CN108734103B - Method for detecting and tracking moving target in satellite video - Google Patents
Method for detecting and tracking moving target in satellite video Download PDFInfo
- Publication number
- CN108734103B CN108734103B CN201810358177.7A CN201810358177A CN108734103B CN 108734103 B CN108734103 B CN 108734103B CN 201810358177 A CN201810358177 A CN 201810358177A CN 108734103 B CN108734103 B CN 108734103B
- Authority
- CN
- China
- Prior art keywords
- target
- vehicle
- tracking
- moving target
- image
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 230000033001 locomotion Effects 0.000 claims abstract description 23
- 238000001514 detection method Methods 0.000 claims abstract description 15
- 239000013598 vector Substances 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 10
- 230000000877 morphologic effect Effects 0.000 claims description 9
- 230000001133 acceleration Effects 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 2
- 238000012937 correction Methods 0.000 claims description 2
- 230000001186 cumulative effect Effects 0.000 claims description 2
- 238000009795 derivation Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 abstract description 9
- 238000012545 processing Methods 0.000 abstract description 4
- 230000003287 optical effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000033228 biological regulation Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/207—Analysis of motion for motion estimation over a hierarchy of resolutions
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of image and video processing, and particularly relates to a method for detecting and tracking a moving target in a satellite video. The method comprises the following steps: detecting a moving target, namely a vehicle, in a city in a satellite video by adopting a detection technology based on motion and local clutter modeling; identifying the detected moving target by adopting an identification technology based on region growing and multiple clues; and associating the detected multiple vehicle targets with the multiple tracks being tracked by adopting a target tracking technology based on a Kalman filter, and further tracking the ground moving target. The invention has strong practicability, can be used in the fields of traffic monitoring, city monitoring, military monitoring and the like, and has wide application prospect.
Description
Technical Field
The invention belongs to the technical field of image processing and video processing, and particularly relates to a method for detecting and tracking a moving target in a satellite video.
Background
Earth Observation (EO) has a wide field of view, can reach the observation range of city size, and is a hot spot of current research and a future research trend. The method plays a vital role in civil and military fields such as urban traffic systems, marine monitoring, air spy, battlefield monitoring and the like. Many of the current observation techniques including optical satellite imaging, satellite-borne synthetic aperture radar and imaging in the aerospace field have evolved and are of great interest. Both optical satellite images and satellite-borne SAR can observe a wide field of view with high resolution. However, the optical satellite image and the SAR image are static pictures, and cannot observe the dynamic change of an object, thereby limiting the application range thereof.
As Unmanned Aerial Vehicles (UAVs) have evolved, more aerial imagery may be used. However, aerial imagery must rely on complex stabilization processes due to the effects of fluctuations in platform instability. For target tracking, image registration needs to be implemented to separate camera motion from target motion. The height of taking photo by plane is limited to according to new civilian aviation regulation, unmanned aerial vehicle's application scope is compressed for the video scope that the video that obtains of taking photo by plane reduces greatly.
Satellite video is an advanced monitoring technology that uses optical sensors to acquire continuous images on a moving satellite platform. Compare with unmanned aerial vehicle's aerial image, except wider field of view, make the satellite video more stable than the aerial video because the stability of satellite motion. Another advantage of satellite video is that it can always aim the shots at the region of interest (ROI) by flying itself.
Disclosure of Invention
The invention aims to provide a method for detecting and tracking a moving target in a satellite imaging video (hereinafter referred to as a satellite video).
The invention is the first algorithm specially aiming at satellite video processing, and can automatically realize the detection and tracking of vehicle targets in the satellite video. The invention provides a new moving target detection algorithm and a new false alarm target elimination algorithm.
The invention provides a method for detecting and tracking a moving target in a satellite video, wherein the moving target mainly aims at a vehicle, and the method comprises the following specific steps:
(1) detecting a moving target, namely a vehicle, in a city in a satellite video by adopting a detection technology based on motion and local clutter modeling;
(2) identifying the detected moving target by adopting an identification technology based on region growing and multiple clues;
(3) the method comprises the steps that a multi-target tracking technology based on a Kalman filter is adopted, detected states of a plurality of vehicle targets are associated with a plurality of tracks which are being tracked, and the ground moving target is further tracked; here, the object refers to a vehicle detected in an arbitrary frame, the object state is a 6-dimensional state vector composed of its position, speed and acceleration, the trajectory refers to a set of state vectors having a certain vehicle object for a continuous period of time, and an arbitrary trajectory has a unique number of the entire video.
In the step (1), the detection technology based on motion and local clutter modeling assumes that an original image is the sum of an image signal and a noise signal, so an inter-frame difference image (an image obtained by subtracting two frame images and then taking an absolute value) represents the noise signal, an abnormal pixel in the difference image represents a position where a vehicle target may exist, an exponential distribution is selected to fit the noise clutter distribution, a binarization threshold value is calculated, binarization of the inter-frame difference image is realized, a foreground in the binary image represents the vehicle target, a background in the binary image represents an image background area, and the current position of the target is determined by using the intersection of two continuous binary images.
In the step (2), the identification technology based on the region growing and the multi-clue reconstructs the geometric shape of the moving object, and utilizes various morphological clues to identify the moving object and the false alarm object.
In the step (3), the target tracking technology based on the Kalman filter describes the motion of the vehicle by using a linear model, estimates the current state of the moving target by combining the observed value and the predicted value, and updates the system state.
In the step (3), the detected association between the plurality of vehicle targets and the plurality of tracks being tracked is realized by using a Hungarian algorithm to optimally match the vehicle targets and the tracks, matching the rear part of the targets with the upper part of the tracks after matching is completed, initializing a new tracking track for the targets which are not matched with the upper tracks, and searching the matched targets again for the tracks which are not matched with the upper tracks by using a nearest neighbor search technology.
The algorithm of each step is further described as follows:
1. the detection technology based on motion and local clutter modeling specifically comprises the following steps:
(1) rasterization, namely: in each frame, dividing an image into two-dimensional grids along the vertical and horizontal directions, wherein each grid unit comprises a plurality of pixels, and a detection algorithm is respectively executed in each unit;
(1a) the image is considered as the sum of the original image signal and the noise signal, i.e.:
Gi(x，y)＝gi(x，y)+ni(x，y) (1)
wherein G isi(x, y) is the amplitude of the pixel (x, y) in the ith frame of video, gi(x, y) is the amplitude of the point (x, y) in the original image in the ith frame of video, ni(x, y) magnitude of the corresponding noise signal;
(1b) selecting an exponential distribution to fit to the distribution of the noise signal, the Cumulative Density Function (CDF) of the exponential distribution being:
wherein, cE(. x) is the CDF of the exponential distribution, and λ is a parameter of the exponential distribution.
(1c) Finding out an area with large gray level change as a target, presetting a false alarm rate, and calculating a binary threshold value:
wherein,
is an inverse function, pfaIs a preset probability value;
(1d) calculating a binarization threshold value by using a predefined probability value to classify pixels, wherein the pixel amplitude value is larger than th and is a foreground, the pixel amplitude value is smaller than th and is a background, and in the binary image, '1' is the foreground and '0' is the background;
(1e) and taking an intersection image of the two obtained binary images, namely performing AND operation pixel by pixel to eliminate scene targets at the front moment and the rear moment, and thus only keeping the current position of the moving vehicle target.
2. The identification technology based on region growing and multiple clues comprises the following specific steps:
(2a) estimating parameters of Gaussian distribution by using a region growing algorithm based on Gaussian distribution and all foreground pixels of a connected region detected by a detector; the PDF of the Gaussian distribution is;
where μ and σ represent the mean and standard deviation, respectively, and the CDF of the Gaussian distribution is:
wherein erf () represents the correlation error function;
(2b) predefined lower boundary probability
And upper bound probability
The lower and upper bounds of the amplitude values are calculated separately, i.e.:
wherein,
and
a lower bound and an upper bound representing a threshold, respectively;
(2c) re-judging the foreground object in the area with amplitude value
Pixels within the range are foreground;
(2d) all pixels communicated with the original foreground are reserved to construct a new communicated region, namely a region growing result;
(2e) and finally, performing morphological identification on the region growing result by taking the area, the duty ratio, the length and the eccentricity as morphological clues so as to eliminate non-vehicle targets in the detected targets, wherein the definition of the duty ratio is the ratio of the area of the connected region pixels to the area of the connected region boundary frame.
3. The target tracking technology based on the Kalman filter comprises the following specific steps:
(3a) the KF is generally suitable for target tracking of linear motion, while the nonlinear motion can be generally decomposed into a series of linear motions, firstly, only considering the KF tracking problem of a moving target, and the dynamic equation of a prediction system is as follows:
xi＝Fi·xi-1+vi (8)
wherein, FiFor the evolution matrix at the ith frame time of the satellite video, viIs a noise vector, xi-1Is the state vector, x, at the instant of the i-1 th frameiIs the state vector of the current moment;
state vector xiIs defined as:
xi＝[x，y，vx，vy，ax，ay]T (9)
wherein, x, y, vx，vy，ax，ayRespectively showing the position in the X-axis direction, the position in the Y-axis direction, the speed in the X-axis direction, the speed in the Y-axis direction, the acceleration in the X-axis direction and the acceleration in the Y-axis direction;
suppose a vehicle object is moving at a time between two framesLinear motion with uniform acceleration of the rows, so that the matrix F evolvesiComprises the following steps:
where τ represents the time interval between two frames in the video, in "seconds", i.e., the reciprocal of the frame rate.
(3b) The area and the position of the detected vehicle are used as observed values, and an observation equation is obtained as follows:
wherein, yi,HiAnd niRespectively representing a measurement vector, a measurement matrix and measurement noise;
observation matrix HiComprises the following steps:
(3c) derivation of the State vector x by Bayes' theoremiThe posterior Probability Density Function (PDF) is simplified by using a chain rule, that is:
wherein the probability p (x) is derived by the Chapman-Kolmogorov equationi|y1：i-1) Namely:
p(xi|y1：i-1)＝∫p(xi|xi-1)p(xi-1|y1：i-1)dxi-1 (14)
(3d) using equations (13) and (14), the state vector x is obtainediThe posterior probability density function PDF:
(3e) and (5) estimating the current state by combining the observed value and the predicted value and utilizing a formula (15), and continuously updating the state of the system.
4. The method comprises the following steps of associating a plurality of vehicle detection and tracking steps:
(4a) in a target tracking technology based on a Kalman filter, a motion trail is associated with a moving target through a Hungarian algorithm, and the associated target is tracked;
(4b) and searching the track of the unassociated target for the associated moving target by using the nearest neighbor search based on SSIM (structural similarity index), if the association is successful, continuing to finish tracking after correction, and if the track still has no associated target, terminating the track.
The method has high algorithm precision and practical value, can be used in the fields of traffic monitoring, city monitoring, military monitoring and the like, and has wide application prospect.
Drawings
Fig. 1 is a schematic diagram of an example of satellite video of the present invention. Where (a) the video location is in valencia, spain and (b) is the corresponding optical map downloaded from Google Earth.
FIG. 2 is a flow chart of a detection algorithm of the present invention that utilizes an exponential distribution to fit the noise clutter distribution.
Fig. 3 is a schematic diagram of foreground segmentation results and region growing results for moving objects and false alarms in accordance with the present invention. Wherein (a), (b), (c) and (d) represent four hypotheses, respectively, and the first, second and third lines represent a color map, a foreground segmentation result and a region growing result, respectively. For ease of viewing, the original gray scale map is correspondingly transformed into the image shown in the first row.
FIG. 4 is a schematic representation of a moving vehicle sample and a non-moving vehicle sample of the present invention. Where (a), (b) and (c) are true moving objects, but the black vehicle in (b) has no significant difference from the background, and (d) is a false alarm object, which may be a stationary vehicle similar to a true moving object.
Fig. 5 is a schematic diagram of the annotation region of a satellite video used in the experiments of the present invention.
Fig. 6 is an annotated effect graph of a satellite video used in the experiments of the present invention. Wherein (a) the location of the vehicle is displayed and (b) its corresponding name is displayed. Annotated in the figure are representative scenarios.
FIG. 7 is a schematic diagram of vehicle detection and tracking results in four frames of satellite video according to the present invention: respectively, frame 50, frame 100, frame 150, and frame 200, each vehicle having its own number.
Fig. 8 is a flow chart illustration of the present invention.
Detailed Description
Firstly, a moving target in a satellite video is detected, and a method based on motion and local clutter modeling is utilized, wherein the specific implementation mode is as follows:
the method comprises the following steps: in each frame, dividing an image into two-dimensional grids along the vertical and horizontal directions, wherein each grid unit comprises a plurality of pixels;
step two: considering the image as the sum of the original image signal and the noise signal, and selecting an exponential distribution to fit the distribution of the noise signal;
step three: the region with large gray scale change is found as the target, and the false alarm rate is preset to be 1 × 10-2Calculating a binary threshold value, and classifying the pixels by using the threshold value, namely: if the pixel amplitude value is larger than the threshold, it is foreground, and conversely, it is background (in the binary image, '1' is foreground and '0' is background);
step four: and taking an intersection image of the two obtained binary images, namely performing AND operation pixel by pixel to eliminate scene targets at the front moment and the rear moment, so that the current position of the moving vehicle target is uniquely reserved to obtain a detection result.
Then, identifying the moving target in the detected satellite video, and utilizing a method based on region growing and multi-clue, wherein the specific implementation mode is as follows:
the method comprises the following steps: estimating parameters of Gaussian distribution by using a region growing algorithm based on Gaussian distribution and all foreground pixels of a connected region detected by a detector, and further obtaining PDF of Gaussian distribution;
step two: predefining values of upper resolution probability and lower boundary probability, and designing the lower boundary probability to be 1 × 10-2The upper bound probability is designed to be 1-1 × 10-2Respectively calculating the upper boundary and the lower boundary of the amplitude value, and judging the foreground target of the area again by using a new amplitude boundary value;
step three: all pixels communicated with the original foreground are reserved to construct a new communicated region, and a region growing result is obtained;
step four: the area, the duty ratio, the length and the eccentricity are used as morphological clues, parameters of the morphological clues can be designed according to actual conditions, morphological identification is carried out on the result of the region growing, non-vehicle targets in the detected targets are removed, and identification results are obtained.
Finally, tracking and associating the moving target in the detected and identified satellite video, and utilizing a target tracking method of a Kalman filter, wherein the specific implementation mode is as follows:
the method comprises the following steps: firstly, only considering the KF tracking problem of a moving target, predicting a dynamic equation of a system, and obtaining an evolution matrix by using a state vector;
step two: obtaining an observation equation by using the area and the position of the detected vehicle as observation values, and obtaining an observation matrix by using an observation vector;
step three: deducing a posterior probability density function of the state vector by a Bayes rule, and simplifying the posterior probability density function by utilizing a chain rule;
step four: estimating the current state by combining the observed value and the predicted value, and updating the system state;
step five: the method comprises the steps of associating a detected vehicle target with a track being tracked by using a Hungarian algorithm, updating the state of the corresponding track by using the state of the target if the detected vehicle target can be associated with the track, initializing a new track and tracking if the target does not associate with the track, searching the corresponding target again by using an SSIM algorithm if the track does not associate with the target, updating the state of the track by using the state of the searched target if the track can be searched, and terminating the tracking of the track if the track cannot be searched.
The algorithm was tested on the video shown in fig. 1, with a frame rate of 20 frames/second, a resolution of 1 meter, a duration of 29 seconds, an image height of 3072 pixels, and an image width of 4096 pixels. As shown in FIG. 6, some regions are manually marked, and the data manually marked by the regions are compared with the result output by the algorithm so as to quantitatively evaluate the performance of the algorithm, and an evaluation test shows that the algorithm has 63.06% recall rate, 81.04% accuracy rate, high algorithm precision and practical value.
Claims (5)
1. A method for detecting and tracking a moving target in a satellite video is characterized by comprising the following specific steps:
(1) detecting a moving target, namely a vehicle, in a city in a satellite video by adopting a detection technology based on motion and local clutter modeling;
(2) identifying the detected moving target by adopting an identification technology based on region growing and multiple clues;
(3) a multi-target tracking technology based on a Kalman filter is adopted, a plurality of detected vehicle targets are associated with a plurality of tracks which are being tracked, and the ground moving target is further tracked; the target refers to a vehicle detected in any frame, the target state refers to a 6-dimensional state vector consisting of the position, the speed and the acceleration of the vehicle, the track refers to a set of state vectors of a certain vehicle target in a continuous time, and any track has a unique serial number of the whole video;
in the step (1), the detection technology based on motion and local clutter modeling is used for detecting a moving target, namely a vehicle, in a city in a satellite video, assuming that an original image is the sum of an image signal and a noise signal, an interframe difference image, namely an image obtained by subtracting two frames of images and then taking an absolute value to represent the noise signal, an abnormal pixel in the difference image represents a position where the vehicle target possibly exists, an exponential distribution is selected to fit the noise clutter distribution, a binarization threshold value is calculated, binarization of the interframe difference image is realized, a foreground in the binary image represents the vehicle target, a background in the binary image represents an image background area, and the intersection of two continuous binary images is used for determining the current position of the target;
in the step (2), the identification technology based on the region growing and the multiple clues identifies the detected moving target, namely reconstructs the geometric shape of the moving target and utilizes various morphological clues to identify the moving target and the false alarm target;
in the step (3), the target tracking technology based on the Kalman filter describes the motion of the vehicle by using a linear model, estimates the current state of the moving target by combining an observed value and a predicted value, and updates the state of the system;
in the step (3), the detected association between the plurality of vehicle targets and the plurality of tracks being tracked is realized by using a Hungarian algorithm to optimally match the vehicle targets and the tracks, matching the rear part of the targets with the upper part of the tracks after matching is completed, initializing a new tracking track for the targets which are not matched with the upper tracks, and searching the matched targets again for the tracks which are not matched with the upper tracks by using a nearest neighbor search technology.
2. The method for detecting and tracking the moving target in the satellite video according to claim 1, wherein the detection based on the motion and local clutter modeling in the step (1) comprises the following specific steps:
(2a) rasterization, namely: in each frame, dividing an image into two-dimensional grids along the vertical and horizontal directions, wherein each grid unit comprises a plurality of pixels, and a detection algorithm is respectively executed in each unit;
(2b) the image is considered as the sum of the original image signal and the noise signal, i.e.:
Gi(x，y)＝gi(x，y)+ni(x，y) (1)
wherein G isi(x, y) is the amplitude of the pixel (x, y) in the ith frame of video, gi(x, y) is the amplitude of the point (x, y) in the original image in the ith frame of video, ni(x, y) magnitude of the corresponding noise signal;
(2c) the exponential distribution is chosen to fit the distribution of the noise signal, the cumulative density function CDF of the exponential distribution being:
wherein, cE() CDF of exponential distribution, λ is a parameter of exponential distribution;
(2d) finding out an area with large gray level change as a target, presetting a false alarm rate, and calculating a binary threshold value:
wherein,
is an inverse function, pfaIs a preset probability value;
(2e) calculating a binarization threshold value by using a predefined probability value to classify pixels, wherein the pixel amplitude value is larger than th and is a foreground, the pixel amplitude value is smaller than th and is a background, and in the binary image, '1' is the foreground and '0' is the background;
(2f) and taking an intersection image of the two obtained binary images, namely performing AND operation pixel by pixel to eliminate scene targets at the front moment and the rear moment, and thus only keeping the current position of the moving vehicle target.
3. The method for detecting and tracking a moving object in a satellite video according to claim 1, wherein the step (2) based on region growing and multi-cue identification comprises the following specific steps:
(3a) estimating parameters of Gaussian distribution by using a region growing algorithm based on Gaussian distribution and all foreground pixels of a connected region detected by a detector; the PDF of the Gaussian distribution is;
where μ and σ represent the mean and standard deviation, respectively, and the CDF of the Gaussian distribution is:
wherein erf () represents the correlation error function;
(3b) predefined lower boundary probability
And upper bound probability
The lower and upper bounds of the amplitude values are calculated separately, i.e.:
wherein,
and
a lower bound and an upper bound representing a threshold, respectively;
(3c) re-judging the foreground object in the area with amplitude value
Pixels within the range are foreground;
(3d) all pixels communicated with the original foreground are reserved to construct a new communicated region, namely a region growing result;
(3e) and finally, performing morphological identification on the region growing result by taking the area, the duty ratio, the length and the eccentricity as morphological clues so as to eliminate non-vehicle targets in the detected targets, wherein the definition of the duty ratio is the ratio of the area of the connected region pixels to the area of the connected region boundary frame.
4. The method for detecting and tracking the moving target in the satellite video according to claim 1, wherein the step (3) of target tracking based on the kalman filter KF comprises the following specific steps:
(4a) the KF is suitable for target tracking of linear motion, decomposes nonlinear motion into a series of linear motion, only considers the KF tracking problem of a moving target at first, and the dynamic equation of the prediction system is as follows:
xi＝Fi·xi-1+vi (8)
wherein, FiFor the evolution matrix at the ith frame time of the satellite video, viIs a noise vector, xi-1Is the state vector, x, at the instant of the i-1 th frameiIs the state vector of the current moment;
state vector xiIs defined as:
xi＝[x，y，vx，vy，ax，ay]T (9)
wherein, x, y, vx，vy，ax，ayRespectively showing the position in the X-axis direction, the position in the Y-axis direction, the speed in the X-axis direction, the speed in the Y-axis direction, the acceleration in the X-axis direction and the acceleration in the Y-axis direction;
assuming that the vehicle object makes a linear motion with uniform acceleration at a time between two frames, the matrix F evolvesiComprises the following steps:
wherein, τ represents the time interval between two frames in the video, and the unit is "second", that is, the reciprocal of the frame rate;
(4b) the area and the position of the detected vehicle are used as observed values, and an observation equation is obtained as follows:
wherein, yi,HiAnd niRespectively representing a measurement vector, a measurement matrix and measurement noise;
observation matrix HiComprises the following steps:
(4c) derivation of the State vector x by Bayes' theoremiThe posterior probability density function PDF is simplified by utilizing a chain rule, namely:
wherein the probability p (x) is derived by the Chapman-Kolmogorov equationi|y1：i-1) Namely:
p(xi|y1：i-1)＝∫p(xi|xi-1)p(xi-1|y1：i-1)dxi-1 (14)
(4d) using equations (13) and (14), the state vector x is obtainediThe posterior probability density function PDF:
(4e) and (5) estimating the current state by combining the observed value and the predicted value and utilizing a formula (15), and continuously updating the state of the system.
5. The method for detecting and tracking the moving target in the satellite video according to claim 1, wherein the step (3) of associating the plurality of detected vehicle targets with the plurality of tracks being tracked comprises the following specific steps:
(5a) in a target tracking technology based on a Kalman filter, a motion trail is associated with a moving target through a Hungarian algorithm, and the associated target is tracked;
(5b) and searching the track of the unassociated target for the associated moving target again by using nearest neighbor search based on SSIM, if the association is successful, continuing to finish tracking after correction, and if the track still has no associated target, terminating the track.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810358177.7A CN108734103B (en) | 2018-04-20 | 2018-04-20 | Method for detecting and tracking moving target in satellite video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810358177.7A CN108734103B (en) | 2018-04-20 | 2018-04-20 | Method for detecting and tracking moving target in satellite video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108734103A CN108734103A (en) | 2018-11-02 |
| CN108734103B true CN108734103B (en) | 2021-08-20 |
Family
ID=63939136
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810358177.7A Active CN108734103B (en) | 2018-04-20 | 2018-04-20 | Method for detecting and tracking moving target in satellite video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108734103B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109712169A (en) * | 2018-11-15 | 2019-05-03 | 上海卫星工程研究所 | Moving-target motion profile prediction technique and method for autonomous tracking based on EO-1 hyperion |
| CN109708627B (en) * | 2018-11-15 | 2022-10-18 | 上海航天控制技术研究所 | Method for rapidly detecting space dynamic point target under moving platform |
| CN109816606B (en) * | 2019-01-18 | 2022-01-04 | 中国科学院空天信息创新研究院 | Method for tracking target by using optical remote sensing satellite |
| CN110057449B (en) * | 2019-04-29 | 2021-06-18 | 上海卫星工程研究所 | Air target height and speed calculation method based on field-of-view spectroscopic hyperspectral detection |
| CN110969643B (en) * | 2019-12-18 | 2023-04-25 | 中国人民解放军国防科技大学 | On-board autonomous prediction method for ground target movement track |
| CN111369570B (en) * | 2020-02-24 | 2023-08-18 | 成都空御科技有限公司 | Multi-target detection tracking method for video image |
| CN111462488B (en) * | 2020-04-01 | 2021-09-10 | 北京工业大学 | Intersection safety risk assessment method based on deep convolutional neural network and intersection behavior characteristic model |
| CN112182133B (en) * | 2020-09-29 | 2022-02-15 | 南京北斗创新应用科技研究院有限公司 | AIS data-based ship loitering detection method |
| CN112990124B (en) * | 2021-04-26 | 2021-08-06 | 湖北亿咖通科技有限公司 | Vehicle tracking method and device, electronic equipment and storage medium |
| CN113283330B (en) * | 2021-05-20 | 2024-10-29 | 南京航空航天大学 | Video SAR moving target detection method based on deep learning and multi-target tracking algorithm |
| CN115480273A (en) * | 2022-10-21 | 2022-12-16 | 北京汇通天下物联科技有限公司 | Vehicle positioning system detection method and device |
| CN116030633B (en) * | 2023-02-21 | 2023-06-02 | 天津汉云工业互联网有限公司 | Vehicle tunnel early warning method and device |
| CN117315498B (en) * | 2023-10-10 | 2024-05-24 | 中国人民解放军战略支援部队航天工程大学 | False alarm discrimination method based on space target detection result |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105374026A (en) * | 2015-10-09 | 2016-03-02 | 上海海事大学 | A maritime infrared small target detection method suitable for coastal defense monitoring |
| WO2016085552A1 (en) * | 2014-10-31 | 2016-06-02 | Raytheon Company | System for real-time moving target detection using vision based image segmentation |
| CN106846374A (en) * | 2016-12-21 | 2017-06-13 | 大连海事大学 | The track calculating method of vehicle under multi-cam scene |
| CN107292911A (en) * | 2017-05-23 | 2017-10-24 | 南京邮电大学 | A kind of multi-object tracking method merged based on multi-model with data correlation |
| CN107730515A (en) * | 2017-10-12 | 2018-02-23 | 北京大学深圳研究生院 | Panoramic picture conspicuousness detection method with eye movement model is increased based on region |
-
2018
- 2018-04-20 CN CN201810358177.7A patent/CN108734103B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016085552A1 (en) * | 2014-10-31 | 2016-06-02 | Raytheon Company | System for real-time moving target detection using vision based image segmentation |
| CN105374026A (en) * | 2015-10-09 | 2016-03-02 | 上海海事大学 | A maritime infrared small target detection method suitable for coastal defense monitoring |
| CN106846374A (en) * | 2016-12-21 | 2017-06-13 | 大连海事大学 | The track calculating method of vehicle under multi-cam scene |
| CN107292911A (en) * | 2017-05-23 | 2017-10-24 | 南京邮电大学 | A kind of multi-object tracking method merged based on multi-model with data correlation |
| CN107730515A (en) * | 2017-10-12 | 2018-02-23 | 北京大学深圳研究生院 | Panoramic picture conspicuousness detection method with eye movement model is increased based on region |
Non-Patent Citations (3)
| Title |
|---|
| "基于帧差法和自适应阂值区域生长的红外运动目标检测";李冰 等;《光电技术应用》;20170228;正文第156-160页 * |
| Frame Difference And Kalman Filter Techniques For Detection Of Moving Vehicles In Video Surveillance;C.Srinivas Rao,et al;《International Journal of Engineering Research and Applications》;20121231;正文第1168-1170页 * |
| 基于O2O学习的多目标检测与跟踪技术研究;车少帅;《中国优秀硕士学位论文全文数据库 信息科技辑》;20160515;I136-599 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108734103A (en) | 2018-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734103B (en) | Method for detecting and tracking moving target in satellite video | |
| Zhang et al. | Eye in the sky: Drone-based object tracking and 3d localization | |
| CN111693972B (en) | Vehicle position and speed estimation method based on binocular sequence images | |
| CN110472553B (en) | Target tracking method, computing device and medium for fusion of image and laser point cloud | |
| CN113269098A (en) | Multi-target tracking positioning and motion state estimation method based on unmanned aerial vehicle | |
| TWI393074B (en) | Apparatus and method for moving object detection | |
| CN110866887A (en) | Target situation fusion sensing method and system based on multiple sensors | |
| KR101628155B1 (en) | Method for detecting and tracking unidentified multiple dynamic object in real time using Connected Component Labeling | |
| CN108108697B (en) | Real-time unmanned aerial vehicle video target detection and tracking method | |
| CN103149939A (en) | Dynamic target tracking and positioning method of unmanned plane based on vision | |
| EP1949339A2 (en) | Tracking using an elastic cluster of trackers | |
| EP3593322B1 (en) | Method of detecting moving objects from a temporal sequence of images | |
| CN110617802A (en) | Satellite-borne moving target detection and speed estimation method | |
| CN117095029B (en) | Method and device for detecting small target in air flight | |
| Ersoy et al. | Interactive target tracking for persistent wide-area surveillance | |
| CN111311640B (en) | Unmanned aerial vehicle identification and tracking method based on motion estimation | |
| Pollard et al. | GM-PHD filters for multi-object tracking in uncalibrated aerial videos | |
| Kim | Aerial map-based navigation using semantic segmentation and pattern matching | |
| Cigla et al. | Image-based visual perception and representation for collision avoidance | |
| CN116862832A (en) | Three-dimensional live-action model-based operator positioning method | |
| Ao et al. | Detecting tiny moving vehicles in satellite videos | |
| Farooq et al. | Efficient object proposals extraction for target detection in VHR remote sensing images | |
| Siew et al. | Simultaneous localization and mapping with moving object tracking in 3d range data using probability hypothesis density (phd) filter | |
| Chen et al. | Research on optimization of visual navigation algorithm for real-time obstacle avoidance of uav moving target tracking | |
| Fehlmann et al. | Application of detection and recognition algorithms to persistent wide area surveillance |
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 |