Intelligent control integration method and system based on cloud computing

Abstract

The invention relates to the technical field of image data processing, in particular to an intelligent control integration method and system based on cloud computing, and the method comprises the steps: obtaining each frame of target monitoring image of a to-be-detected community in a preset time period; feature point matching is carried out on character areas in all adjacent frames of target monitoring images; determining an exercise performance index corresponding to each matching point group; determining the non-possibility of the same person between every two matching point groups between the same two adjacent frames of target monitoring images; screening out a person area representing the same person to be detected; determining an abnormal behavior performance index corresponding to each person to be detected; and if the abnormal behavior performance index corresponding to the to-be-detected person is greater than a preset abnormal threshold, determining that the to-be-detected person has an abnormal behavior, and controlling to send abnormal prompt information. According to the invention, the target monitoring image is analyzed to identify the motion trail of the person in the community to be detected, and then the abnormal behavior performance is analyzed, so that the timeliness of identifying the abnormal person is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and particularly relates to an intelligent control integration method and system based on cloud computing. Background Art

[0002] With the rapid development of information technology, cloud computing, as an emerging computing model, has gradually penetrated into multiple industries. Cloud computing provides users with efficient, convenient, and scalable IT (Information Technology) services by integrating computing resources, storage resources, and application resources. And intelligent control, as an important branch in the field of automation, is also constantly developing and improving. Combining cloud computing with intelligent control can give full play to the advantages of both, realizing more efficient and intelligent system control, which can be used to assist community security.

[0003] In most communities, community security is mainly ensured by means such as manual patrol and manual video surveillance. Although these means can ensure community safety to a certain extent, they are relatively dependent on the subjective judgment of community security guards, and community security guards are often difficult to maintain a high level of vigilance in real time, especially when the human body is prone to fatigue at night, resulting in poor timeliness of abnormal behavior identification, and thus reducing the community's security prevention ability. Summary of the Invention

[0004] In order to solve the technical problem of poor timeliness of abnormal behavior identification, the present invention proposes an intelligent control integration method and system based on cloud computing.

[0005] In a first aspect, the present invention provides an intelligent control integration method based on cloud computing, and the method includes:

[0006] Obtain each frame of target monitoring image of the community to be detected within a preset time period, and identify the human area from each frame of target monitoring image;

[0007] Perform feature point matching on the human areas in all adjacent frames of target monitoring images to obtain a set of matching points;

[0008] Determine the motion performance index corresponding to each set of matching points according to the motion distance, gray scale distribution difference, and moving direction difference between the feature points in each set of matching points;

[0009] Determine the non - possibility of the same person between each two sets of matching points in the same adjacent two frames of target monitoring images according to the distance between the feature points in each two sets of matching points in the same adjacent two frames of target monitoring images and the difference between the motion performance indexes corresponding to these two sets of matching points;

[0010] According to the improbability of the same person among different matching point groups of all target monitoring images, filter out the person regions representing the same person to be detected;

[0011] According to the motion trajectory situation among all person regions representing each person to be detected, determine the abnormal behavior performance index corresponding to each person to be detected;

[0012] If the abnormal behavior performance index corresponding to the person to be detected is greater than the preset abnormal threshold, it is determined that the person to be detected has an abnormal behavior, and control is sent to send an abnormal prompt message.

[0013] Combined with the above first aspect, in a possible implementation manner, the determining the motion performance index corresponding to each matching point group according to the motion distance, gray-scale distribution difference, and moving direction difference between feature points in each matching point group includes:

[0014] Determine the distance between the pixel coordinates corresponding to the feature points in each matching point group as the motion distance corresponding to each matching point group;

[0015] According to the gray-scale difference between each feature point in each matching point group and its corresponding preset neighborhood, determine the gray-scale distribution difference corresponding to each matching point group;

[0016] According to the pixel coordinates corresponding to the feature points in each matching point group, determine the moving direction difference corresponding to each matching point group;

[0017] According to the motion distance, gray-scale distribution difference, and moving direction difference corresponding to each matching point group, determine the motion performance index corresponding to each matching point group, where the motion distance, gray-scale distribution difference, and moving direction difference are all positively correlated with the motion performance index.

[0018] Combined with the above first aspect, in a possible implementation manner, the determining the moving direction difference corresponding to each matching point group according to the pixel coordinates corresponding to the feature points in each matching point group includes:

[0019] Determine any one matching point group as the marked matching point group, and determine two feature points in the marked matching point group as the first reference point and the second reference point respectively;

[0020] Determine the target monitoring image to which the first reference point belongs as the first monitoring image, and determine the coordinate origin of the first monitoring image as the first origin;

[0021] Filter out the pixel points with the same position as the second reference point from the first monitoring image as the temporary points;

[0022] Connect the first reference point and the first origin to obtain the first straight line, and connect the temporary point and the first origin to obtain the second straight line;

[0023] Determine the angle between the first straight line and the second straight line as the movement direction difference corresponding to the marker matching point group.

[0024] Combined with the above first aspect, in a possible implementation manner, the formula for the gray-scale distribution difference corresponding to the matching point group is:

[0025] where, ΔG t，t+1，i is the gray-scale distribution difference corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; t is the frame number of the target monitoring image; i is the serial number of the matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; || is the absolute value function; N is the number of pixel points in the preset neighborhood; j is the serial number of the pixel point in the preset neighborhood; G t，i is the gray-scale value corresponding to the feature point belonging to the t-th frame target monitoring image in the i-th matching point group; G t+1，i is the gray-scale value corresponding to the feature point belonging to the (t + 1)-th frame target monitoring image in the i-th matching point group; G t，ij is the gray-scale value corresponding to the j-th pixel point in the preset neighborhood corresponding to the feature point belonging to the t-th frame target monitoring image in the i-th matching point group; G t+1，ij is the gray-scale value corresponding to the j-th pixel point in the preset neighborhood corresponding to the feature point belonging to the (t + 1)-th frame target monitoring image in the i-th matching point group.

[0026] Combined with the above first aspect, in a possible implementation manner, the formula for the non - possibility of the same person between any two matching point groups between two adjacent target monitoring images is:

[0027] ρ t，t+1，i，i+1 =|SP t，t+1，i -SP t，t+1，i+1 |×|d t，i，i+1 -d t+1，i，i+1 |; where, ρ t，t+1，i，i+1 is the non - possibility of the same person between the i-th matching point group and the (i + 1)-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; t is the frame number of the target monitoring image; i is the serial number of the matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; || is the absolute value function; SP t，t+1，i is the movement performance index corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; SP t，t+1，i+1 is the movement performance index corresponding to the (i + 1)-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image; d t，i，i+1is the distance between the feature points of the i-th matching point group and the (i + 1)-th matching point group in the target monitoring image of the t-th frame; d t+1，i，i+1 is the distance between the feature points of the i-th matching point group and the (i + 1)-th matching point group in the target monitoring image of the (t + 1)-th frame.

[0028] Combined with the above first aspect, in a possible implementation manner, the screening of the person regions representing the same person to be detected according to the non-possibility of the same person between different matching point groups in all target monitoring images includes:

[0029] Determine any two adjacent target monitoring images as the first marked image and the second marked image respectively;

[0030] Form a marked group set with all the matching point groups between the first marked image and the second marked image;

[0031] Form a marked category with the matching point groups in the marked group set whose normalized value of the non-possibility of the same person is less than or equal to a preset same-possibility threshold;

[0032] Determine the person region in the first marked image as the first marked region, and determine the person region in the second marked image as the second marked region;

[0033] If the number of matching point groups belonging to the same marked category between the first marked region and the second marked region is greater than a preset quantity threshold, it is determined that the first marked region and the second marked region represent the same person to be detected.

[0034] Combined with the above first aspect, in a possible implementation manner, the determination of the abnormal behavior performance index corresponding to each person to be detected according to the motion trajectory situation between all the person regions representing each person to be detected includes:

[0035] Determine the center point of each person region as the person representative point, and determine any one person to be detected as the marked detected person;

[0036] Connect each adjacent pair of person representative points corresponding to the marked detected person to construct a target vector between each adjacent pair of person representative points corresponding to the marked detected person;

[0037] Determine the abnormal behavior performance index corresponding to the marked detected person according to all the target vectors corresponding to the marked detected person.

[0038] Combined with the above first aspect, in a possible implementation manner, the determination of the abnormal behavior performance index corresponding to the marked detected person according to all the target vectors corresponding to the marked detected person includes:

[0039] Divide the target vectors with continuously the same direction among all the target vectors corresponding to the marker detector into the same vector group;

[0040] Connect the first human representative point and the last human representative point corresponding to the marker detector to construct the overall displacement vector corresponding to the marker detector;

[0041] Determine the abnormal behavior performance index corresponding to the marker detector according to the number of vector groups and the modulus of the overall displacement vector.

[0042] Combined with the above first aspect, in a possible implementation manner, the formula corresponding to the abnormal behavior performance index of the marker detector is:

[0043] where β is the abnormal behavior performance index corresponding to the marker detector; norm() is the normalization function; T is the time duration between the acquisition time corresponding to the first human area and the acquisition time corresponding to the last human area of the marker detector; H is the number of vector groups; M is the modulus of the overall displacement vector corresponding to the marker detector.

[0044] In the second aspect, the present invention provides an intelligent control integration system based on cloud computing, and the system includes:

[0045] An image acquisition and recognition module, configured to acquire each frame of target monitoring image of the community to be detected within a preset time period, and recognize the human area from each frame of target monitoring image;

[0046] A feature point matching module, configured to perform feature point matching on the human areas in all adjacent-frame target monitoring images to obtain a set of matching points;

[0047] A motion performance index determination module, configured to determine the motion performance index corresponding to each set of matching points according to the motion distance, gray distribution difference, and moving direction difference between the feature points in each set of matching points;

[0048] A same-person improbability determination module, configured to determine the improbability of the same person between each two sets of matching points between the same adjacent two-frame target monitoring images according to the distance between the feature points in each two sets of matching points between the same adjacent two-frame target monitoring images and the difference between the motion performance indexes corresponding to these two sets of matching points;

[0049] A region screening module, configured to screen out the human areas representing the same person to be detected according to the improbability of the same person between different sets of matching points between all target monitoring images;

[0050] An abnormal behavior performance index determination module, configured to determine the abnormal behavior performance index corresponding to each person to be detected according to the motion trajectory situation between all the human areas representing each person to be detected.

[0051] An information control sending module, configured to determine that there is an abnormal behavior of a person to be detected if the abnormal behavior performance index corresponding to the person to be detected is greater than a preset abnormal threshold, and control the sending of an abnormal prompt message.

[0052] In a third aspect, a server is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the method in the first aspect or any possible implementation manner of the first aspect.

[0053] In a fourth aspect, a computer program product is provided, including: computer program code, when the computer program code runs on a computer, enabling the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0054] In a fifth aspect, a computer-readable storage medium is provided, storing computer program code, when the computer program code runs on a computer, enabling the computer to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0055] The present invention has the following beneficial effects:

[0056] An intelligent control integration method based on cloud computing of the present invention analyzes a target monitoring image to identify the movement trajectory of a person in a community to be detected, and then analyzes their abnormal behavior performance, solving the technical problem of poor timeliness in identifying abnormal behavior persons, thereby improving the timeliness of identifying abnormal behavior persons. Compared with relying on manual identification of abnormal behavior persons, the present invention analyzes multiple frames of target monitoring images of the community to be detected, relatively objectively quantifies multiple characteristics related to abnormal behavior persons, such as movement performance indicators, non-possibility of the same person, and abnormal behavior performance indicators, thereby realizing the determination of abnormal behavior, and further realizing the identification of abnormal behavior persons, and improving the timeliness of identifying abnormal behavior persons. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0058] Figure 1 It is a flowchart of an intelligent control integration method based on cloud computing of the present invention;

[0059] Figure 2 It is a schematic diagram of the composition structure of an intelligent control integration system based on cloud computing according to the present invention;

[0060] Figure 3 It is a schematic diagram of the structure of a computer device according to the present invention. Specific embodiments

[0061] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following specifically describes the specific embodiments, structures, features and effects of the technical solutions proposed according to the present invention in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0063] Refer to Figure 1 , which shows the flow of some embodiments of an intelligent control integration method based on cloud computing according to the present invention. The intelligent control integration method based on cloud computing includes the following steps:

[0064] Step S1, obtain each frame of target monitoring image of the community to be detected within a preset time period, and identify the human area from each frame of target monitoring image.

[0065] Among them, the community to be detected can be an area where abnormal behavior people need to be monitored. For example, the community to be detected can be a community. The preset time period can be a pre-set time period with the current moment as the end moment. The duration corresponding to the preset time period can be 30 seconds. The target monitoring image can be the monitoring image of the community to be detected after image preprocessing. Image preprocessing can include, but is not limited to: image enhancement and image denoising. The human area can represent the people in the community to be detected.

[0066] As an example, this step can include the following steps:

[0067] In the first step, each frame of monitoring image of the community to be detected can be collected through a camera, and each frame of monitoring image can be enhanced through a histogram equalization algorithm to achieve image preprocessing of the monitoring image, and the enhanced monitoring image can be used as the target monitoring image.

[0068] It should be noted that in actual situations, when the area of the community to be detected is large, it may not be possible to capture the entire community with a single camera. In this case, multiple cameras can be used to capture different parts of the community to be detected, and multiple images collected at the same moment can be stitched together to form a surveillance image covering the community to be detected.

[0069] In the second step, the area representing a person can be identified from each frame of the target surveillance image through a neural network model or threshold segmentation technology as the human area.

[0070] In step S2, feature point matching is performed on the human areas in all adjacent frames of the target surveillance images to obtain a set of matching points.

[0071] As an example, through the SIFT (Scale-Invariant Feature Transform) feature matching algorithm, feature points can be extracted from each human area in each frame of the target surveillance image, and the feature points in every two adjacent frames of the target surveillance images are matched, and two mutually matching feature points form a set of matching points.

[0072] In step S3, according to the motion distance, gray distribution difference, and moving direction difference between the feature points in each set of matching points, the motion performance index corresponding to each set of matching points is determined.

[0073] As an example, this step may include the following steps:

[0074] In the first step, the distance between the pixel coordinates corresponding to the feature points in each set of matching points is determined as the motion distance corresponding to each set of matching points.

[0075] Among them, the pixel coordinates corresponding to the feature points can be the coordinates of the feature points in the pixel coordinate system. The pixel coordinate system can be a coordinate system with the upper left corner of the target surveillance image as the origin, the horizontal right as the horizontal axis, and the vertical down as the vertical axis.

[0076] For example, the formula for determining the motion distance corresponding to the set of matching points can be:

[0077] Among them, ΔS t,t+1,i is the motion distance corresponding to the i-th set of matching points between the t-th frame of the target surveillance image and the (t + 1)-th frame of the target surveillance image. t is the frame number of the target surveillance image. i is the serial number of the set of matching points between the t-th frame of the target surveillance image and the (t + 1)-th frame of the target surveillance image. x t，i is the abscissa of the pixel coordinates corresponding to the feature point belonging to the t-th frame of the target surveillance image in the i-th set of matching points. x t+1，i is the abscissa of the pixel coordinates corresponding to the feature point belonging to the (t + 1)-th frame of the target surveillance image in the i-th set of matching points. y t，iis the ordinate of the pixel coordinates corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the t-th frame. y t+1，i is the ordinate of the pixel coordinates corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the (t + 1)-th frame.

[0078] It should be noted that when ΔS t，t+1，i is larger, it often indicates that the relative unit movement distance of the part represented by the i-th matching point group is relatively larger.

[0079] Second step, according to the gray-scale differences between each feature point in each matching point group and its corresponding preset neighborhood, determine the gray-scale distribution difference corresponding to each matching point group.

[0080] For example, the formula for determining the gray-scale distribution difference corresponding to the matching point group can be:

[0081] where ΔG t，t+1，i is the gray-scale distribution difference corresponding to the i-th matching point group between the target monitoring image of the t-th frame and the target monitoring image of the (t + 1)-th frame. t is the frame number of the target monitoring image. i is the serial number of the matching point group between the target monitoring image of the t-th frame and the target monitoring image of the (t + 1)-th frame. || is the absolute value function. N is the number of pixel points in the preset neighborhood. j is the serial number of the pixel points in the preset neighborhood. G t，i is the gray-scale value corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the t-th frame. G t+1，i is the gray-scale value corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the (t + 1)-th frame. G t，ij is the gray-scale value corresponding to the j-th pixel point in the preset neighborhood corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the t-th frame. G t+1，ij is the gray-scale value corresponding to the j-th pixel point in the preset neighborhood corresponding to the feature points in the i-th matching point group that belong to the target monitoring image of the (t + 1)-th frame.

[0082] It should be noted that can represent the gray-scale distribution around the feature points in the i-th matching point group that belong to the target monitoring image of the t-th frame. can represent the gray-scale distribution around the feature points in the i-th matching point group that belong to the target monitoring image of the (t + 1)-th frame. When ΔG t，t+1，i is smaller, it often indicates that the gray-scale distribution among the feature points in the i-th matching point group is more similar.

[0083] Third step, according to the pixel coordinates corresponding to the feature points in each matching point group, determining the movement direction difference corresponding to each matching point group may include the following sub-steps:

[0084] The first sub-step: Designate any one of the matching point groups as the marked matching point group, and designate two feature points in the above-mentioned marked matching point group as the first reference point and the second reference point respectively.

[0085] The second sub-step: Designate the target monitoring image to which the first reference point belongs as the first monitoring image, and designate the coordinate origin of the above-mentioned first monitoring image as the first origin.

[0086] Among them, the coordinate origin of the first monitoring image can be the origin of the pixel coordinate system, that is, the upper left corner of the first monitoring image.

[0087] The third sub-step: Screen out the pixel points with the same position as the second reference point from the above-mentioned first monitoring image as the temporary points.

[0088] The fourth sub-step: Connect the first reference point and the first origin to obtain the first straight line, and connect the temporary point and the first origin to obtain the second straight line.

[0089] The fifth sub-step: Designate the angle between the above-mentioned first straight line and the second straight line as the movement direction difference corresponding to the above-mentioned marked matching point group.

[0090] Step 4: Determine the movement performance index corresponding to each matching point group according to the movement distance, gray level distribution difference, and movement direction difference corresponding to each matching point group.

[0091] Among them, the movement distance, gray level distribution difference, and movement direction difference can all have a positive correlation with the movement performance index.

[0092] For example, the formula for determining the movement performance index corresponding to the matching point group can be:

[0093] SP t，t+1，i =ΔS t，t+1，i ×ΔG t，t+1，i ×Δθ t，t+1，i ; where SP t，t+1，i is the movement performance index corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image. t is the frame number of the target monitoring image. i is the serial number of the matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image. ΔS t，t+1，i is the movement distance corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image. ΔG t，t+1，i is the gray level distribution difference corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image. Δθ t，t+1，i is the movement direction difference corresponding to the i-th matching point group between the t-th frame target monitoring image and the (t + 1)-th frame target monitoring image.

[0094] It should be noted that when ΔS t，t+1，i is larger, it often indicates that the relative unit movement distance of the part represented by the i-th matching point group is relatively larger. When ΔG t，t+1，i is smaller, it often indicates that the gray-scale distribution among the feature points in the i-th matching point group is more similar. When Δθ t，t+1，i is larger, it often indicates that the movement direction of the part represented by the i-th matching point group is more likely to have changed within a unit time. Therefore, SP t，t+1，i can represent the movement condition of the part represented by the i-th matching point group within a unit time.

[0095] Step S4: Determine the non - possibility of the same person between every two matching point groups in the same adjacent two - frame target monitoring images according to the distance between the feature points in every two matching point groups and the difference between the corresponding motion performance indicators of these two matching point groups.

[0096] As an example, the formula corresponding to the non - possibility of the same person between any two matching point groups in the same adjacent two - frame target monitoring images can be:

[0097] ρ t，t+1，i，i+1 = |SP t，t+1，i - SP t，t+1，i+1 | × |d t，i，i+1 - d t+1，i，i+1 |; where ρ t，t+1，i，i+1 is the non - possibility of the same person between the i - th matching point group and the (i + 1)-th matching point group between the t - th frame target monitoring image and the (t + 1)-th frame target monitoring image. t is the frame number of the target monitoring image. i is the serial number of the matching point group between the t - th frame target monitoring image and the (t + 1)-th frame target monitoring image. || is the absolute - value function. SP t，t+1，i is the motion performance indicator corresponding to the i - th matching point group between the t - th frame target monitoring image and the (t + 1)-th frame target monitoring image. SP t，t+1，i+1 is the motion performance indicator corresponding to the (i + 1)-th matching point group between the t - th frame target monitoring image and the (t + 1)-th frame target monitoring image. d t，i，i+1 is the distance between the feature points of the i - th matching point group and the (i + 1)-th matching point group in the t - th frame target monitoring image. d t+1，i，i+1 is the distance between the feature points of the i - th matching point group and the (i + 1)-th matching point group in the (t + 1)-th frame target monitoring image.

[0098] It should be noted that in actual situations, the movement of the human body is often holistic, that is, for the same human body, most of the positions on it often move synchronously. SP t，t+1，i can represent the movement condition of the part represented by the i - th matching point group within a unit time. SPt，t+1，i+1 It can characterize the movement of the part represented by the (i + 1)-th matching point group within a unit time. d t，i，i+1 It can characterize the distance between the part represented by the i-th matching point group and the part represented by the (i + 1)-th matching point group at the acquisition moment corresponding to the t-th frame of the target monitoring image. d t+1，i，i+1 It can characterize the distance between the part represented by the i-th matching point group and the part represented by the (i + 1)-th matching point group at the acquisition moment corresponding to the (t + 1)-th frame of the target monitoring image. When |SP t，t+1，i -SP t，t+1，i+1 | is smaller, it often indicates that the parts represented by the i-th matching point group and the (i + 1)-th matching point group are more likely to move synchronously. When |d t，i，i+1 -d t+1，i，i+1 | is smaller, it often indicates that the parts represented by the i-th matching point group and the (i + 1)-th matching point group are more likely to move synchronously. Therefore, when ρ t，t+1，i，i+1 is smaller, it often indicates that the parts represented by the i-th matching point group and the (i + 1)-th matching point group are more likely to move synchronously, and it often indicates that the i-th matching point group and the (i + 1)-th matching point group are more likely to represent different positions of the same person during normal movement.

[0099] Step S5: According to the non - possibility of the same person between different matching point groups in all target monitoring images, screen out the person regions representing the same person to be detected.

[0100] As an example, this step may include the following steps:

[0101] First step: Determine any two adjacent target monitoring images as the first marked image and the second marked image respectively.

[0102] Second step: Form a marked group set with all the matching point groups between the first marked image and the second marked image.

[0103] Third step: Form a marked category with the matching point groups whose normalized value of the non - possibility of the same person between them in the above - mentioned marked group set is less than or equal to a preset same - possibility threshold.

[0104] Among them, the preset same - possibility threshold can be a threshold set in advance, and it can be 0.4.

[0105] It should be noted that the matching point groups in the same marked category are often the matching point groups with a relatively large non - possibility of the same person between them.

[0106] Fourth step: Determine the person region in the first marked image as the first marked region, and determine the person region in the second marked image as the second marked region.

[0107] Step 5: If the number of groups of matching points belonging to the same marking category between the first marking region and the second marking region is greater than a preset quantity threshold, it is determined that the first marking region and the second marking region represent the same person to be detected.

[0108] Among them, the preset quantity threshold can be a quantity threshold set in advance, and it can be equal to half of the total number of all groups of matching points between the first marking region and the second marking region.

[0109] Step S6: According to the movement trajectory situation among all the person regions representing each person to be detected, determine the abnormal behavior performance index corresponding to each person to be detected.

[0110] As an example, this step may include the following steps:

[0111] First step: Determine the center point of each person region as the person representative point, and determine any person to be detected as the marking detection person.

[0112] Second step: Connect each adjacent pair of person representative points corresponding to the above-mentioned marking detection person to construct a target vector between each adjacent pair of person representative points corresponding to the above-mentioned marking detection person.

[0113] Among them, the target monitoring images to which two adjacent person representative points belong may be two adjacent frames of target monitoring images.

[0114] For example, in the same pixel coordinate system, connect the pixel coordinates corresponding to two person representative points, take the length of this connection line segment as the modulus of the target vector between these two person representative points, and take the direction of this connection line segment pointing to the later person representative point among these two person representative points as the direction of the target vector between these two person representative points.

[0115] Third step: According to all the target vectors corresponding to the above-mentioned marking detection person, determining the abnormal behavior performance index corresponding to the above-mentioned marking detection person may include the following sub-steps:

[0116] First sub-step: Divide the target vectors with continuously the same direction among all the target vectors corresponding to the above-mentioned marking detection person into the same vector group.

[0117] For example, if all the target vectors corresponding to the marked detector are successively the first vector, the second vector, the third vector, the fourth vector, the fifth vector, the sixth vector, the seventh vector, the eighth vector, and the ninth vector, and the directions of the first vector and the second vector are the first direction, the direction of the third vector is the second direction, the directions of the fourth vector and the fifth vector are the third direction, the direction of the sixth vector is the second direction, and the directions of the seventh vector, the eighth vector, and the ninth vector are the first direction, then 5 vector groups can be obtained, and these 5 vector groups can be successively: {the first vector, the second vector}, {the third vector}, {the fourth vector, the fifth vector}, {the sixth vector}, and {the seventh vector, the eighth vector, the ninth vector}.

[0118] The second sub-step is to connect the first human representative point and the last human representative point corresponding to the marked detector to construct the overall displacement vector corresponding to the marked detector.

[0119] Among them, the first human representative point corresponding to the marked detector can be the center point of the earliest human area representing the marked detector. The last human representative point corresponding to the marked detector can be the center point of the latest human area representing the marked detector.

[0120] For example, in the same pixel coordinate system, connect the pixel coordinates corresponding to the first human representative point and the pixel coordinates corresponding to the last human representative point of the marked detector, take the length of this connection line segment as the modulus of the overall displacement vector corresponding to the marked detector, and take the direction of this connection line segment pointing to its last human representative point as the direction of the overall displacement vector corresponding to the marked detector.

[0121] The third sub-step is to determine the abnormal behavior performance index corresponding to the marked detector according to the number of vector groups and the modulus of the overall displacement vector.

[0122] For example, the formula for determining the abnormal behavior performance index corresponding to the marked detector can be:

[0123] Among them, β is the abnormal behavior performance index corresponding to the marked detector. norm() is the normalization function. T is the time duration between the acquisition time corresponding to the first human area of the marked detector and the acquisition time corresponding to the last human area, that is, the time duration between the acquisition time corresponding to the earliest human area of the marked detector and the acquisition time corresponding to the latest human area. H is the number of vector groups. M is the modulus of the overall displacement vector corresponding to the marked detector.

[0124] It should be noted that in actual situations, most normal pedestrians in the community usually do not stay at a certain position for a long time, nor do they frequently change their walking directions. When T is larger, it usually indicates that the marked detector is more likely to stay for a long time on the community road. When H is larger, it usually indicates that the marked detector is more likely to frequently change its walking direction on the community road. When M is smaller, it usually indicates that the marked detector is more likely to wander for a long time on the community road. Therefore, when β is larger, it usually indicates that the marked detector is more likely to have abnormal behavior on the community road.

[0125] Step S7, if the abnormal behavior performance index corresponding to the person to be detected is greater than the preset abnormal threshold, it is determined that the person to be detected has abnormal behavior, and an abnormal prompt message is controlled to be sent.

[0126] Among them, the preset abnormal threshold can be a threshold set in advance, which can be 0.7. The abnormal prompt message can be an abnormal behavior prompt message. For example, the abnormal prompt message can be "The person to be detected has abnormal behavior, and it is necessary for the staff to understand the on-site situation and take corresponding measures".

[0127] Reference Figure 2 , based on the same inventive concept as the above method embodiment, the present invention provides an intelligent control integration system based on cloud computing. The system includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the above computer program is executed by the processor, it realizes the steps of an intelligent control integration method based on cloud computing, which can specifically include:

[0128] An image acquisition and recognition module 201, configured to acquire each frame of target monitoring image of the community to be detected within a preset time period, and identify the person area from each frame of target monitoring image;

[0129] A feature point matching module 202, configured to perform feature point matching on the person areas in all adjacent frame target monitoring images to obtain a set of matching points;

[0130] A motion performance index determination module 203, configured to determine the motion performance index corresponding to each set of matching points according to the motion distance, gray-scale distribution difference, and moving direction difference between the feature points in each set of matching points;

[0131] A same-person improbability determination module 204, configured to determine the improbability of the same person between each two sets of matching points between the same adjacent two-frame target monitoring images according to the distance between the feature points in each two sets of matching points between the same adjacent two-frame target monitoring images and the difference between the motion performance indexes corresponding to these two sets of matching points;

[0132] The region screening module 205 is configured to screen out the person regions representing the same person to be detected according to the non - possibility of the same person among different matching point groups between all target monitoring images;

[0133] The abnormal behavior performance index determination module 206 is configured to determine the abnormal behavior performance index corresponding to each person to be detected according to the motion trajectory situation among all the person regions representing each person to be detected;

[0134] The information control and sending module 207 is configured to, if the abnormal behavior performance index corresponding to the person to be detected is greater than the preset abnormal threshold, determine that the person to be detected has an abnormal behavior, and control the sending of an abnormal prompt message.

[0135] Figure 3 It is a schematic structural diagram of a computer device provided by an embodiment of the present invention. Exemplarily, as Figure 3 shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302. Wherein, when the processor 302 executes the computer program 303, the computer device can execute any one of the above - introduced intelligent control integration methods based on cloud computing.

[0136] Based on the same inventive concept as the above - mentioned method embodiment, the present invention provides a server, including a memory and a processor. The memory is used to store executable program codes, and the processor is used to call and run the executable program codes from the memory, so that the device executes any one of the above - mentioned intelligent control integration methods based on cloud computing.

[0137] Based on the same inventive concept as the above - mentioned method embodiment, the present invention provides a computer program product, which includes: computer program codes. When the computer program codes run on a computer, the computer is enabled to execute any one of the above - mentioned intelligent control integration methods based on cloud computing.

[0138] Based on the same inventive concept as the above - mentioned method embodiment, the present invention provides a computer - readable storage medium, which stores computer program codes. When the computer program codes run on a computer, the computer is enabled to execute any one of the above - mentioned intelligent control integration methods based on cloud computing.

[0139] In summary, compared with relying on manual identification of abnormal behavior persons, the present invention analyzes multiple target monitoring images of the community to be detected, and relatively objectively quantifies multiple features related to abnormal behavior persons, such as motion performance indicators, non - possibility of the same person, and abnormal behavior performance indicators, thereby realizing the determination of abnormal behavior, further realizing the identification of abnormal behavior persons, and improving the timeliness of identifying abnormal behavior persons.

[0140] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A cloud computing-based intelligent control integration method, characterized in that: The following steps are involved: Obtain each frame of the target monitoring image of the community to be detected within a preset time period, and identify the human area from each frame of the target monitoring image; Perform feature point matching on the human area in all adjacent frame target monitoring images to obtain a matching point group; Determine the motion performance index corresponding to each matching point group according to the motion distance, grayscale distribution difference and moving direction difference between the feature points in each matching point group; According to the distance between the feature points in each two matching point groups between the same two adjacent target monitoring images, and the difference between the motion performance indicators corresponding to the two matching point groups, the impossibility of the same person between each two matching point groups between the same two adjacent target monitoring images is determined; According to the impossibility of the same person between different matching point groups among all target monitoring images, the person region representing the same person to be detected is screened out; According to the motion trajectories between all the character regions representing each person to be detected, the abnormal behavior performance index corresponding to each person to be detected is determined; If the abnormal behavior performance index corresponding to the person to be detected is greater than the preset abnormal threshold, it is determined that the person to be detected has abnormal behavior, and the control sends an abnormal prompt message.

2. The intelligent control integration method based on cloud computing according to claim 1 is characterized in that: Determining the motion performance index corresponding to each matching point group according to the motion distance, grayscale distribution difference and moving direction difference between the feature points in each matching point group includes: The distance between the pixel coordinates corresponding to the feature points in each matching point group is determined as the motion distance corresponding to each matching point group; Determine the grayscale distribution difference corresponding to each matching point group according to the grayscale difference between each feature point in each matching point group and its corresponding preset neighborhood; Determine the moving direction difference corresponding to each matching point group according to the pixel coordinates corresponding to the feature points in each matching point group; The motion performance index corresponding to each matching point group is determined according to the motion distance, grayscale distribution difference and moving direction difference corresponding to each matching point group, wherein the motion distance, grayscale distribution difference and moving direction difference are all positively correlated with the motion performance index.

3. The cloud computing-based intelligent control integration method according to claim 2, characterized in that: The step of determining the moving direction difference corresponding to each matching point group according to the pixel coordinates corresponding to the feature points in each matching point group includes: Determine any matching point group as a marked matching point group, and determine two feature points in the marked matching point group as a first reference point and a second reference point, respectively; Determine the target monitoring image to which the first reference point belongs as the first monitoring image, and determine the coordinate origin of the first monitoring image as the first origin; Filtering out pixel points at the same position as the second reference point from the first monitoring image as temporary points; Connecting the first reference point and the first origin to obtain a first straight line, and connecting the temporary point and the first origin to obtain a second straight line; The angle between the first straight line and the second straight line is determined as the difference in movement directions corresponding to the marked matching point group.

4. The cloud computing-based intelligent control integration method according to claim 2, characterized in that: The formula corresponding to the grayscale distribution difference of the matching point group is: Among them, ΔG t,t+1,i is the grayscale distribution difference between the target monitoring image of the tth frame and the target monitoring image of the t+1th frame, corresponding to the i-th matching point group; t is the frame number of the target monitoring image; i is the serial number of the matching point group between the target monitoring image of the tth frame and the target monitoring image of the t+1th frame; || is the absolute value function; N is the number of pixels in the preset neighborhood; j is the serial number of the pixel in the preset neighborhood; G t,i is the gray value corresponding to the feature point of the target monitoring image of the tth frame in the i-th matching point group; G t+1,i is the gray value corresponding to the feature point of the target monitoring image of the t+1th frame in the i-th matching point group; G t,ij G is the gray value corresponding to the jth pixel in the preset neighborhood corresponding to the feature point of the target monitoring image in the tth frame in the i-th matching point group; t+1,ij It is the gray value corresponding to the jth pixel point in the preset neighborhood corresponding to the feature point of the target monitoring image of the t+1th frame in the i-th matching point group.

5. The cloud computing-based intelligent control integration method according to claim 1, characterized in that: The formula for the improbable correspondence between any two matching point groups of the same person between two adjacent target surveillance images is: ρ t,t+1,i,i+1 =|SP t,t+1,i -SP t,t+1,i+1 |×|d t,i,i+1 -d t+1,i,i+1 |; Among them, ρ t,t+1,i,i+1 is the impossibility of the same person between the target monitoring image of the t-th frame and the target monitoring image of the t+1-th frame, and between the i-th matching point group and the i+1-th matching point group; t is the frame number of the target monitoring image; i is the sequence number of the matching point group between the target monitoring image of the t-th frame and the target monitoring image of the t+1-th frame; || is the absolute value function; SP t,t+1,i is the motion performance index corresponding to the i-th matching point group between the t-th frame target monitoring image and the t+1-th frame target monitoring image; SP t,t+1,i+1 is the motion performance index corresponding to the i+1th matching point group between the tth frame target monitoring image and the t+1th frame target monitoring image; d t,i,i+1 is the distance between the feature points of the i-th matching point group and the i+1-th matching point group in the target monitoring image of the t-th frame; d t+1,i,i+1 It is the distance between the feature points of the i-th matching point group and the i+1-th matching point group in the t+1-th frame target monitoring image.

6. The cloud computing-based intelligent control integration method according to claim 1, characterized in that: The method of screening out the person region representing the same person to be detected according to the impossibility of the same person between different matching point groups among all target monitoring images comprises: Determine any two adjacent target monitoring images as a first marked image and a second marked image respectively; All matching point groups between the first labeled image and the second labeled image form a labeled group set; The matching point groups whose normalized values ​​of the same person's improbability in the tag group set are less than or equal to the same preset possibility threshold value are used to form a tag category; Determine the person region in the first marked image as the first marked region, and determine the person region in the second marked image as the second marked region; If the number of matching point groups between the first marking area and the second marking area that belong to the same marking category is greater than a preset number threshold, it is determined that the first marking area and the second marking area represent the same person to be detected.

7. The intelligent control integration method based on cloud computing according to claim 1 is characterized in that: Determining the abnormal behavior performance index corresponding to each person to be detected based on the motion trajectory between all the character regions representing each person to be detected includes: The center point of each character area is determined as the character representative point, and any person to be detected is determined as the marked detection person; Connecting every two adjacent character representative points corresponding to the marked detected person, and constructing a target vector between every two adjacent character representative points corresponding to the marked detected person; According to all target vectors corresponding to the marked detection person, an abnormal behavior performance index corresponding to the marked detection person is determined.

8. The intelligent control integration method based on cloud computing according to claim 7 is characterized in that: The step of determining the abnormal behavior performance index corresponding to the marked detection person according to all target vectors corresponding to the marked detection person includes: Dividing the target vectors with the same continuous direction among all the target vectors corresponding to the marked detected person into the same vector group; Connecting the first character representative point and the last character representative point corresponding to the marked detected person to construct an overall displacement vector corresponding to the marked detected person; According to the number of vector groups and the modulus of the overall displacement vector, an abnormal behavior performance index corresponding to the marked detection person is determined.

9. The cloud computing-based intelligent control integration method according to claim 8, characterized in that: The formula corresponding to the abnormal behavior performance index of the marked detection person is: Among them, β is the abnormal behavior performance indicator corresponding to the marked person; norm() is the normalization function; T is the time between the acquisition time corresponding to the first person area of ​​the marked person and the acquisition time corresponding to the last person area; H is the number of vector groups; M is the modulus of the overall displacement vector corresponding to the marked person.

10. An intelligent control integrated system based on cloud computing, characterized in that: It comprises a processor and a memory, wherein the processor is used to process instructions stored in the memory to implement a cloud computing-based intelligent control integration method as described in any one of claims 1 to 9.

Images