CN112465870B - Thermal image alarm intrusion detection method and device under complex background - Google Patents
Thermal image alarm intrusion detection method and device under complex background Download PDFInfo
- Publication number
- CN112465870B CN112465870B CN202011437523.4A CN202011437523A CN112465870B CN 112465870 B CN112465870 B CN 112465870B CN 202011437523 A CN202011437523 A CN 202011437523A CN 112465870 B CN112465870 B CN 112465870B
- Authority
- CN
- China
- Prior art keywords
- motion
- target
- moving
- alarm
- track
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30241—Trajectory
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Burglar Alarm Systems (AREA)
- Image Analysis (AREA)
- Closed-Circuit Television Systems (AREA)
Abstract
The invention provides a thermal image alarm intrusion detection method and a device under a complex background, wherein the method comprises the following steps: s1, acquiring a thermal imaging video frame, creating a motion background model, inputting the thermal imaging video frame into the motion background model, and extracting a moving target; s2, detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the moving state of the target by adopting a filtering algorithm, completing the matching of the historical frame, and generating a moving track; s3, setting sampling points according to the motion trail, and extracting and counting morphological features and motion features of historical targets of the sampling points; s4, false alarm filtering is carried out according to morphological characteristics and motion characteristics of the motion trail, and a real alarm is output. The thermal image alarm intrusion detection method and device provided by the invention effectively filter out a large number of false alarms, and realize a smaller false alarm rate while maintaining higher sensitivity.
Description
Technical Field
The invention belongs to the technical field of thermal image perimeter intrusion detection, and particularly relates to a thermal image alarm intrusion detection method and device under a complex background.
Background
GMM, which is a short term for Gauss ian Mixture Model, is also referred to as MOG, and is a gaussian mixture model, which precisely quantizes things by using a gaussian probability density function (normal distribution curve) and decomposes one thing into a plurality of models formed based on the gaussian probability density function (normal distribution curve).
KNN is short for K-NearestNeighbor, and refers to a proximity algorithm or K nearest neighbor classification algorithm, and is one of the simplest methods in the data mining classification technology.
The environment of the frontier defense has the characteristics of wide visual field and complex environment, and the moving target is usually very small and is generally below 15 pixels. In order to detect small moving objects, a high sensitivity must be maintained, and at high sensitivity, due to the complexity of the environment, a large number of false alarms, such as wind blowing, thermal waves formed by wheat waves, projection of clouds on the ground, etc., may be caused. Conventional alarm intrusion is generally based on GMM to perform motion background modeling, detect a motion target, extract characteristics of a single motion target, and filter. However, the targets of the border protection environment are very small, the characteristics are not obvious, the characteristics of a single target have high randomness, and the single target cannot be distinguished from a large number of false alarm points.
This is a deficiency of the prior art, and therefore, it is necessary to provide a method and apparatus for detecting thermal image alarm intrusion in a complex background, aiming at the above-mentioned drawbacks of the prior art.
Disclosure of Invention
Aiming at the defects that the targets of the side protection environment in the prior art are very small, the characteristics are not obvious, the characteristics of a single target have very large randomness and cannot be distinguished from a large number of false alarm points, the invention provides a thermal image alarm intrusion detection method and device under a complex background, and aims to solve the technical problems.
In a first aspect, the present invention provides a thermal imaging alarm intrusion detection method under a complex background, including the steps of:
s1, acquiring a thermal imaging video frame, creating a motion background model, inputting the thermal imaging video frame into the motion background model, and extracting a moving target;
s2, detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the moving state of the target by adopting a filtering algorithm, completing the matching of the historical frame, and generating a moving track;
s3, setting sampling points according to the motion trail, and extracting and counting morphological features and motion features of historical targets of the sampling points;
s4, false alarm filtering is carried out according to morphological characteristics and motion characteristics of the motion trail, and a real alarm is output.
Further, the specific steps of step S1 are as follows:
s11, acquiring a thermal imaging video frame through thermal imaging equipment;
s12, modeling a motion background based on a KNN algorithm to generate a KNN motion background model;
s13, inputting the thermal imaging video frame into a KNN motion background model, and extracting a motion target.
Further, the specific steps of step S2 are as follows:
s21, detecting a moving target by adopting a KNN algorithm;
s22, generating an initial track based on the maximum intersection ratio of the current frame and the historical frame of the moving target;
s23, counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
s24, performing historical frame matching based on the Hungary algorithm on the smoothed initial track to generate a motion track.
Further, the specific steps of step S3 are as follows:
s31, setting an adopted frequency, and extracting a corresponding historical target in a motion track according to the adopted frequency;
s32, recording morphological characteristics of a historical target, wherein the morphological characteristics comprise an aspect ratio and a target pixel area;
s33, recording motion characteristics of a historical target, wherein the motion characteristics comprise the moving speed of the target and the moving distance of the target in an adjacent frame;
s34, after the motion trail is generated for a set period of time, counting the speed, the aspect ratio, the target pixel area, the dispersion of the target moving speed and the linear pixel distance of the moving target after the trail is finished.
Further, the specific steps of step S4 are as follows:
s41, adjusting the duration of the motion trail, and filtering alarms with duration not meeting a threshold value;
s42, adjusting a dispersion threshold value, and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being greater than the dispersion threshold value;
s43, filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
s44, setting the filtered alarm as false alarm, and outputting the rest real alarm.
Further, the specific steps of step S41 are as follows:
s411, setting a motion trail duration threshold as a first threshold;
s412, judging whether the duration of the motion trail is greater than a first threshold value;
if yes, go to step S413;
if not, go to step S42;
s413, filtering the motion trail corresponding to the alarm.
Further, the specific steps of step S42 are as follows:
s421, calculating the dispersion of the morphological characteristics of the target and the dispersion of the moving speed characteristics of the target according to the historical information of the moving target recorded on the moving track;
s422, setting a dispersion threshold value of the target morphological characteristics as a second threshold value and setting a dispersion threshold value of the target moving speed as a third threshold value;
s423, judging whether the dispersion degree of the target morphological characteristics of the motion trail is larger than a second threshold value;
if yes, go to step S424;
if not, go to step S43;
s424, whether the dispersion of the target moving speed is larger than a third threshold value or not;
if yes, go to step S425;
if not, go to step S43;
s425, filtering the motion trail corresponding to the alarm.
Further, the specific steps of step S43 are as follows:
s431, setting the linear pixel distance of the target movement as a fourth threshold value;
s432, after the motion trail lasts for a set time period, obtaining the linear pixel distance of the motion object;
s433, judging whether the linear pixel distance of the moving object is smaller than a fourth threshold value;
if yes, go to step S434;
if not, go to step S44;
s434, filtering the motion trail corresponding to the alarm.
In a second aspect, the present invention provides a thermal imaging alarm intrusion detection device under a complex background, including:
the moving target extraction module is used for acquiring a thermal imaging video frame, creating a moving background model, inputting the thermal imaging video frame into the moving background model, and extracting a moving target;
the motion track generation module is used for detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the motion state of the target by adopting a filtering algorithm, and completing the matching of the historical frame to generate the motion track;
the motion trail feature extraction module is used for setting sampling points according to the motion trail and extracting and counting morphological features and motion features of historical targets of the sampling points;
and the false alarm filtering module is used for carrying out false alarm filtering according to the morphological characteristics and the motion characteristics of the motion trail and outputting a real alarm.
Further, the moving object extraction module includes:
a thermal imaging video frame acquisition unit for acquiring a thermal imaging video frame by a thermal imaging device;
the motion background model generation unit is used for carrying out motion background modeling based on a KNN algorithm to generate a KNN motion background model;
the moving target extraction unit is used for inputting the thermal imaging video frame into the KNN moving background model to extract a moving target;
the motion trail generation module comprises:
the moving target detection unit is used for detecting the moving target by adopting a KNN algorithm;
an initial track generation unit for generating an initial track based on the maximum intersection ratio of the current frame and the history frame of the moving object;
the initial track smoothing unit is used for counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
the historical frame matching and motion track generating unit is used for performing historical frame matching based on a Hungary algorithm on the smoothed initial track to generate a motion track;
the motion trail feature extraction module comprises:
the historical target extraction unit is used for setting the adoption frequency and extracting a corresponding historical target in the motion trail according to the adoption frequency;
a morphology feature recording unit for recording morphology features of a history target, the morphology features including aspect ratio and target pixel area;
the motion characteristic recording unit is used for recording motion characteristics of the historical target, wherein the motion characteristics comprise the moving speed of the target and the moving distance of the target in the adjacent frames;
the feature statistics unit is used for counting the speed, the aspect ratio, the target pixel area, the dispersion of the target moving speed and the linear pixel distance of the moving target after the moving target is moved after the track is ended after the movement track is generated for a set period of time;
the false alarm filtering module comprises:
the short false alarm filtering unit is used for adjusting the duration time of the motion trail and filtering alarms with duration time not meeting a threshold value;
the large-scale false alarm filtering unit is used for adjusting a dispersion threshold value and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being greater than the dispersion threshold value;
the small deformation false alarm filtering unit is used for filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
and the true alarm output unit is used for setting the filtered alarm as a false alarm and outputting the rest true alarms.
The invention has the advantages that,
according to the thermal image alarm intrusion detection method and device under the complex background, the moving target is detected based on the KNN algorithm, the track of the moving target is generated by combining the Hungary matching algorithm with the front and rear frames of the target, the Kalman filtering algorithm is adopted for track smoothing and target moving state prediction, and meanwhile, feature statistics is carried out on morphological features and moving features of each historical target forming the track, so that a large number of false alarms are effectively filtered, higher sensitivity is maintained, and meanwhile, a smaller false alarm rate is realized.
In addition, the invention has reliable design principle, simple structure and very wide application prospect.
It can be seen that the present invention has outstanding substantial features and significant advances over the prior art, as well as the benefits of its implementation.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the description of the embodiments or the prior art will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.
FIG. 1 is a schematic flow chart of the method of the present invention;
FIG. 2 is a second schematic flow chart of the method of the present invention;
FIG. 3 is a schematic view of the apparatus of the present invention;
in the figure, a 1-moving object extraction module; 1.1-a thermal imaging video frame acquisition unit; 1.2-a motion background model generation unit; 1.3-moving object extraction unit; 2-a motion trail generation module; 2.1 moving object detection unit-; 2.2-an initial trajectory generation unit; 2.3-an initial trajectory smoothing unit; 2.4-a history frame matching and motion trail generating unit; 3-a motion trail feature extraction module; 3.1-a history object extraction unit; 3.2-a morphological feature recording unit; 3.3-a motion feature recording unit; 3.4-feature statistics unit; 4-false alarm filtering module; 4.1-a short false positive filtering unit; 4.2-a large-scale false alarm filtering unit; 4.3-small deformation false alarm filtering unit; 4.4-true alarm output unit.
Detailed Description
In order to make the technical solution of the present invention better understood by those skilled in the art, the technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Example 1:
as shown in fig. 1, the invention provides a thermal image alarm intrusion detection method under a complex background, which comprises the following steps:
s1, acquiring a thermal imaging video frame, creating a motion background model, inputting the thermal imaging video frame into the motion background model, and extracting a moving target;
s2, detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the moving state of the target by adopting a filtering algorithm, completing the matching of the historical frame, and generating a moving track;
s3, setting sampling points according to the motion trail, and extracting and counting morphological features and motion features of historical targets of the sampling points;
s4, false alarm filtering is carried out according to morphological characteristics and motion characteristics of the motion trail, and a real alarm is output.
Example 2:
as shown in fig. 2, the invention provides a thermal image alarm intrusion detection method under a complex background, which comprises the following steps:
s1, acquiring a thermal imaging video frame, creating a motion background model, inputting the thermal imaging video frame into the motion background model, and extracting a moving target; the method comprises the following specific steps:
s11, acquiring a thermal imaging video frame through thermal imaging equipment;
s12, modeling a motion background based on a KNN algorithm to generate a KNN motion background model;
s13, inputting a thermal imaging video frame into a KNN motion background model, and extracting a motion target;
s2, detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the moving state of the target by adopting a filtering algorithm, completing the matching of the historical frame, and generating a moving track; the method comprises the following specific steps:
s21, detecting a moving target by adopting a KNN algorithm;
s22, generating an initial track based on the maximum intersection ratio of the current frame and the historical frame of the moving target; the success rate of the latter Hungary matching is improved based on the initial track generated by the maximum cross ratio;
s23, counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
s24, performing historical frame matching based on a Hungary algorithm on the smoothed initial track to generate a motion track;
s3, setting sampling points according to the motion trail, and extracting and counting morphological features and motion features of historical targets of the sampling points; the method comprises the following specific steps:
s31, setting an adopted frequency, and extracting a corresponding historical target in a motion track according to the adopted frequency;
s32, recording morphological characteristics of a historical target, wherein the morphological characteristics comprise an aspect ratio and a target pixel area;
s33, recording motion characteristics of a historical target, wherein the motion characteristics comprise the moving speed of the target and the moving distance of the target in an adjacent frame;
s34, after a motion track is generated for a set period of time, counting the speed, the aspect ratio, the target pixel area, the dispersion of the target moving speed and the linear pixel distance of the moving target after the motion track is finished;
s4, performing false alarm filtering according to morphological characteristics and motion characteristics of the motion trail, and outputting a real alarm; the method comprises the following specific steps:
s41, adjusting the duration of the motion trail, and filtering alarms with duration not meeting a threshold value;
s42, adjusting a dispersion threshold value, and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being greater than the dispersion threshold value;
s43, filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
s44, setting the filtered alarm as false alarm, and outputting the rest real alarm.
In the above embodiment, the intersection ratio is a concept in object detection, and is the ratio of the generated candidate frame to the original marked frame, that is, the ratio of their intersection to the union, and the optimal situation is complete overlapping, that is, the ratio is 1, that is, the current frame and the history frame are completely matched and optimal.
In certain embodiments, step S41 is specifically as follows:
s411, setting a motion trail duration threshold as a first threshold;
s412, judging whether the duration of the motion trail is greater than a first threshold value;
if yes, go to step S413;
if not, go to step S42;
s413, filtering the motion trail corresponding to the alarm;
the step S41 is used for filtering short false alarms, and the short false alarms mainly come from bare rocks in sunlight, shake of plants and the like, and are characterized by short duration, generally in the millisecond level and small target size;
the false alarm can be filtered by adjusting the duration of the track, and a true alarm can be formed only after the duration reaches a certain degree.
In some embodiments, step S42 is specifically as follows:
s421, calculating the dispersion of the morphological characteristics of the target and the dispersion of the moving speed characteristics of the target according to the historical information of the moving target recorded on the moving track;
s422, setting a dispersion threshold value of the target morphological characteristics as a second threshold value and setting a dispersion threshold value of the target moving speed as a third threshold value;
s423, judging whether the dispersion degree of the target morphological characteristics of the motion trail is larger than a second threshold value;
if yes, go to step S424;
if not, go to step S43;
s424, whether the dispersion of the target moving speed is larger than a third threshold value or not;
if yes, go to step S425;
if not, go to step S43;
s425, filtering the motion trail corresponding to the alarm;
step S42 is implemented by filtering large deformation false alarms, the large deformation false alarms usually have long duration and large deformation scale, the false alarms mainly originate from wheat waves and cloud shadows, particularly, in the daytime and when wind exists, waves of one wave are formed on a thermal image, and after the waves are processed by a KNN algorithm, a moving target with longer duration is formed, and the duration is unequal from 1 second to 8 seconds; in addition, when clouds exist on the mountain, moving targets with longer duration can be formed at the edges of the clouds, and the moving targets are characterized by large deformation scale;
according to the historical information of the target recorded on the track, calculating the dispersion degree of the morphological characteristics and the speed characteristics of the target, finding that the dispersion degree of the false alarm is larger, and according to the characteristics, adjusting the threshold value of the dispersion degree, the large-scale false alarm can be filtered out.
In certain embodiments, step S43 is specifically as follows:
s431, setting the linear pixel distance of the target movement as a fourth threshold value;
s432, after the motion trail lasts for a set time period, obtaining the linear pixel distance of the motion object;
s433, judging whether the linear pixel distance of the moving object is smaller than a fourth threshold value;
if yes, go to step S434;
if not, go to step S44;
s434, filtering the motion trail corresponding to the alarm;
step S43 is implemented by filtering small deformation false alarms, the false alarms have long duration and small deformation, and the false alarms are fewer and are mostly existing in wheat waves, and are characterized in that a target often wanders around a point;
after the motion trail lasts for a period of time, if the linear distance of the object moving is found to be short or the object does not move, the small deformation false alarm can be judged.
Example 3:
as shown in fig. 3, the present invention provides a thermal image alarm intrusion detection device under a complex background, including:
the moving target extraction module 1 is used for acquiring a thermal imaging video frame, creating a moving background model, inputting the thermal imaging video frame into the moving background model, and extracting a moving target; the moving object extraction module 1 includes:
a thermal imaging video frame acquisition unit 1.1 for acquiring a thermal imaging video frame by a thermal imaging device;
the motion background model generating unit 1.2 is used for performing motion background modeling based on a KNN algorithm to generate a KNN motion background model;
the moving target extraction unit 1.3 is used for inputting the thermal imaging video frame into the KNN moving background model to extract a moving target;
the motion track generation module 2 is used for detecting a motion target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the motion target, smoothing the initial track and predicting the motion state of the target by adopting a filtering algorithm, and completing the matching of the historical frame to generate the motion track; the motion trajectory generation module 2 includes:
the moving target detection unit 2.1 is used for detecting the moving target by adopting a KNN algorithm;
an initial trajectory generation unit 2.2 for generating an initial trajectory based on a maximum intersection ratio of a current frame and a history frame of the moving object;
the initial track smoothing unit 2.3 is used for counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
the historical frame matching and motion trail generation unit 2.4 is used for performing historical frame matching based on the Hungary algorithm on the smoothed initial trail to generate a motion trail;
the motion trail feature extraction module 3 is used for setting sampling points according to the motion trail and extracting and counting morphological features and motion features of historical targets of the sampling points; the motion trajectory feature extraction module 3 includes:
a history target extracting unit 3.1, configured to set an adoption frequency, and extract a corresponding history target in the motion trail according to the adoption frequency;
a morphology feature recording unit 3.2 for recording morphology features of the history object, the morphology features including aspect ratio and object pixel area;
a motion characteristic recording unit 3.3 for recording motion characteristics of the historical target, wherein the motion characteristics comprise the moving speed of the target and the moving distance of the target in adjacent frames;
the feature statistics unit 3.4 is used for counting the speed, the aspect ratio, the target pixel area, the dispersion of the target moving speed and the linear pixel distance of the moving target moving after the moving target is finished after the moving track is generated for a set period of time;
the false alarm filtering module 4 is used for performing false alarm filtering according to the morphological characteristics and the motion characteristics of the motion trail and outputting a real alarm; the false alarm filtering module 4 includes:
the short false alarm filtering unit 4.1 is used for adjusting the duration time of the motion trail and filtering the alarms with duration time not meeting the threshold value;
the large-scale false alarm filtering unit 4.2 is used for adjusting a dispersion threshold value and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being larger than the dispersion threshold value;
the small deformation false alarm filtering unit 4.3 is used for filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
and the true alarm output unit 4.4 is used for setting the filtered alarm as false alarm and outputting the rest true alarms.
Although the present invention has been described in detail by way of preferred embodiments with reference to the accompanying drawings, the present invention is not limited thereto. Various equivalent modifications and substitutions may be made in the embodiments of the present invention by those skilled in the art without departing from the spirit and scope of the present invention, and it is intended that all such modifications and substitutions be within the scope of the present invention/be within the scope of the present invention as defined by the appended claims. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (6)
1. A thermal image alarm intrusion detection method under a complex background is characterized by comprising the following steps:
s1, acquiring a thermal imaging video frame, creating a motion background model, inputting the thermal imaging video frame into the motion background model, and extracting a moving target;
s2, detecting a moving target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the moving target, smoothing the initial track and predicting the moving state of the target by adopting a filtering algorithm, completing the matching of the historical frame, and generating a moving track; the specific steps of the step S2 are as follows:
s21, detecting a moving target by adopting a KNN algorithm;
s22, generating an initial track based on the maximum intersection ratio of the current frame and the historical frame of the moving target;
s23, counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
s24, performing historical frame matching based on a Hungary algorithm on the smoothed initial track to generate a motion track;
s3, setting sampling points according to the motion trail, and extracting and counting morphological features and motion features of historical targets of the sampling points; the specific steps of the step S3 are as follows:
s31, setting a sampling frequency, and extracting a corresponding historical target in a motion track according to the sampling frequency;
s32, recording morphological characteristics of a historical target, wherein the morphological characteristics comprise an aspect ratio and a target pixel area;
s33, recording motion characteristics of a historical target, wherein the motion characteristics comprise the moving speed of the target and the moving distance of the target in an adjacent frame;
s34, after a motion track is generated for a set period of time, counting the speed, the aspect ratio, the target pixel area, the dispersion of the target moving speed and the linear pixel distance of the moving target after the motion track is finished;
s4, performing false alarm filtering according to morphological characteristics and motion characteristics of the motion trail, and outputting a real alarm; the specific steps of the step S4 are as follows:
s41, adjusting the duration of the motion trail, and filtering alarms with duration not meeting a threshold value;
s42, adjusting a dispersion threshold value, and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being greater than the dispersion threshold value; the specific steps of step S42 are as follows:
s421, calculating the dispersion of the morphological characteristics of the target and the dispersion of the moving speed characteristics of the target according to the historical information of the moving target recorded on the moving track;
s422, setting a dispersion threshold value of the target morphological characteristics as a second threshold value and setting a dispersion threshold value of the target moving speed as a third threshold value;
s423, judging whether the dispersion degree of the target morphological characteristics of the motion trail is larger than a second threshold value;
if yes, go to step S424;
if not, go to step S43;
s424, whether the dispersion of the target moving speed is larger than a third threshold value or not;
if yes, go to step S425;
if not, go to step S43;
s425, filtering the motion trail corresponding to the alarm;
s43, filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
s44, setting the filtered alarm as false alarm, and outputting the rest real alarm.
2. The method for detecting thermal image alarm intrusion under complex background according to claim 1, wherein the specific steps of step S1 are as follows:
s11, acquiring a thermal imaging video frame through thermal imaging equipment;
s12, modeling a motion background based on a KNN algorithm to generate a KNN motion background model;
s13, inputting the thermal imaging video frame into a KNN motion background model, and extracting a motion target.
3. The method for detecting thermal image alarm intrusion under complex background according to claim 1, wherein step S41 comprises the following specific steps:
s411, setting a motion trail duration threshold as a first threshold;
s412, judging whether the duration of the motion trail is greater than a first threshold value;
if yes, go to step S413;
if not, go to step S42;
s413, filtering the motion trail corresponding to the alarm.
4. The method for detecting thermal image alarm intrusion under complex background according to claim 1, wherein step S43 comprises the following specific steps:
s431, setting the linear pixel distance of the target movement as a fourth threshold value;
s432, after the motion trail lasts for a set time period, obtaining the linear pixel distance of the motion object;
s433, judging whether the linear pixel distance of the moving object is smaller than a fourth threshold value;
if yes, go to step S434;
if not, go to step S44;
s434, filtering the motion trail corresponding to the alarm.
5. The utility model provides a thermal image warning intrusion detection device under complicated background which characterized in that includes:
the moving target extraction module (1) is used for acquiring a thermal imaging video frame, creating a moving background model, inputting the thermal imaging video frame into the moving background model, and extracting a moving target;
the motion track generation module (2) is used for detecting a motion target by adopting a KNN algorithm, generating an initial track according to the intersection ratio of a current frame and a historical frame of the motion target, smoothing the initial track and predicting the motion state of the target by adopting a filtering algorithm, and completing the matching of the historical frame to generate the motion track; the motion trail generation module (2) comprises:
the moving target detection unit (2.1) is used for detecting the moving target by adopting a KNN algorithm;
an initial trajectory generation unit (2.2) for generating an initial trajectory based on a maximum intersection ratio of a current frame and a history frame of a moving object;
the initial track smoothing unit (2.3) is used for counting the speed information of the moving object, constructing an initial state matrix and a motion equation of a Kalman filtering algorithm according to the speed information of the moving object, and finishing initial track smoothing and object motion state prediction;
the historical frame matching and motion track generating unit (2.4) is used for performing historical frame matching based on the Hungary algorithm on the initial track after smoothing to generate a motion track;
the motion trail feature extraction module (3) is used for setting sampling points according to the motion trail and extracting and counting morphological features and motion features of historical targets of the sampling points; the motion trail feature extraction module (3) comprises:
a historical target extraction unit (3.1) for setting a sampling frequency and extracting a corresponding historical target in the motion trail according to the sampling frequency;
a morphology feature recording unit (3.2) for recording morphology features of a history object, the morphology features comprising an aspect ratio and an object pixel area;
a motion feature recording unit (3.3) for recording motion features of a history object, the motion features including an object moving speed, a moving distance of the object in an adjacent frame;
a feature statistics unit (3.4) for counting a speed, an aspect ratio, a target pixel area, a dispersion of a target moving speed and a linear pixel distance of a moving target moving after the moving target is finished after the moving track is generated for a set period of time;
the false alarm filtering module (4) is used for carrying out false alarm filtering according to the morphological characteristics and the motion characteristics of the motion trail and outputting a real alarm; the false alarm filtering module (4) comprises:
a short false alarm filtering unit (4.1) for adjusting the duration of the motion trail and filtering alarms with duration not meeting a threshold value;
the large-scale false alarm filtering unit (4.2) is used for adjusting a dispersion threshold value and filtering alarms with the dispersion of the target morphological characteristics and the target moving speed being larger than the dispersion threshold value;
the small deformation false alarm filtering unit (4.3) is used for filtering an alarm that the linear pixel distance of the moving target after the track is finished is smaller than a threshold value;
and the true alarm output unit (4.4) is used for setting the filtered alarm as false alarm and outputting the rest true alarms.
6. The thermal imaging alarm intrusion detection device under complex background according to claim 5, wherein the moving object extraction module (1) comprises:
a thermal imaging video frame acquisition unit (1.1) for acquiring a thermal imaging video frame by a thermal imaging device;
the motion background model generating unit (1.2) is used for carrying out motion background modeling based on a KNN algorithm to generate a KNN motion background model;
and the moving target extraction unit (1.3) is used for inputting the thermal imaging video frame into the KNN moving background model to extract the moving target.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011437523.4A CN112465870B (en) | 2020-12-10 | 2020-12-10 | Thermal image alarm intrusion detection method and device under complex background |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011437523.4A CN112465870B (en) | 2020-12-10 | 2020-12-10 | Thermal image alarm intrusion detection method and device under complex background |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112465870A CN112465870A (en) | 2021-03-09 |
| CN112465870B true CN112465870B (en) | 2023-07-14 |
Family
ID=74800793
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011437523.4A Active CN112465870B (en) | 2020-12-10 | 2020-12-10 | Thermal image alarm intrusion detection method and device under complex background |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112465870B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113238214B (en) * | 2021-04-14 | 2024-03-19 | 杭州萤石软件有限公司 | Target object detection method, target object detection device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103413330A (en) * | 2013-08-30 | 2013-11-27 | 中国科学院自动化研究所 | Method for reliably generating video abstraction in complex scene |
| CN111932580A (en) * | 2020-07-03 | 2020-11-13 | 江苏大学 | Road 3D vehicle tracking method and system based on Kalman filtering and Hungary algorithm |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101739691B (en) * | 2009-12-04 | 2011-12-14 | 北京智安邦科技有限公司 | Method and device for detecting video false alarm target |
| CN103489202A (en) * | 2013-01-15 | 2014-01-01 | 上海盈觉智能科技有限公司 | Intrusion detection method based on videos |
| CN104574340A (en) * | 2013-10-18 | 2015-04-29 | 上海睿穹信息科技有限公司 | Video intrusion detection method based on historical images |
| CN105141903B (en) * | 2015-08-13 | 2018-06-19 | 中国科学院自动化研究所 | A kind of method for carrying out target retrieval in video based on colouring information |
| CN110866943B (en) * | 2019-11-21 | 2022-06-14 | 无锡中科水质环境技术有限公司 | Fish position tracking method for water quality monitoring |
| CN111507235B (en) * | 2020-04-13 | 2024-05-28 | 北京交通大学 | Railway perimeter foreign matter intrusion detection method based on video |
| CN111582060B (en) * | 2020-04-20 | 2023-04-18 | 浙江大华技术股份有限公司 | Automatic line drawing perimeter alarm method, computer equipment and storage device |
-
2020
- 2020-12-10 CN CN202011437523.4A patent/CN112465870B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103413330A (en) * | 2013-08-30 | 2013-11-27 | 中国科学院自动化研究所 | Method for reliably generating video abstraction in complex scene |
| CN111932580A (en) * | 2020-07-03 | 2020-11-13 | 江苏大学 | Road 3D vehicle tracking method and system based on Kalman filtering and Hungary algorithm |
Non-Patent Citations (5)
| Title |
|---|
| Implementation of algorithms for Point target detection and tracking in Infrared image sequences;Ravi Vaishnavi.et al;《IEEE》;全文 * |
| 一种低复杂度的监控视频异物入侵检测算法;刘峰等;南京邮电大学学报(自然科学版)(第01期);全文 * |
| 冯莹莹等.《智能监控视频中运动目标跟踪方法研究》.长春：吉林大学出版社,2018,全文. * |
| 基于YOLOv3与卡尔曼滤波的多目标跟踪算法;任珈民等;计算机应用与软件(第05期);全文 * |
| 基于移动平台热成像的行人检测与跟踪研究;诸葛琳娜;《中国优秀硕士学位论文全文数据库 信息科技》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112465870A (en) | 2021-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109101944B (en) | Real-time video monitoring method for identifying garbage thrown into river channel | |
| KR101271092B1 (en) | Method and apparatus of real-time segmentation for motion detection in surveillance camera system | |
| CN103824070A (en) | Rapid pedestrian detection method based on computer vision | |
| Zhang et al. | Object detection/tracking toward underwater photographs by remotely operated vehicles (ROVs) | |
| CN110490043A (en) | A kind of forest rocket detection method based on region division and feature extraction | |
| CN102034240A (en) | Method for detecting and tracking static foreground | |
| Liang et al. | UAV detection using continuous wave radar | |
| Manikandan et al. | Human object detection and tracking using background subtraction for sports applications | |
| Xia et al. | In situ sea cucumber detection based on deep learning approach | |
| CN112465870B (en) | Thermal image alarm intrusion detection method and device under complex background | |
| Gupta et al. | Early wildfire smoke detection in videos | |
| CN108921826A (en) | The transmission line of electricity that super-pixel segmentation is combined with deep learning invades object detecting method | |
| Fahn et al. | Abnormal maritime activity detection in satellite image sequences using trajectory features | |
| Muchtar et al. | A unified smart surveillance system incorporating adaptive foreground extraction and deep learning-based classification | |
| CN113920585A (en) | Behavior recognition method and device, equipment and storage medium | |
| Zhou et al. | Dynamic background subtraction using spatial-color binary patterns | |
| Tsetsos et al. | A forest fire detection system: The Meleager approach | |
| Kim et al. | Three plot correlation-based small infrared target detection in dense sun-glint environment for infrared search and track | |
| US20240221181A1 (en) | Image processing system | |
| Apewokin et al. | Embedded real-time surveillance using multimodal mean background modeling | |
| Sawalakhe et al. | Foreground background traffic scene modeling for object motion detection | |
| CN103177237A (en) | Video monitoring method and device based on on-line lasers | |
| Kuppuswamy et al. | Similar Object Detection and Tracking in H. 264 Compressed Video Using Modified Local Self Similarity Descriptor and Particle Filtering. | |
| Tong et al. | Human positioning based on probabilistic occupancy map | |
| Mishra et al. | Landslide detection system using computer vision approach and Raspberry Pi |
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 |