TW201430721A - Method for detecting violent behavior of persons - Google Patents

Abstract

A method for detecting violent behavior of persons is executed by a monitor system, and includes: recognizing whether a person /persons enters a target zone in accordance with images delivered form a camera; comparing a present image with prior sequential images and calculating a motion value representing a variation between each two images; determining whether the motion value is greater than a default threshold, if yes, adding 1 to an abnormal count and recording the time representing the present image; determining whether an interval between the time representing the present image and a first abnormal time or a last abnormal time is not greater than a default time period, if not, setting the abnormal count to zero, if yes, determining whether the abnormal count greater than a default count, and if yes, outputting information corresponding to the occurrence of violent behavior of a person. Thus, violent behavior of a person can be detected precisely and promptly.

Description

Personnel behavior detection method

The invention relates to an image recognition method, in particular to a method for detecting a strong behavior of a person.

In the halls of buildings, elevators, ATMs, and alleys, it is often necessary to install monitors. The surveillance screens of the monitors are transmitted to the remote security company or management office for security management by remote monitoring.

With the evolution of image recognition technology, various human behavior detection technologies have been developed in recent years, and alarms are issued when abnormal conditions are detected, thereby reducing the burden of manual monitoring. Among them, Motion History Images (MHI) technology can record the information of the human body in time and space, and use MHI information to further analyze the behavior of people.

The existing MHI information processing methods mostly distinguish the limbs in detail, and define various actions and postures in advance, and then identify the actions in a manner of comparing features. However, in real life, the human body is tall and thin, and the body movements are very diverse. It is not easy to make a precise and universal definition, so that the prior art has poor recognition effect on abnormal conditions, often false alarms or unable to detect abnormal conditions. Failure to issue an alert in a timely manner.

Therefore, the object of the present invention is to provide a method for detecting a strong behavior of a person who can accurately recognize the intense behavior of a person and perform relevant output in time.

Therefore, the method for detecting a strong behavior of the present invention is performed by a monitoring system including a camera that takes an image toward a target area, and A processing unit that receives images from the camera. The method comprises the following steps performed by the processing unit:

(A) According to the image transmitted by the camera, it is recognized that a person enters the target area, and the number of abnormal times is set to zero.

(B) Calculating a momentum value representative of the degree of change for comparing a current image with a previous plurality of sequence images.

(C) determining whether the momentum value is greater than a predetermined threshold, and if so, performing step (D).

(D) Add 1 to the number of abnormalities and record the time corresponding to the current image.

(E) determining whether the time corresponding to the current image and the interval between the first abnormal time or the last abnormal time does not exceed a predetermined length of time, and if yes, proceeding to step (F), otherwise returning to step (A).

(F) determining whether the abnormal number of times is greater than a predetermined number of times, if yes, proceeding to step (G), and if not, returning to step (B).

(G) It is judged that a person's intense behavior occurs and relevant output is performed.

In the step (B), the current image is compared with the previous image and the previous image, and the values representing the degree of change between the two images are respectively obtained, and then added to obtain the momentum value. The value of the degree of change between the two images is, for example, calculated by subtracting the RGB component values of the respective pixels. The step (C) may further determine whether the current image has a hand feature, and if the two determination conditions are all satisfied, the step (D) is performed; the step (F) is an action of determining that a person is performing the destruction.

Alternatively, the step (C) also determines whether the contour of the person in the current image is Overlap, if the two determination conditions are met, step (D) is performed; even, the step (C) identifies the hand region for the current image, and determines whether the horizontal edge of the contour of the hand region is more than the vertical edge If the three judgment conditions are satisfied, the step (D) is performed.

The utility model has the advantages that the comprehensive logical judgment is made by using parameters such as the degree of image change and time, and the abnormal situation of the intense behavior of the personnel can be accurately found without consuming a large amount of calculation resources.

The above and other technical contents, features and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments.

Referring to FIG. 1 and FIG. 2, the method for detecting the intense behavior of the inventor of the present invention is executed by a monitoring system 100. When a person enters a target area, the detecting personnel can initiate a fierce behavior of destroying the device. When multiple personnel enter the target area, the user can also start. Detect if there is a fierce behavior of fighting. The monitoring system 100 includes a camera 1 that captures images of the target area, and a processing unit 2 that receives images from the camera 1. The aforementioned target area is, for example, a room in which the cash dispenser is placed.

The method includes the following steps:

Step S11 - The camera 1 continues to take images of the target area to obtain a sequence image.

The following steps are processed by the processing unit 2 in accordance with the image transmitted by the camera 1.

Step S12 - detecting whether a person enters, that is, whether a single or plural person enters the target area. The specific technical means of this step are not The specific means are limited, and the prior art, such as foreground tracking or region-based tracking, can be used to detect the changed region by subtracting the current image from the background image, and further setting rules for screening; or using contour tracking (Contour) -Based Tracking), find the contour line and track it according to the change of the contour line; or use Feature-Based Tracking to first extract features for the object to be tracked, such as center of gravity, area, etc. The feature is used to track the object; or Model-Based Tracking is used to first build the object model, build the motion model, and then search for objects from the continuous image. If there is a person entering, the process proceeds to step S13, and if not, the process proceeds to step S12.

Step S13 - setting an abnormal number of times to 0 and counting parameter i to be 0. Next, the momentum condition determination in step S14 and the characteristic condition determination in step S15 are performed, and the condition judgment result is comprehensively judged in step S16. Step S14 includes step S141 and step S142.

In step S141, a value representing the degree of change between the two images is obtained separately from the previous image and the previous image, and then added to obtain a newly defined momentum value of the invention. The specific calculation method is detailed below. In this embodiment, when the first and second images are executed in this step, since there is no previous or previous image, the process returns to step S13, and the next image is taken again. The conditions of S14 and S15 are judged.

For example, in the processing situation after the third image, the camera captures an image every 30 milliseconds (ms). This step firstly targets the image F3 with a time of 1 minute 10 seconds 00 milliseconds and the time is 1 minute 09 seconds 30 Milliseconds of each The RGB component values of the pixels are differentiated and compared - if there is a person moving, the image F3 will have RGB component values of some pixels (0 to 255 each) which are different from the image F2, and the difference of these RGB component values is greater than a pixel variation threshold. The pixel is directly set to 255 in the image F3, and the pixel whose value of the RGB component value is less than or equal to the pixel change threshold is set to 0 in the image F3.

Then, an interesting sampling area is locked. In this embodiment, the monitoring person presets to select a rectangular block in which the RGB component value of the image is 255, which may be densely populated as a sampling area. Find the number of pixels in the sampling area with RGB component value of 255 as the numerator. The number of pixels in the whole sampling area is the denominator, and obtain a value between 0 and 1, which is defined as an initial momentum value. The value of the degree of change between the two images is used to quantify the intensity of the behavior.

At the same time, the same method for the RGB component values of each pixel and the initial momentum value calculation are performed by using the above method for the image F3 with a time of 1 minute 10 seconds and 00 milliseconds and the image F1 with a time of 1 minute 09 seconds 00 milliseconds. And then summed to obtain the aforementioned momentum value for subsequent calculation.

Step S142 - determining whether the momentum value is greater than a predetermined threshold, and recording the judgment result of the step, that is, the momentum condition judgment result.

Step S15 - Identify whether there is a hand feature for the current image, and record the feature condition judgment result. The hand feature of this embodiment is defined as a prominent point of the foreground; therefore, this step first uses the existing foreground detection technology to find the foreground, and then uses the contour to detect the protruding portion. If the protruding portion is detected, it is determined that there is a hand feature.

Step S16 - The momentum condition judgment result and step for step S142 S15 characteristic condition judgment result, analyze whether both conditions are true? If so, it means that there is a rapid movement of the hand in the current image capturing, so step S17 is performed. If not, the process returns to step S13, and the next image is taken to determine the conditions of steps S14 and S15.

Step S17 - Adding the number of abnormalities by 1, incrementing the counting parameter i by 1, and recording the time T _i as the time corresponding to the current image.

Step S18 - determining whether the time T _i corresponding to the current image and the interval between the first abnormal time T ₁ or the last abnormal time T _i-1 does not exceed a preset time length, and if so, the rapid movement of the representative hand may be continuous , instead of the occasional transient, therefore proceeding to step S19, if otherwise returning to step S13 to return to zero, the number of abnormalities is restarted. This embodiment is exemplified by T _i -T ₁ , and the preset time length is 3 seconds, but the invention is not limited thereto.

Step S19 - determining whether the abnormal number of times is greater than a preset number of times, if yes, indicating that the current image captures a rapid movement of the hand of the person at the time, and may be subjected to a violent behavior of destruction, so step S20 is performed, otherwise, returning to step S13 Thereafter, the next image is taken and the condition determinations of steps S14 and S15 are performed.

In step S20, it is determined that a person's intense behavior occurs and relevant output is performed, for example, the alarm of the target area is sounded, or the display screen of the coordinated security management center indicates an abnormal condition.

With the above calculation technique, for the sequence image shown in FIG. 3, each time the flow proceeds to step S17 during the process, the number of abnormalities is mostly accumulated, and when the preset number of times is accumulated, an alarm is issued. For the sequence image shown in Figure 4, since the person is not able to move quickly, there is no rapid movement of the hand. Proceeding to step S17, the number of abnormalities does not accumulate, so no alarm is issued.

Referring to FIG. 5, when step S12 detects that a plurality of people enter the target area, the fierce behavior detection of the fight is additionally performed.

Step S21 - setting an abnormal number of times to 0 and counting parameter j to be 0. Next, the person contact condition determination in step S22, the momentum condition determination in step S23, and the characteristic condition determination in step S24 are performed, and the condition judgment result is comprehensively judged in step S25. Step S23 includes step S231 and step S232; step S24 includes step S241 and step S242. However, the present invention is not limited to the comprehensive judgment of the three conditions of the embodiment, and may be any two of the conditions, for example, only determining whether the person is in contact and whether the momentum condition is satisfied.

Step S22 - It is judged whether the contours of the persons overlap.

The specific calculation means in this embodiment is for moving objects, that is, in step S12, for detecting the person entering the target area, for example, using a component labeling algorithm to individually select the moving object to form a tracking frame, this step is judged. Whether the tracking frames are connected, or whether the same foreground is used - meaning whether the same pixel is included in the component symbol algorithm at the same time by multiple tracking frame tags.

In step S231, a value representing the degree of change between the two images is obtained separately from the previous image and the previous image, and then added to obtain the momentum value. For specific technical means, please refer to the foregoing step S141, and the details are not repeated here.

Step S232 - determining whether the momentum value is greater than a predetermined threshold, and recording the momentum condition determination result.

Step S241 - Identify the hand area for the current image. The specific technical means is to set the upper half area as the hand area by using the connected tracking frame of step S22.

Step S242 - It is judged whether or not the horizontal side of the outline of the hand area is more than the vertical side, and the condition judgment result is recorded. Specifically, the hand region is first cut into 16×16 blocks, and then the horizontal edge and the vertical edge are calculated for the hand contour in each block. In this embodiment, in detail, between +45°~-45° and -135°~+135° in the contour are classified as horizontal edges, between +45° and +135°, and Between -45° and -135° is classified as a vertical side, and the horizontal and vertical sides are compared.

Step S25 - For the human contact condition determination result of step S22, the momentum condition determination result of step S232, and the characteristic condition determination result of step S242, it is analyzed whether all three conditions are satisfied. If yes, it means that the current image captures the rapid movement of the plurality of people at the time and has the raised hand, so step S26 is performed, and if not, the process returns to step S21, and the next image is taken to perform steps S22, S23 and S24. The condition is judged.

Step S26 - incrementing the number of abnormalities by one, incrementing the counting parameter j by one, and recording the time _Tj as the time corresponding to the current image.

Step S27 - determining whether the time T _j corresponding to the current image and the interval between the first abnormal time T ₁ or the last abnormal time T _j-1 does not exceed a preset time length, and if so, the plurality of people are in contact and have a raised hand The fast motion may be continuous, rather than an accidental transient, so that step S28 is followed, and if otherwise, returning to step S21 for zeroing, the number of abnormalities is restarted. This embodiment is exemplified by T _j -T ₁ , and the preset time length is 5 seconds, but the invention is not limited thereto.

Step S28 - determining whether the current abnormal number is greater than a preset number of times, if yes, indicating that the current image captures the plurality of people at the time and the rapid movement of the raised hand is continuous, and the intense behavior of the fight may be ongoing, so step S29 is performed. If not, after returning to step S21, the next image is taken to perform the condition determination of steps S22, S23 and S24.

In step S29, it is determined that a person's intense behavior occurs and relevant output is performed, for example, the alarm of the target area is sounded, or the display screen of the coordinated security management center indicates an abnormal condition.

In summary, the preferred embodiment of the method for detecting a strong behavior of the present invention utilizes parameters such as image change degree and time, and even feature identification to perform comprehensive logical judgment, which can be accurately performed without consuming large computational resources. The abnormal situation of the personnel's fierce behavior is found, and the object of the present invention can be achieved.

The above is only the preferred embodiment of the present invention, and the scope of the invention is not limited thereto, that is, the simple equivalent changes and modifications made by the scope of the invention and the description of the invention are All remain within the scope of the invention patent.

100‧‧‧Monitoring system

1‧‧‧ camera

2‧‧‧Processing unit

S11~S20‧‧‧Steps

S141, S142‧‧‧ steps

S21~S29‧‧‧Steps

S231, S232‧‧‧ steps

S241, S242‧‧‧ steps

1 is a block diagram schematically illustrating hardware for performing a method for detecting a strong behavior of a person of the present invention; FIG. 2 is a flowchart illustrating a fierce behavior of a preferred embodiment of the method for detecting a strong behavior of a person of the present invention Detection process; Figure 3 and Figure 4 are a sequence of image images showing the damage behavior and The detection result of the normal money-collection behavior; and FIG. 5 is a flowchart illustrating the detection process of the fierce behavior of the fight in the preferred embodiment.

S11~S20‧‧‧Steps

Claims (10)

A method for detecting a person's intense behavior is performed by a monitoring system including a camera that takes an image toward a target area, and a processing unit that receives an image from the camera, the method comprising the following steps performed by the processing unit : (A) According to the image transmitted by the camera, it is recognized that a person enters the target area, and an abnormal number of times is set to 0; (B) for a current image and a previous plurality of sequence images, respectively, a momentum representative of the degree of change is calculated. (C) determining whether the momentum value is greater than a predetermined threshold, if yes, performing step (D); (D) incrementing the abnormal number by one, and recording the time corresponding to the current image; (E) determining the current image Whether the interval between the corresponding time and the first abnormal time or the last abnormal time does not exceed a preset time length, if yes, proceed to step (F), if not, return to step (A); (F) determine whether the abnormal number of times is If it is greater than a preset number of times, if yes, proceed to step (G), if otherwise, return to step (B); and (G) determine that a person's intense behavior occurs and perform relevant output. The method for detecting a strong behavior of a person as claimed in claim 1, wherein the step (C) further determines whether the current image has a hand feature, and if the two determination conditions are met, the step (D) is performed; F) is an act of judging that a person has committed damage. The method for detecting a strong behavior of a person as claimed in claim 2, wherein the hand The judgment of the feature is to use the foreground detection technology to find the foreground, and then use the contour to detect the protruding part and define it as the hand. The method for detecting a violent behavior of a person as claimed in claim 1, wherein the step (A) identifies that a plurality of persons enter the target area, and the step (C) further determines whether the contours of the person in the current image overlap. The second judgment condition is established, and step (D) is performed. The method for detecting a strong behavior of a person as claimed in claim 4, wherein the judging of the overlapping of the contours of the person is to use a component symbol algorithm to individually select the moving object to form a tracking frame, and determine whether the same pixel is simultaneously marked by the plurality of tracking frames. Incorporate, if it is, the contours overlap. The method for detecting a strong behavior of a person as claimed in claim 4, wherein the step (C) further identifies a hand region for the current image, and determines whether a horizontal edge of the contour of the hand region is more than a vertical edge, if The third judgment condition is established, and step (D) is performed. The method for detecting a strong behavior of a person according to claim 6, wherein the determination of whether the horizontal edge of the outline of the hand region is more than the vertical edge is to first block the hand region and then for each region. The hand outline in the block calculates the horizontal and vertical edges, between +45° and -45° and between -135° and +135°, which is classified as horizontal, between +45° and +135. Between ° and between -45° and -135° are classified as vertical edges, and the horizontal and vertical sides are compared. The method for detecting a strong behavior of a person according to any one of claims 1 to 7, wherein the step (B) is to compare the RGB component values of the pixels of the current image with the RGB component values of the pixels of the earlier image. And get the generation The momentum value is obtained by the value of the degree of change between the two images. The method for detecting a strong behavior of a person according to claim 8, wherein the value of the degree of change between the two images is subtracted by using the RGB component values of each pixel and compared with a pixel change threshold, which is greater than the pixel change threshold. a pixel, wherein the current image setting component value is 255, and the pixel having the pixel change threshold is less than or equal to the current image setting component value of 0, and the number of pixels whose component value is 255 is a numerator, and the total pixel number is a denominator. This represents the value of the degree of change between the two images. The method for detecting a strong behavior of a person according to claim 9, wherein the step (B) is to determine a value of the degree of change between the current image and the previous image and the previous image respectively. And add the total to get the momentum value.