xij (t),

α = weighting factor.

A matrix is then calculated D_ {ij} (t) = | B_ {ij} (t) -X_ {ij} (t) | subtraction for at least one area that moves. By means of the weighting based on the color of the matrix D_ {ij} (t) of subtraction is finally obtained the matrix [S_ {ij} (t)] of weighted subtraction based on color.

This subtraction matrix [S_ {ij} (t)] It is calculated from

one

being Luma {Dij} the part of clarity of Dij, ChromaU (Dij) the percentage of color U of Dij and ChromaU (Dij) the percentage of color V of I said

The probable existence of smoke in the place (i, j) is finally determined for example through the projection of the weighted subtraction matrix [S_ {ij} (t)] of the color on the x / y axis of a coordinate system Cartesian.

The projection on a coordinate system Cartesian has the following form in this regard:

2

In this example Sij has the size HxV (H = smoke speed x of smoke movement = V). Obviously the Coordinate system choice is random. So they can used for example also spherical coordinates, coordinates cylindrical, etc.

With the help of the matrix [S_ {ij} (t)] of weighted subtraction depending on the color can be checked then a probable existence of smoke in an area that moves of the video image. In the case of a probable existence of smoke defines a video image zone (ROI = Region of Interest, relevant interest area) reduced with respect to the original image Obviously several zones can also be defined ROI in a video image or in several channels. By means of the data reduction to approximately 1: 100, the size of the ROI can be in this respect for example 8x128 pixels, it is reduced considerably the processor load for true analysis or evaluation. If there is smoke in an area that moves from the image of captured video is clarified by at least one information smoke characteristic In the present example the five following information to increase the reliability of detection.

As characteristic smoke information is consider the speed of smoke (smoke movement), the number of pixels (active pixels) that describe this movement, the luminance variation (clarity variation) of the at least one video image with respect to the background, color variation (color change) of the smoke moved and the movement of the smoke (position and in the histogram).

For each ROI zone, the following smoke characteristic information:

-

he SROI smoke movement (t): v (t) = correlation of projection time and SROI (t), for example pyj (t),

-

the variance of BROI (t) and relative XROI (t) with respect to background (normal): l (t) = 1 - var {B_ {ROI} (t)} / var {B_ {ROI} (t)} for the determination of the variation of clarity,

-

active pixels of S_ {ROI} (t): a (t) = number of pixels of S_ {ROI} (t) with a value greater than 0,

-

change color: c (t) = number of pixels with {1- | ChromaU (D_ {ROI} (t)) - ChromaV (D_ {ROI} (t) |} <value threshold,

-

position and on the histogram: h (t) = projection and S_ {ROI} (t) values, for example p_ {yj} (t) is used to create a histogram with 64 channels

Then they integrate for a while determined and thus by several images the information v (t), l (t), a (t), c (t) and h (t) smoke characteristics The function has in this regard by Example as follows:

3

Based on the information integrated into the over time the respective mean value is determined.

Average movement value of smoke

F_ {V} = V

Average exchange value of clarity

F_ {L} = L

Average pixel value assets

F_ {A} = A

Average exchange value of color

F_ {C} = C

Average position value and in the histogram

F_ {H} = H.

Next it is calculated for each of these average values the probability of the existence of smoke. This is performed through pattern recognition. For each value means a discriminating value \ Psi is determined. A value δ threshold (or also a probability function) can be defined by example the discriminator as follows:

for the variation of clarity

4

or 0 \ gamma (F_ {L}) \ leq 1, with \ Gamma (x) as a function of probability.

The smoke pattern is defined by the product of all discriminators

5

or as the average value of all discriminators

6

where N_ {F} = 5 is the number of information.

Finally the decision is made if in the case of the area that moves from the video image is about the smoke image In this regard an integrator is determined I (t) that increases or decreases by a value \ sigma

I (t = 0) = 0;

if \ aleph (t) = 1, then I (t) = I (t-1) + \ sigma + (se add to S + if I (t)> S +), if not I (t) = I (t-1) - sig - (added to S - (usually 0) if I (t) < S -),

adopting \ sigma +, \ sigma - usually the value +1.

Smoke is detected and triggers for example the alarm, if I (t) exceeds a critical value?

if I (t) < \kappa,

then there is smoke

quad

If not, there is no smoke.

Figure 2 shows a representation Simplified of a VB video image. The image contains a zone that moves that represents smoke. In addition the video VB image shows a ROI zone that was determined according to the description with with respect to figure 1.

Figure 3 shows a decision diagram for smoke detection, as described in figure 1. Yes I (t) exceeds a given threshold value \ kappa is triggered the alarm and with a high probability smoke was detected. For what I (t) does not increase to infinity and thus reduce unnecessarily the reaction time for smoke detection is Define a maximum I_ {T} value. As critical time is called the Time to trigger the alarm. This time should be the most short possible.

Figure 4 shows a VR device according to the invention with a receiving unit E and a sending unit S to communicate for example with other units, such as sensors, central units, etc., and a processing unit V to perform the procedure according to figure 1. The device can be integrated in this regard in a camcorder, a central unit, etc., or represent a separate unit.

Claims (13)

         \ global \ parskip0.930000 \ baselineskip
      

1. Smoke detection procedure by analyzing at least one video image captured by a video camera that monitors an area, checking at least one area that moves from the at least one video image using the determination of the direction and size of the area that moves regarding the probable existence of smoke, being evaluated in the case of a positive test result at least a part of the at least one zone that moves based on at least one characteristic smoke information regarding the existence of smoke, characterized in that the at least one video image is generated with a certain frequency and at least one image [X_ {ij} (t)] of intensity is obtained from it, and because a matrix [B_ {ij} (t )] of background accumulation, which is obtained from the [X_ {ij} (t)] intensity images weighted with a weighting factor, indicating the weighting factor to what extent the intensity images influence the matrix [ B_ {ij} (t)] of accumulation. 2. Method according to claim 1, characterized in that, as at least one characteristic smoke information, the smoke velocity, the number of pixels describing this movement, the luminance variation of the at least one video image with respect to the video are used. background, the color variation of the smoke moved and the movement of the smoke. 3. Method according to claim 1, characterized in that with the aid of a matrix D_ {ij} (t) = | B_ {ij} (t) - X_ {ij} (t) | subtraction determines at least one area that moves. 4. Method according to claim 3, characterized in that a weighted subtraction matrix [S_ {ij} (t)] is determined from color based on the subtraction matrix [D_ {{j} (t)]. 5. Method according to claim 4, characterized in that, with the help of the weighted subtraction matrix [S_ {ij} (t)], a probable existence of smoke is verified in a moving area of the video image, and in the case of a positive test result, an area (ROI) of relevant video image reduced with respect to the original image is defined. Method according to claim 5, characterized in that the relevant video image zone (ROI) represents at least a part of the moving area of the video image. Method according to claim 5, characterized in that the relevant video image area (ROI) is represented as a rectangle with the length / width ratio of 16: 1. Method according to claim 4, characterized in that the probable existence of smoke in the place (i, j) is determined by projection of the weighted subtraction matrix [S_ {ij} (t)] as a function of the color on the axis x / y of a Cartesian coordinate system. Method according to one of the preceding claims, characterized in that the at least one characteristic smoke information is evaluated in the relevant video image area. Method according to claim 9, characterized in that the at least one characteristic smoke information mentioned is integrated for a certain time and in this way by several images and its average value is determined, and because for each of these average values it is calculated the probability of the existence of smoke. Method according to claim 10, characterized in that the probability of the existence of smoke is determined by comparison with a threshold value δ and / or by a probability function \ (x). 12. Method according to claim 10 and 11, characterized in that from the probabilities of the average values a total probability of the existence of smoke in the relevant video image area is calculated, because this total probability is integrated by several images and because when exceeding a threshold (\ kappa) an alarm is triggered by the integrated value. 13. Device (VR) for smoke detection by analyzing at least one video image captured by a video camera that monitors an area, - with a receiving unit (E) and a unit (S) broadcast to communicate with other units, - with a processing unit (V) for check the probable existence of smoke by determining of the direction and size of an area that moves from the at least a video image, in the case of a test result positive for the evaluation of at least a part of the at least one area that moves based on at least one information smoke characteristic with respect to the existence of smoke, to generate at least one video image with a certain frequency and to get at least one image [X_ {ij} (t)] of intensity of it and to get a matrix [B_ {ij} (t)] of fund accumulation from [X_ {ij} (t)] intensity images weighted with a factor weighting, indicating the weighting factor to what extent intensity images influence the matrix [B_ {ij} (t)] of accumulation.