CN106898014A - A kind of intrusion detection method based on depth camera - Google Patents

Abstract

The invention provides an intrusion detection method based on a depth camera, which is detected through the following steps: first, the user sets the monitoring range within the range of the camera; initializes the number of background pixels in the preset monitoring range, which is recorded as num0; records the intrusion state as state, state=0 means there is no one, state=1 means there are people, initialized to 0; for each frame of distance image data, count the number of pixels in the preset space range, and record it as numi; if numi<num0+thr1, perform the following operations : Num0=(A*num0+B*numi)/(A+B), thr1 is the first threshold of judgment, which reflects the background pixel number noise fluctuation situation, and A and B are used to adjust the update speed of the background pixel number; If numi>num0+thr2, perform the following operations: State is set to 1; where thr2 is the second threshold for judgment, by adjusting thr2 can eliminate the misjudgment of small animal invasion; otherwise, the State remains unchanged. When there is an intrusion within the preset monitoring range, the points will change, and it will be judged whether there is an intrusion according to the changed situation. Compared with the current mainstream intrusion detection methods, this method has the advantages of being basically not affected by ambient light and having high recognition accuracy.

Description

An Intrusion Detection Method Based on Depth Camera

technical field

The invention relates to the field of video surveillance, in particular to an intrusion detection method based on a depth camera.

Background technique

Video surveillance is an important part of the security solution system. In recent years, with the advancement of computer, network, image processing, and transmission technologies, video surveillance technology has also made great progress. Intrusion detection is an important part of intelligent video surveillance, and it is widely used in military and civilian fields. At present, there are mainly methods based on image and video analysis or infrared detectors for intrusion detection. Among them, the method based on image and video analysis is vulnerable to the application of ambient light changes, and it is easy to generate false alarms in scenes with complex ambient light; while the method based on infrared detectors It is susceptible to false alarms caused by environmental factors such as small animals, fog, rain and snow. In addition, differences in installation methods, angles, and locations will also cause certain effects.

Contents of the invention

The invention discloses an intrusion detection method based on a depth camera. The distance information from all objects in the scene to the camera is obtained through the depth camera, and then the point cloud of the scene (each sampling point in the space coordinate system) can be obtained through other parameters of the camera. Three-dimensional coordinate set) data, calculate the number of pixels in the preset space monitoring range and perform dynamic update, when there is an intrusion in the preset monitoring range, the number of points will change, and judge whether someone has invaded according to the changing situation

Its technical scheme is as follows:

An intrusion detection method based on a depth camera, characterized in that:

Including the following steps:

1) Configure the monitoring range: the user sets the monitoring range within the range of the camera;

2) Initialize the number of background pixels within the preset monitoring range. The initialization value can be set to the number of pixels corresponding to the entire image (for example, the distance image with a resolution of 320*240 is set to 320*240=76800) or the first frame is initialized after power-on The number of pixels in the corresponding preset monitoring range or other values greater than the number of points in the estimated monitoring range, denoted as num0;

3) record the intrusion state as state, state=0 represents no one, state=1 represents people, initialized to 0;

4) For each frame of distance image data, count the number of pixels in the preset space range, denoted as numi;

5) If numi<num0+thr1, perform the following operations: Num0=(A*num0+B*numi)/(A+B),

State is set to 0; among them, thr1, A, and B are all adjustable parameters, and thr1 is the first threshold for judgment, which reflects the noise fluctuation of the number of background pixels, and A and B are used to adjust the update speed of the number of background pixels;

6) If numi>num0+thr2, perform the following operations: State is set to 1; where thr2 is the second threshold for judgment, which is an adjustable parameter, and by adjusting thr2, misjudgment of small animals and other invasions can be eliminated;

7) If numi neither satisfies <num0+thr1 nor >num0+thr2, perform the following operations: State remains unchanged;

Further, thr1 in the step 5) is determined by the following method:

5-1) Grab N times of distance data continuously for a fixed scene (for example, N=100);

5-2) The standard deviation (mean square deviation) σ of the N-time data is calculated according to the following standard calculation formula:

Suppose there is a set of values X ₁ , X ₂ , X ₃ , ... Xn (all real numbers), the average value (arithmetic mean value) is μ, the formula is shown in the figure

The standard deviation is also known as the standard deviation, or the experimental standard deviation, and the formula is

5-3) Set thr1 to a multiple of σ, usually according to the normal distribution, you can set thr1 to 4 times of σ, and the corresponding areas of each multiple are as follows:

95.449974% of the area is within two standard deviations 2σ of the mean;

99.730020% of the area is within the range of three standard deviations 3σ around the mean;

■ 99.993666% of the area is within four standard deviations 4σ of the mean.

Further, thr2 in the step 6) is determined by the following method:

6-1) The calculation formula of the space volume represented by each sampling point is

Volume L=(d/f)^3*A, where d is the distance, f is the focal length of the camera, and A is a coefficient related to the camera sensor;

6-2) Suppose the volume of the object to be detected is greater than M, then thr2 can be set to a fixed multiple of M/L, usually 0.5 times M/L, but the value must be at least greater than thr1;

Beneficial effects: the present invention provides an intrusion detection method based on a depth camera. The depth camera obtains the distance information from all objects in the scene to the camera, and then obtains the point cloud of the scene through other parameters of the camera (each sampling point in the space coordinate system The three-dimensional coordinate set) data, calculate the number of pixels in the preset spatial monitoring range and update it dynamically. When there is an intrusion in the preset monitoring range, the number of points will change, and judge whether there is an intrusion according to the changing situation. Compared with the current mainstream intrusion detection methods, this method has the advantages of being basically not affected by ambient light and having high recognition accuracy.

Description of drawings

The present invention is described in further detail in conjunction with accompanying drawing:

Fig. 1 is a schematic diagram of the process flow of the application method;

detailed description

Hereinafter, the present invention will be explained in detail based on a preferred embodiment shown in the drawings, but the present invention is not limited to this embodiment.

In conjunction with accompanying drawing 1, this technical solution is introduced:

An intrusion detection method based on a depth camera, characterized in that:

Including the following steps:

1) Configure the monitoring range: the user sets the monitoring range within the range of the camera;

2) Initialize the number of background pixels within the preset monitoring range. The initialization value can be set to the number of pixels corresponding to the entire image (for example, the distance image with a resolution of 320*240 is set to 320*240=76800) or the first frame is initialized after power-on The number of pixels in the corresponding preset monitoring range or other values greater than the number of points in the estimated monitoring range, denoted as num0;

3) record the intrusion state as state, state=0 represents no one, state=1 represents people, initialized to 0;

4) For each frame of distance image data, count the number of pixels in the preset space range, denoted as numi;

5) If numi<num0+thr1, perform the following operations: Num0=(A*num0+B*numi)/(A+B),

State is set to 0; among them, thr1, A, and B are all adjustable parameters, and thr1 is the first threshold for judgment, which reflects the noise fluctuation of the number of background pixels, and A and B are used to adjust the update speed of the number of background pixels;

6) If numi>num0+thr2, perform the following operations: State is set to 1; where thr2 is the second threshold for judgment, which is an adjustable parameter, and by adjusting thr2, misjudgment of small animals and other invasions can be eliminated;

7) If numi neither satisfies <num0+thr1 nor >num0+thr2, perform the following operations: State remains unchanged.

Further, thr1 in the step 5) is determined by the following method:

5-1) Grab N times of distance data continuously for a fixed scene (for example, N=100);

5-2) The standard deviation (mean square deviation) σ of the N-time data is calculated according to the following standard calculation formula:

Suppose there is a set of values X ₁ , X ₂ , X ₃ , ... Xn (all real numbers), the average value (arithmetic mean value) is μ, the formula is shown in the figure

The standard deviation is also known as the standard deviation, or the experimental standard deviation, and the formula is

5-3) Set thr1 to a multiple of σ, usually according to the normal distribution, you can set thr1 to 4 times of σ, and the corresponding areas of each multiple are as follows:

95.449974% of the area is within two standard deviations 2σ of the mean;

99.730020% of the area is within the range of three standard deviations 3σ around the mean;

■ 99.993666% of the area is within four standard deviations 4σ of the mean.

Further, thr2 in the step 6) is determined by the following method:

6-1) The calculation formula of the space volume represented by each sampling point is

Volume L=(d/f)^3*A, where d is the distance, f is the focal length of the camera, and A is a coefficient related to the camera sensor;

6-2) Suppose the volume of the object to be detected is greater than M, then thr2 can be set to a fixed multiple of M/L, usually 0.5 times M/L, but the value must be at least greater than thr1;

The present invention has following characteristics:

1. The present invention performs intrusion detection based on a depth camera, and the spatial coordinate accuracy and data volume acquired by it are much higher than that of ordinary infrared detectors. Combining with the algorithm provided by the present invention, very stable and accurate intrusion detection can be realized;

2 The present invention designs a unique background update method, which relies on calculating changes in spatial object coordinates rather than the original illuminance information acquired by sensors, which can eliminate the interference of ambient light changes, compared with ordinary cameras that rely on brightness changes for background update methods , the method provided by the present invention has better robustness and more accurate updates;

3. The present invention designs a method for judging the size of intruding objects using the number of sampling points acquired within the monitoring range, which can well eliminate the interference of small-sized objects intruding (such as small animals, etc.).

The above specific embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to examples, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without Any deviation from the spirit and scope of the technical solutions of the present invention shall be covered by the scope of the claims of the present invention.

Claims (5)

1. An intrusion detection method based on a depth camera, characterized in that: Including the following steps: 1) Configure the monitoring range: the user sets the monitoring range within the range of the camera; 2) Initialize the number of background pixels in the preset monitoring range: the initialization value can be set to the number of pixels corresponding to the entire image or the number of pixels in the preset monitoring range corresponding to the first frame of power-on initialization or other values larger than the estimated monitoring range The value of the points, denoted as num0; 3) record the intrusion state as state, state=0 represents no one, state=1 represents people, initialized to 0; 4) For each frame of distance image data, count the number of pixels in the preset space range, denoted as numi; 5) If numi<num0+thr1, perform the following operations: Num0=(A*num0+B*numi)/(A+B), State is set to 0; among them, thr1, A, and B are all adjustable parameters, and thr1 is the first threshold for judgment, which reflects the noise fluctuation of the number of background pixels, and A and B are used to adjust the update speed of the number of background pixels; 6) If numi>num0+thr2, perform the following operations: State is set to 1; where thr2 is the second threshold for judgment, which is an adjustable parameter, and the misjudgment of small animal invasion can be eliminated by adjusting thr2; 7) If numi neither satisfies <num0+thr1 nor >num0+thr2, perform the following operations: State remains unchanged. 2. a kind of intrusion detection method based on depth camera according to claim 1, is characterized in that: thr1 in described step 5) is determined by the following method: 5-1) Continuously capture N times distance data for fixed scenes; 5-2) The standard deviation σ of the N-time data is calculated according to the following standard calculation formula: Suppose there is a set of values X ₁ , X ₂ , X ₃ ,...Xn, the average value of which is μ, the formula is $\mathrm{<span\; class="notranslate">\; \&sigma;</span>}<span\; class="notranslate">\; =</span>\sqrt{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$ 5-3) Set thr1 to a certain multiple of σ. 3. An intrusion detection method based on a depth camera according to claim 2, characterized in that: N is set to 100, and thr1 is set to 4 times of σ. 4. a kind of intrusion detection method based on depth camera according to claim 1, is characterized in that: thr2 in described step 6) is determined by the following method: 6-1) The calculation formula of the space volume represented by each sampling point is: volume L=(d/f)^3*A, where d is the distance, f is the focal length of the camera, and A is a coefficient related to the camera sensor; 6-2) Suppose the volume of the object to be detected is larger than M, then thr2 can be set to a fixed multiple of M/L. 5. An intrusion detection method based on a depth camera according to claim 4, characterized in that: thr2 is set to 0.5 times M/L, and at least greater than thr1.