CN108764124B - Crowd movement detection method and device - Google Patents
Crowd movement detection method and device Download PDFInfo
- Publication number
- CN108764124B CN108764124B CN201810513092.1A CN201810513092A CN108764124B CN 108764124 B CN108764124 B CN 108764124B CN 201810513092 A CN201810513092 A CN 201810513092A CN 108764124 B CN108764124 B CN 108764124B
- Authority
- CN
- China
- Prior art keywords
- spectrogram
- crowd
- determining
- frame image
- subsequent frame
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Abstract
The embodiment of the invention discloses a method and a device for detecting crowd movement, wherein the method comprises the following steps: acquiring a frame image in a video monitoring image; performing Fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image; determining the size of the crowd according to the high-frequency component in the spectrogram; when the crowd size is larger than a preset size threshold, extracting subsequent frame images according to a preset rule; performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image; and determining the direction and the distance of the crowd movement according to the first spectrogram and the second spectrogram. The method has the advantages of relatively simple operation, capability of effectively reducing the operation amount and higher judgment accuracy.
Description
Technical Field
The invention relates to the technical field of computer image processing, in particular to a method and a device for detecting crowd movement.
Background
The crowd information can provide important basis for management and decision making of public places, the phenomena of crowd congestion, crowd congestion and the like in public places such as traffic hubs, large-scale activity sites, large malls and the like are more and more, and hidden dangers caused by excessive crowd congestion and congestion of crowds are more and more. In real life, people can accurately know the movement ability of people, a manual method is usually adopted, manual statistics consumes a large amount of manpower and material resources, and the error rate is high.
A large number of arrangements of video monitoring provide a foundation for counting detailed information of crowds, intelligently predicting and intelligently managing public areas, and real-time and accurate crowd information of various areas can be obtained. Crowd motion may be detected based on video surveillance.
At present, people movement is detected based on video monitoring, and people image characteristics can be combined with a neural network mode generally. Specifically, people are identified from images acquired by the monitoring video through the image characteristics of the people, and the number of people in the people is identified based on the image characteristics of the people. And training the neural network based on the front frame image and the back frame image and the artificially determined motion vector. After training is finished, front and rear frame images acquired from the monitoring video are input to the trained neural network, and crowd movement is detected according to an output result of the neural network. However, the above method has the following problems: because the number scale of the crowd is detected by utilizing the gray characteristic of the image in the image acquired by the monitoring video, the number scale of the crowd can be obtained only by carrying out a large amount of calculation, and the server serving a plurality of scene monitoring is difficult to bear the corresponding calculation amount; meanwhile, a neural network needs massive monitoring images of different scenes, and the neural network without approximate scene training has certain errors on image detection results of different backgrounds.
Disclosure of Invention
In view of this, embodiments of the present invention provide a group motion detection method and apparatus, so as to solve the technical problems of a large amount of computation and a high error rate in detecting group motion by using a surveillance video in the prior art.
In a first aspect, an embodiment of the present invention provides a method for detecting motion of a crowd, including:
acquiring a frame image in a video monitoring image;
performing Fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image;
determining the size of the crowd according to the high-frequency component in the spectrogram;
when the crowd size is larger than a preset size threshold, extracting subsequent frame images according to a preset rule;
performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
and determining the direction and distance of the movement of the crowd according to the first spectrogram and the second spectrogram.
Further, the determining the size of the crowd according to the high-frequency component in the spectrogram comprises:
acquiring the number of side peaks in the high-frequency component;
and comparing the side peak quantity with a preset population scale side peak quantity to determine the population scale.
Further, the determining the crowd movement direction and distance according to the phase between the high-frequency signal in the first spectrogram and the high-frequency signal in the second spectrogram comprises:
calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases;
and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
Further, the calculating the cross-power spectral phase according to the first spectrogram and the second spectrogram includes:
cross-power spectral phase calculation using the following formula:
Further, the determining the size of the crowd according to the high-frequency component in the spectrogram further includes:
extracting subsequent frame images according to a preset rule;
performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
determining the direction and distance of the movement of the crowd according to the first spectrogram and the second spectrogram;
determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image;
and when the movement speed exceeds a preset speed threshold value, determining that the crowd size is smaller than a preset size threshold value.
Further, after determining the orientation and distance that enable crowd movement from the first and second spectrograms, the method further comprises:
and predicting the position of the crowd according to the direction and the distance of the crowd movement.
In a second aspect, an embodiment of the present invention further provides a device for detecting motion of a crowd, including:
the acquisition module is used for acquiring a frame image in the video monitoring image;
the first transformation module is used for carrying out Fourier transformation on the frame image to obtain a first spectrogram corresponding to the frame image;
the scale determining module is used for determining the scale of the crowd according to the high-frequency component in the spectrogram;
the extraction module is used for extracting the subsequent frame images according to a preset rule when the crowd size is larger than a preset size threshold;
the second transformation module is used for carrying out Fourier transformation on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
and the motion determining module is used for determining the direction and the distance of the motion of the crowd according to the first spectrogram and the second spectrogram.
Further, the scale determining module includes:
calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases;
and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
Further, the motion determination module comprises:
a subsequent extraction unit for extracting subsequent frame images according to a preset rule;
the subsequent transformation unit is used for carrying out Fourier transformation on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
the determining unit is used for determining the direction and the distance of the crowd movement according to the first spectrogram and the second spectrogram;
the speed determining unit is used for determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image;
and the size threshold determining unit is used for determining that the crowd size is smaller than a preset size threshold when the movement speed exceeds a preset speed threshold.
Further, the scale determining module includes:
an acquisition unit configured to acquire the number of side peaks in the high-frequency component;
and the comparison unit is used for comparing the side peak quantity with a preset crowd size side peak quantity to determine the size of the crowd.
According to the method and the device for detecting the crowd movement, the monitoring video image is subjected to Fourier transform to obtain the spectrogram of the image, the spectrogram is analyzed, and the crowd size is determined according to the high-frequency component in the spectrogram. And Fourier transform can be carried out on the subsequently acquired video images, and the moving direction and distance of the crowd can be determined by utilizing the spectrograms corresponding to the subsequently acquired video images and combining the phases of the high-frequency signals in the two spectrograms. The method has the advantages of relatively simple operation, capability of effectively reducing the operation amount and higher judgment accuracy.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments made with reference to the following drawings:
fig. 1 is a schematic flow chart of a method for detecting motion of a human group according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a method for detecting motion of a crowd according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a group motion detection apparatus according to a third embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a method for detecting a crowd movement according to an embodiment of the present invention, where the method is applicable to a situation of detecting a crowd movement in a video surveillance image, and the method can be executed by a crowd movement detection device and can be integrated in a video surveillance server, and specifically includes the following steps:
and S110, acquiring a frame image in the video monitoring image.
Currently, surveillance videos may be collected by video surveillance equipment. Typically, the surveillance video is composed of frame images, and for example, a1 second video may be composed of 30-60 frame images. Therefore, frame images can be extracted from the surveillance video. For example, according to the preset time, the frame image corresponding to the current time is extracted from the video, or the frame image corresponding to the specified time may be extracted from the video. Further, the acquired frame image may be used as the first frame image.
And S120, performing Fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image.
The spectrogram is short for frequency spectral density and is a distribution curve of frequency. The image can be viewed as a two-dimensional signal, with a two-dimensional fourier transform being a superposition of the fourier transforms of the one-dimensional fourier transform on each of the row and column scan lines. Before fourier transformation, where an image is a set of a series of points obtained by sampling in a continuous space (real space), we used a two-dimensional matrix to represent each point in the space, the image could be represented by z ═ f (x, y). From the physical effect, the fourier transform is to convert the image from the spatial domain to the frequency domain, and the inverse transform is to convert the image from the frequency domain to the spatial domain. Namely, the gray distribution function of the image is transformed into the frequency distribution function of the image, and the inverse fourier transform is the transformation of the frequency distribution function of the image into the gray distribution function.
Illustratively, the two-dimensional fourier transform of an image can be implemented using the following formula:
where F (x, y) represents a matrix of size M x N, where x-0, 1,2, M-1 and y-0, 1,2, N-1, and F (u, v) represents the fourier transform of F (x, y). It can be converted to a trigonometric representation method where u and v can be used to determine the frequency of the sine and cosine. The coordinate system of F (u, v) is called the frequency domain, and the M x N matrix defined by u-0, 1,2, M-1 and v-0, 1,2, N-1 is often called the frequency domain matrix.
The image can be subjected to two-dimensional Fourier transform by using the formula, and a spectrogram of the frame image is generated according to the transformation. And the spectrogram can be used as a first spectrogram.
And S130, determining the size of the crowd according to the high-frequency component in the spectrogram.
The low frequency component in the image may be a place where the image intensity transition is gentle, and the high frequency component in the image may be a place where the image intensity (brightness/grayscale) changes drastically. Illustratively, the image intensity may be an image brightness or a gray scale. In surveillance video images, especially in indoor surveillance video images, background images are generally relatively similar due to brightness or gray scale, and are generally low frequency components. The peripheral outline of the crowd in the image can be high-frequency components due to the large difference with the background image.
The low frequency component is a comprehensive measure of the intensity of the whole image. If the intensity of each position of one image is equal, the image only has low-frequency components, and only has one main peak on the spectrogram of the image, and the main peak is positioned at the position with the frequency of zero. If the intensity of each position of one image changes violently, the image has not only low-frequency components but also a plurality of high-frequency components, and from the spectrum of the image, not only one main peak but also a plurality of side peaks exist. If the population is large, the profile is relatively large, and the high-frequency components are more various. Correspondingly, if the population is small, the population size is small, and the types of high-frequency components in the spectrogram are small. Therefore, the size of the crowd can be determined according to the high-frequency component in the spectrogram corresponding to the video monitoring image.
For example, the determining the size of the crowd according to the high frequency component in the spectrogram may include: acquiring the number of side peaks in the high-frequency component; and comparing the side peak quantity with a preset population scale side peak quantity to determine the population scale. Since the main peak exists in the spectrogram corresponding to each type of image, the number of people needs to be determined according to the number of the side peaks. And further determining the size of the population.
In this embodiment, the corresponding relationship between the crowd size and the number of side peaks may be preset. For example, the number of side peaks corresponding to a population scale of 5 or less may be preset to be 13 or less; the number of the side peaks corresponding to the population scale of 5-10 people is 27-50, etc. According to the preset corresponding relation, the number of the side peaks of the crowd scale corresponding to different crowd scales can be determined and compared to determine the crowd scale.
And S140, when the crowd size is larger than a preset size threshold, extracting the subsequent frame image according to a preset rule.
If the crowd size is smaller than a certain size, for example, 5 people away, no vicious events such as congestion and treading will occur. Therefore, motion detection of the population is not required. Only when the crowd size is larger than a certain size, the crowd carries out motion detection. In the embodiment, the obtained crowd size is compared with a preset size threshold, and when the crowd size is determined to be larger than the preset size threshold, the movement of the crowd is detected.
Subsequent images are required to detect the motion of the population. Therefore, in the present embodiment, the subsequent frame image can be extracted according to a preset rule. The preset rule may be that the images are extracted at time intervals or image intervals. For example, a frame image 2 seconds later may be extracted as an image of a subsequent frame, or a 60 th frame image after an image corresponding to the first spectrogram may be extracted as an image of a subsequent frame.
S150, carrying out Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image.
Illustratively, the subsequent frame image may still be fourier transformed in the manner described above. Illustratively, the following equation may still be used to implement a two-dimensional fourier transform of an image:
where F (x, y) represents a matrix of size M x N, where x-0, 1,2, M-1 and y-0, 1,2, N-1, and F (u, v) represents the fourier transform of F (x, y). It can be converted to a trigonometric representation method where u and v can be used to determine the frequency of the sine and cosine. The coordinate system of F (u, v) is called the frequency domain, and the M x N matrix defined by u-0, 1,2, M-1 and v-0, 1,2, N-1 is often called the frequency domain matrix.
The image can be subjected to two-dimensional Fourier transform by using the formula, and a spectrogram of a subsequent frame image is generated according to the transform.
And S160, determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram.
The spectrogram has rich information. The information includes not only high frequency components and low frequency components, but also amplitude, phase and other information. Since the object displacement in the time domain corresponds to the phase angle in the frequency domain, the detection of the object motion can be achieved from the phases in the multiple spectrograms of interest.
For example, the determining the orientation and distance of the crowd movement according to the first spectrogram and the second spectrogram may include: calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases; and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
Because the video image frames corresponding to the first spectrogram and the second spectrogram have an incidence relation, namely, part of objects in the images, namely people, generate corresponding displacement, the images of the first spectrogram and the second spectrogram can be expressed by an F (x, y) ═ F (x + ax, y + by) formula, Fourier change is simultaneously carried out on two sides of the formula, a pulse function is obtained by solving the Fourier inverse transformation of a cross-power spectrum, and then a coordinate corresponding to a peak point of the function is searched, so that a displacement vector to be obtained can be obtained. In actual operation, the inverse transformation of the cross-power spectral phase of two images always contains a correlation peak value representing the registration point of the two images and some non-correlation peak values, and the correlation peak value directly reflects the displacement vector between the two images, namely the moving distance and the moving direction of people.
For example, the calculating the cross-power spectral phase according to the first spectrogram and the second spectrogram may include:
cross-power spectral phase calculation using the following formula:
According to the method and the device for detecting the crowd movement, the spectrogram of the monitoring video image is obtained by performing Fourier transform on the image, the spectrogram is analyzed, and the crowd size is determined according to the high-frequency component in the spectrogram. And Fourier transform can be carried out on the subsequently acquired video images, and the moving direction and distance of the crowd can be determined by utilizing the spectrograms corresponding to the subsequently acquired video images and combining the phases of the high-frequency signals in the two spectrograms. The method has the advantages of relatively simple operation, capability of effectively reducing the operation amount and higher judgment accuracy.
Example two
Fig. 2 is a schematic flow chart of a method for detecting a crowd movement according to a second embodiment of the present invention. In this embodiment, the determining the size of the crowd according to the high-frequency component in the spectrogram is specifically optimized as follows: extracting subsequent frame images according to a preset rule; performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image; determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram; determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image; and when the movement speed exceeds a preset speed threshold value, determining that the crowd size is smaller than a preset size threshold value.
Correspondingly, the method for screening live content provided by the embodiment specifically includes:
s210, acquiring a frame image in the video monitoring image.
S220, performing Fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image.
And S230, extracting the subsequent frame image according to a preset rule, and performing Fourier transform on the subsequent frame image to obtain a comparison spectrogram of the subsequent frame image.
In the video acquisition process, if the crowd is closer to a video acquisition device, such as a camera, the occupied range of the crowd in the acquired frame image is larger. Correspondingly, if only one person is at a short distance from the video acquisition device, the more high-frequency vectors in the spectrogram corresponding to the acquired frame image. A large-scale misjudgment of the population may occur. In order to avoid the above misjudgment, in this embodiment, the crowd size may be judged by using the subsequent frame image. Illustratively, the subsequent frame image may be extracted according to a preset rule. The preset rule may be that the images are extracted at time intervals or image intervals. And Fourier transform is carried out on the subsequently extracted image to generate a spectrogram of a subsequent frame image, namely a comparison spectrogram for comparison.
S240, determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram, and determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image.
For example, the direction and distance of the movement of the crowd can be determined by the method described in the above embodiment, and the time difference between the subsequent frame image and the frame image can be determined by using the time difference. For the images acquired according to the frame number rule, the interval time between the images and the monitoring video acquisition device can be determined according to the number of the image frames acquired per second set by the monitoring video acquisition device, and the movement speed of the crowd in the images can be determined based on the displacement and the time.
And S250, when the movement speed exceeds a preset speed threshold, determining that the crowd size is smaller than a preset size threshold.
The monitoring video acquisition device has a smaller shooting range for the closer person, so that the time length of the closer person appearing in the image is shorter. The range of image acquisition is typically moved in a short time. Which generally appears to be faster in speed performance. Therefore, it is possible to determine whether the size of the crowd is erroneously determined due to the too close distance based on the speed.
For example, a speed threshold is set, the movement speed determined in the above manner is compared with the speed threshold, and when the movement speed exceeds the speed threshold, the size of the crowd is judged to be misjudged due to too close distance. Wherein the speed threshold may be empirically chosen. For example, the image capturing range may be determined according to the optimal distance of the video monitoring, and the speed of the normal movement of the crowd in the image capturing range may be used as the speed threshold.
And S260, when the crowd size is larger than a preset size threshold, extracting the subsequent frame image according to a preset rule.
S270, carrying out Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image.
And S280, determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram.
In this embodiment, the determining the size of the crowd according to the high frequency component in the spectrogram is specifically optimized as follows: extracting subsequent frame images according to a preset rule; performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image; determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram; determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image; and when the movement speed exceeds a preset speed threshold value, determining that the crowd size is smaller than a preset size threshold value. The staff scale misjudgment caused by the fact that the distance between the crowd and the video acquisition device is too close can be effectively avoided, and the accuracy of the crowd scale judgment is improved.
In a preferred implementation of this embodiment, after determining the orientation and distance capable of crowd movement from the first and second spectrograms, the method may further include the following steps: and predicting the position of the crowd according to the direction and the distance of the crowd movement. After the crowd scale is judged and the crowd movement is detected, the crowd movement needs to be predicted so as to realize risk prediction on whether congestion treading and the like possibly occur subsequently. Illustratively, the prediction may be performed by continuously acquiring multiple frames of images, and obtaining a motion line of the human group based on the multiple frames of images, where the motion line includes: direction and displacement. And establishing an svm classifier based on the obtained direction and displacement, adjusting the weight parameters in the classifier according to the direction and displacement corresponding to the subsequent multi-frame image of the classifier, and predicting the crowd movement according to the adjusted weight parameters. Exemplary y1 is a1x1+ a2x2+ … … + anxn, and y2 is b1y1+ b2y2+ … … + bnyn, where xn is the displacement represented by the nth image, y1 is the displacement prediction amount, an is the motion weight parameter, yn is the motion direction represented by the nth image, y2 is the orientation prediction amount, and bn is the orientation weight parameter c and is a constant. The prediction of crowd movement can be effectively realized, and then the risk early warning can be provided accordingly, and various risks caused by crowd congestion are effectively reduced.
EXAMPLE III
Fig. 3 is a schematic structural diagram of a device for detecting motion of a human group according to a third embodiment of the present invention, as shown in fig. 3, the device includes:
an obtaining module 310, configured to obtain a frame image in a video surveillance image;
a first transform module 320, configured to perform fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image;
a size determining module 330, configured to determine the size of the crowd according to the high-frequency component in the spectrogram;
the extracting module 340 is configured to extract subsequent frame images according to a preset rule when the crowd size is larger than a preset size threshold;
a second transform module 350, configured to perform fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
and the motion determining module 360 is used for determining the direction and the distance of the motion of the crowd according to the first spectrogram and the second spectrogram.
The device for detecting crowd movement provided by this embodiment obtains a spectrogram of a monitoring video image by performing fourier transform on the image, analyzes the spectrogram, and determines the crowd size according to a high-frequency component in the spectrogram. And Fourier transform can be carried out on the subsequently acquired video images, and the moving direction and distance of the crowd can be determined by utilizing the spectrograms corresponding to the subsequently acquired video images and combining the phases of the high-frequency signals in the two spectrograms. The method has the advantages of relatively simple operation, capability of effectively reducing the operation amount and higher judgment accuracy.
On the basis of the foregoing embodiments, the scale determining module includes:
calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases;
and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
On the basis of the foregoing embodiments, the scale determining module includes:
an acquisition unit configured to acquire the number of side peaks in the high-frequency component;
and the comparison unit is used for comparing the side peak quantity with a preset crowd size side peak quantity to determine the size of the crowd.
On the basis of the foregoing embodiments, the motion determination module includes:
a subsequent extraction unit for extracting subsequent frame images according to a preset rule;
the subsequent transformation unit is used for carrying out Fourier transformation on the subsequent frame image to obtain a comparison spectrogram of the subsequent frame image;
the determining unit is used for determining the direction and the distance of the crowd movement according to the first spectrogram and the second spectrogram;
the speed determining unit is used for determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image;
and the size threshold determining unit is used for determining that the crowd size is smaller than a preset size threshold when the movement speed exceeds a preset speed threshold.
The device for detecting the crowd movement provided by the embodiment of the invention can execute the method for detecting the crowd movement provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.
Those of ordinary skill in the art will understand that: all or a portion of the steps of implementing the above-described method embodiments may be performed by hardware associated with program instructions. The program may be stored in a computer-readable storage medium. When executed, the program performs steps comprising the method embodiments described above; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. A method for detecting motion of a population, comprising:
acquiring a frame image in a video monitoring image;
performing Fourier transform on the frame image to obtain a first spectrogram corresponding to the frame image;
determining the size of the crowd according to the high-frequency component in the spectrogram;
when the crowd size is larger than a preset size threshold, extracting subsequent frame images according to a preset rule;
performing Fourier transform on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
determining the direction and distance of the crowd movement according to the first spectrogram and the second spectrogram;
the determining the size of the crowd according to the high-frequency component in the spectrogram further comprises:
extracting subsequent frame images according to a preset rule;
performing Fourier transform on the subsequent frame image to obtain a comparison spectrogram of the subsequent frame image, wherein the comparison spectrogram is a spectrogram of the subsequent frame image;
determining the direction and distance of the crowd movement according to the first spectrogram and the comparison spectrogram;
determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image;
and when the movement speed exceeds a preset speed threshold value, determining that the crowd size is smaller than a preset size threshold value.
2. The method of claim 1, wherein determining the size of the population based on the high frequency components in the spectrogram comprises:
acquiring the number of side peaks in the high-frequency component;
and comparing the side peak quantity with a preset population scale side peak quantity to determine the population scale.
3. The method of claim 1, wherein determining a crowd motion azimuth and distance from a phase between a high frequency signal in the first spectrogram and a high frequency signal in a second spectrogram comprises:
calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases;
and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
4. The method of claim 3, wherein calculating the cross-power spectral phase from the first and second spectrograms comprises:
cross-power spectral phase calculation using the following formula:
5. The method of claim 1, after determining the orientation and distance at which crowd motion is possible from the first and second spectrograms, the method further comprising:
and predicting the position of the crowd according to the direction and the distance of the crowd movement.
6. A device for detecting motion of a person, comprising:
the acquisition module is used for acquiring a frame image in the video monitoring image;
the first transformation module is used for carrying out Fourier transformation on the frame image to obtain a first spectrogram corresponding to the frame image;
the scale determining module is used for determining the scale of the crowd according to the high-frequency component in the spectrogram;
the extraction module is used for extracting the subsequent frame images according to a preset rule when the crowd size is larger than a preset size threshold;
the second transformation module is used for carrying out Fourier transformation on the subsequent frame image to obtain a second spectrogram of the subsequent frame image;
the motion determination module is used for determining the direction and the distance of the motion of the crowd according to the first spectrogram and the second spectrogram;
the motion determination module comprising:
a subsequent extraction unit for extracting subsequent frame images according to a preset rule;
the subsequent transformation unit is used for carrying out Fourier transformation on the subsequent frame image to obtain a comparison spectrogram of the subsequent frame image, and the comparison spectrogram is a spectrogram of the subsequent frame image;
the determining unit is used for determining the direction and the distance of the crowd movement according to the first spectrogram and the second spectrogram;
the speed determining unit is used for determining the movement speed of the crowd according to the time interval between the subsequent frame image and the current frame image;
and the size threshold determining unit is used for determining that the crowd size is smaller than a preset size threshold when the movement speed exceeds a preset speed threshold.
7. The apparatus of claim 6, wherein the motion determination module comprises:
calculating cross-power spectrum phases according to the first spectrogram and the second spectrogram, and calculating two-dimensional inverse discrete Fourier transform of the cross-power spectrum phases;
and determining the position of the maximum peak value of the phase correlation function based on the two-dimensional inverse discrete Fourier transform, and determining the displacement and the direction of the movement of the crowd according to the position of the maximum peak value.
8. The apparatus of claim 6, wherein the scale determining module comprises:
an acquisition unit configured to acquire the number of side peaks in the high-frequency component;
and the comparison unit is used for comparing the side peak quantity with a preset crowd size side peak quantity to determine the size of the crowd.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810513092.1A CN108764124B (en) | 2018-05-25 | 2018-05-25 | Crowd movement detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810513092.1A CN108764124B (en) | 2018-05-25 | 2018-05-25 | Crowd movement detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764124A CN108764124A (en) | 2018-11-06 |
| CN108764124B true CN108764124B (en) | 2022-04-26 |
Family
ID=64005638
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810513092.1A Active CN108764124B (en) | 2018-05-25 | 2018-05-25 | Crowd movement detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764124B (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05233816A (en) * | 1992-02-24 | 1993-09-10 | Nippon Telegr & Teleph Corp <Ntt> | Moving image feature detecting method |
| CN103985140B (en) * | 2014-05-22 | 2015-06-03 | 中国人民解放军国防科学技术大学 | Anti-noise moving object detecting method based on frequency domain processing |
| CN104574352A (en) * | 2014-09-02 | 2015-04-29 | 重庆大学 | Crowd density grade classification method based on foreground image |
| CN104268528B (en) * | 2014-09-28 | 2017-10-17 | 中智科创机器人有限公司 | A kind of crowd massing method for detecting area and device |
| CN106534614A (en) * | 2015-09-10 | 2017-03-22 | 南京理工大学 | Rapid movement compensation method of moving target detection under mobile camera |
-
2018
- 2018-05-25 CN CN201810513092.1A patent/CN108764124B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764124A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN102968799B (en) | Integral image-based quick ACCA-CFAR SAR (Automatic Censored Cell Averaging-Constant False Alarm Rate Synthetic Aperture Radar) image target detection method | |
| Huang et al. | Automatic moving object extraction through a real-world variable-bandwidth network for traffic monitoring systems | |
| CN111951313B (en) | Image registration method, device, equipment and medium | |
| KR20140045854A (en) | Method and apparatus for monitoring video for estimating gradient of single object | |
| KR101348680B1 (en) | Target acquisition method for video tracker, and target acquisition apparatus using the same | |
| EP3332388B1 (en) | Temporally adaptive processing | |
| CN116309781B (en) | Cross-modal fusion-based underwater visual target ranging method and device | |
| CN111353496B (en) | Real-time detection method for infrared dim targets | |
| Niu et al. | Moving point target detection based on higher order statistics in very low SNR | |
| CN109684986B (en) | Vehicle analysis method and system based on vehicle detection and tracking | |
| JP2018531532A6 (en) | Adaptive processing in time | |
| Ma et al. | A lightweight neural network for crowd analysis of images with congested scenes | |
| CN108764124B (en) | Crowd movement detection method and device | |
| CN107767401B (en) | Infrared target real-time tracking method and device based on nuclear correlation filtering | |
| CN106709516B (en) | Fuzzy image detection method based on naive Bayes method | |
| Zhao et al. | Shooting for smarter motion detection in cameras: improvements for the visual background extractor algorithm using optical flow | |
| Lagorio et al. | Automatic detection of adverse weather conditions in traffic scenes | |
| CN105631419B (en) | Face identification method and device | |
| CN114757848A (en) | Camera jelly effect detection method and system based on artificial intelligence | |
| Okarma et al. | A fast image analysis technique for the line tracking robots | |
| Almalki et al. | Characterizing scattered occlusions for effective dense-mode crowd counting | |
| Mohana et al. | Vehicle detection and counting by using real time traffic flux through differential technique and performance evaluation | |
| CN113496136A (en) | Unmanned aerial vehicle and image identification method thereof | |
| KR20150120805A (en) | Method and system for detecting human in range image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |