CN113034533B - Infrared small target detection method based on space-time stationarity - Google Patents
Infrared small target detection method based on space-time stationarity Download PDFInfo
- Publication number
- CN113034533B CN113034533B CN202110376498.1A CN202110376498A CN113034533B CN 113034533 B CN113034533 B CN 113034533B CN 202110376498 A CN202110376498 A CN 202110376498A CN 113034533 B CN113034533 B CN 113034533B
- Authority
- CN
- China
- Prior art keywords
- image
- result
- space
- target
- stationarity
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 230000011218 segmentation Effects 0.000 claims abstract description 20
- 238000010586 diagram Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000001914 filtration Methods 0.000 claims abstract description 14
- 230000000877 morphologic effect Effects 0.000 claims abstract description 10
- 238000005070 sampling Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000010339 dilation Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 230000005855 radiation Effects 0.000 abstract description 3
- 230000004927 fusion Effects 0.000 abstract description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 230000007123 defense Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/10048—Infrared image
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an infrared small target detection method based on space-time stationarity, and relates to the technical field of infrared image processing and target detection. The method is based on the stationarity of background components in the infrared small target image in space and time, the time average of a plurality of frames before the current frame is obtained, the average result is subtracted from the current frame and threshold segmentation is carried out, then impulse noise in the segmentation result is filtered, and morphological closed operation is carried out, so that a time stationarity target response graph is obtained. And carrying out rapid edge-preserving filtering on the current frame to obtain a space stationarity target response diagram. And fusing the two target response graphs by using AND operation, and analyzing a connected domain of a fusion result to further obtain the estimation of the target centroid position. And finally, marking the detection result in the current frame. The method has better algorithm real-time performance and detection accuracy, and has better robustness for detecting the infrared small target under the complex conditions of radiation intensity change, target scale change, background clutter interference and the like.
Description
Technical Field
The invention relates to the technical field of infrared image processing and target detection, in particular to an infrared small target detection method based on space-time stationarity.
Background
The infrared target detection is a key technology in an infrared detection system, has very important significance and value in early warning systems, sea defense systems, air defense systems and the like in the field of military reconnaissance, and is also widely applied to civil fields such as medical imaging, robots, automatic driving, traffic management and the like. However, since the infrared imaging system is usually far from the target, the imaging of the target on the image is small, usually only a few or dozens of pixels, and the infrared light is absorbed and scattered by the atmosphere, the contrast and the signal-to-noise ratio of the infrared target in the image are often low, and the infrared target is easily submerged in noise points and background clutter, which brings great difficulty to the detection work, and the signal-to-noise ratio refers to the ratio of the peak-to-peak value of the output signal of the brightness channel of the camera to the effective value of the video clutter under the standard illumination.
Existing infrared small target detection techniques are typically based on modeling of infrared images. It is generally considered that an infrared image is composed of three parts of a target signal, a background signal and a noise signal. The target signal is usually blurred in edge and lacks effective features for description, but the local contrast is relatively high and is weakly correlated with the background; the background signal is basically stable in space and time, has the characteristic of slowly changing low frequency, and has high correlation with the surrounding background signal; the noise signals are generally randomly distributed in time and space, independent of the background image, and the inter-frame distribution is not correlated.
In the prior art, the technology based on infrared small target detection is mainly divided into two types: track Before Detection (TBD) and Track Before Detection (DBT). The TBD algorithm generates a plurality of candidate tracks by tracking the small target, then gradually eliminates false tracks through a certain criterion, and finally obtains the estimation of the target track and the position; the DBT algorithm first estimates and suppresses the background, minimizes the influence of background clutter, and separates the target from the background and noise by threshold segmentation.
The above two methods have disadvantages: the existing DBT algorithm has poor robustness and high false alarm rate in complex environments such as uneven radiation, background clutter interference and the like; the TBD algorithm usually has a large amount of calculation and poor real-time performance.
Disclosure of Invention
The invention provides an infrared small target detection method based on space-time stationarity, aiming at solving the problems of poor instantaneity, high false alarm rate and the like of an infrared small target detection algorithm in complex environments of uneven radiation, background clutter interference and the like in the prior art, and aiming at: the signal-to-clutter ratio of the target is improved, real-time detection of the infrared small target is realized, and the robustness and accuracy of detection are improved.
In order to achieve the purpose, the invention adopts the following technical scheme:
an infrared small target detection method based on space-time stationarity comprises the following steps:
step A: acquiring an image of the infrared small target detection, reading an nth frame of image, preprocessing the nth frame of image, and graying the nth frame of image to obtain a processing result;
and B: acquiring k frame images before the nth frame, calculating the time average of the k frame images before the nth frame, and performing difference calculation on the calculated time average result and the result in the step A to obtain a difference result;
and C: b, threshold segmentation is carried out on the difference result in the step B, impulse noise is filtered out, a filtering result is obtained, then morphological closed operation is carried out on the filtering result, and the obtained morphological closed operation result is used as a time stationarity target response diagram;
step D: b, forming a plane-gray three-dimensional space by taking the gray value as an intensity dimension, mapping the result in the step A to the plane-gray three-dimensional space, and performing average down-sampling to obtain a down-sampling result;
step E: d, performing three-dimensional Gaussian filtering on the down-sampling result in the step D, and performing linear up-sampling on the result obtained after the three-dimensional Gaussian filtering to obtain an up-sampling result;
step F: carrying out normalization processing on the up-sampling result, and then carrying out threshold segmentation to obtain a space stability target response graph;
step G: performing AND operation on the time stationarity target response graph and the space stationarity target response graph to obtain a space-time stationarity target response graph;
step H: and carrying out connected domain analysis on the space-time stationarity target response graph to obtain estimation on the position of the target centroid.
Preferably, the step B includes:
step B.1: defining the gray value of the pixel point positioned in (X, Y) in the nth frame as fn(X, Y), calculating the average value of the gray values of the pixel points of the k frame images before the nth frame, wherein the pixel points are located at (X, Y), and the formula is as follows:
step B.2: then, calculating the pixel value difference image Diff of the mean value of the current n frame image and the previous k frame image, wherein the formula is as follows:
preferably, the specific steps of step C are:
step C.1: and carrying out threshold segmentation on the difference image Diff:
step C.1.1: calculating the maximum pixel value μ in the difference image Diff as max (Diff (X, Y));
step C.1.2: then, threshold segmentation is carried out on the Diff of the differential image, and the formula is as follows:
where α is a constant and μ represents the largest pixel value in the difference image Diff; finally, obtaining an image g subjected to threshold segmentation;
step C.2: then, the obtained image g is subjected to impulse noise filtering:
step C.2.1: designating a window with the size of c X c, which is used for performing sliding window calculation from top to bottom and from left to right on the image g, then obtaining a sliding window with a certain pixel point as a center through the value g (X, Y) of the pixel point in the image g, then sequencing the pixel point values in the range of the sliding window, then replacing the value g (X, Y) of the pixel point with the sequenced median value of the pixel point values, and finally obtaining an image result h after replacing the value of each pixel point in the image g according to the replacement method;
step C.3: and performing morphological closed operation on the image result h:
step C.3.1: selecting a rectangular structural element with the size of 3 multiplied by 3 as se, and setting
In order to etch the operation symbol,
for the sign of the dilation operation, the image result h is eroded and dilated by se as follows:
wherein D ishAnd DseThe definition domains are respectively an image result h and a rectangular structural element se;
step C.3.2: and (3) performing closed operation on the image result h by using a rectangular structural element se to obtain a time stationarity target response diagram T, wherein the formula is as follows:
preferably, the specific steps of step D are:
step D.1: mapping the current frame image to a plane-gray scale three-dimensional space to form a three-dimensional homogeneous vector (wi, w), wherein the formula is as follows:
i(X，Y，I)＝fn(X，Y)
w(X，Y，I)＝δ(I-fn(X，Y))
step D.2: calculating the minimum pixel value I in the current frame imagemin＝min(fi(X，Y))；
Step D.3: coordinate (X, Y, f)n(X, Y)) to a down-sampled form (X, Y, ζ):
wherein [. ]]For the rounding operator, ssSampling rate, s, for spatial dimensionsrZeta is the intensity dimension coordinate after down-sampling;
step D.4: the average down-sampled result of the vector (wi, w) is set as the vector (w)↓i↓，w↓) A vector (w)↓i↓，w↓) Initialized to a zero vector and then paired with the vector (w)↓i↓，w↓) Updating:
(w↓i↓(x，y，ζ)，w↓(x，y，ζ))＝(wi(X，Y，I)，w(X，Y，I))
wherein the coordinate relationship of (X, Y, ζ) to (X, Y, I) is given by the formula in step d.3.
Preferably, the specific steps of step E are:
step E.1: vector (w)↓i↓，w↓) And three-dimensional Gaussian nucleus
Performing convolution to obtain vector
The formula is as follows:
wherein, in
In the method, a coordinate system is established by taking a geometric center of a Gaussian kernel as an origin, and the coordinate is the value of an element at the position of (m, n, epsilon), as shown in a formula:
step E.2: for vector
Linear interpolation is performed to obtain an upsampled original size result vector (W)bIb，Wb)。
Preferably, the specific steps of step F are:
step F.1: for the result vector (W)bIb，Wb) Normalized to obtain IbThen, the result graph f is obtainedbThe formula is as follows:
Ib(X，Y，I)＝WbIb(X，Y，I)/Wb(X，Y，I)
fb(X，Y)＝Ib(X，Y，I)
step F.2: for the result chart fbPerforming threshold segmentation:
step F.2.1: graph f of calculation resultsbInner maximum pixel value mub＝max(fb(X，Y))；
Step F.2.2: to fbPerforming threshold segmentation to obtain a threshold-segmented image gb：
Wherein beta is a constant, mubRepresenting a difference image fbThe largest pixel value.
Preferably, the specific steps of step G are:
step G.1: for the image g and the image gbAnd operation is carried out to obtain a response graph G of the space-time stationarity target, which specifically comprises the following steps:
preferably, the specific steps of step H are: and analyzing the connected domain of the space-time stationarity target response graph G, and taking the obtained centroid position of each connected region as the centroid position of the candidate target to obtain a detection result.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
1. the invention comprehensively considers the space-time stationarity characteristics of three signals of a target, a background and noise in an infrared image, combines single-frame and multi-frame information, effectively inhibits background influence and noise interference, obviously improves the detection capability of infrared weak and small targets, and provides a feasible way for solving the problems of poor real-time performance, low robustness in a complex scene and low detection accuracy rate under the conditions of more noise and low signal-to-noise ratio of infrared small target detection based on multiple frames.
2. Aiming at the problem that a large number of false targets exist in infrared small target detection, the method selects candidate targets supported by a time stationarity response diagram and a space stationarity response diagram together with operation, thereby obviously reducing false alarm rate and enhancing detection real-time property. The system can robustly, efficiently and accurately detect the small targets in various complex scenes.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
fig. 2 is an original image for detecting a small infrared target according to embodiment 1 of the present invention;
FIG. 3 is a time response diagram of the detection of small infrared targets in example 1 of the present invention;
FIG. 4 is a spatial response diagram for detecting infrared small targets in embodiment 1 of the present invention;
FIG. 5 is a diagram of a space-time fusion response of infrared small target detection in embodiment 1 of the present invention;
fig. 6 is a graph of the result of detecting a small infrared target in example 1 of the present invention.
Detailed Description
All of the features disclosed in this specification, or all of the steps in any method or process so disclosed, may be combined in any combination, except combinations of features and/or steps that are mutually exclusive.
The invention will be further described with reference to the accompanying drawings and specific embodiments.
Example 1:
step A: acquiring an image of the infrared small target detection, reading an nth frame of image, preprocessing the nth frame of image, and graying the nth frame of image to obtain a processing result;
the image preprocessing mainly refers to processing the image data type, including that a color image is changed into a gray image, and a pixel value is converted into a double type.
And B: judging whether the number of the nth frame is greater than k, taking k as 5, if not, initializing the time stationarity response diagram as a 1 matrix, namely assigning the value of each pixel of the time stationarity response diagram as 1; if the difference is larger than k, acquiring k frame images before the nth frame, calculating the time average of the k frame images before the nth frame, and carrying out differential calculation on the calculated time average result and the result in the step A to obtain a differential result; the step B comprises the following steps:
step B.1: defining the gray value of the pixel point positioned in (X, Y) in the nth frame as fn(X, Y), calculating the average value of the gray values of the pixel points of the k frame images before the nth frame, wherein the pixel points are located at (X, Y), and the formula is as follows:
step B.2: then, calculating the pixel value difference image Diff of the mean value of the current n frame image and the previous k frame image, wherein the formula is as follows:
and C: b, threshold segmentation is carried out on the difference result in the step B, impulse noise is filtered out, a filtering result is obtained, then morphological closed operation is carried out on the filtering result, and the obtained morphological closed operation result is used as a time stationarity target response diagram; the concrete steps of the step C are as follows:
step C.1: and carrying out threshold segmentation on the difference image Diff:
step C.1.1: calculating the maximum pixel value μ in the difference image Diff as max (Diff (X, Y));
step C.1.2: then, threshold segmentation is carried out on the Diff of the differential image, and the formula is as follows:
wherein alpha is a constant and takes a value of 0.05, and mu represents the maximum pixel value in the differential image DiTf; finally, obtaining an image g subjected to threshold segmentation;
step C.2: then, the obtained image g is subjected to impulse noise filtering:
step C.2.1: designating a window with the size of c multiplied by c, wherein c is set to be 3 and is used for performing sliding window calculation from top to bottom and from left to right on the image g, then obtaining a sliding window with a certain pixel point as a center through the value g (X, Y) of the pixel point in the image g, then sequencing pixel point values in the range of the sliding window, replacing the value g (X, Y) of the pixel point with the sequenced median value of the pixel point, and finally obtaining an image result h after replacing the value of each pixel point in the image g according to the replacement method;
step C.3: and performing morphological closed operation on the image result h:
step C.3.1: selecting a rectangular structural element with the size of 3 multiplied by 3 as se, and setting
In order to etch the operation symbol,
for the sign of the dilation operation, the image result h is eroded and dilated by se as follows:
wherein D ishAnd DseThe definition domains are respectively an image result h and a rectangular structural element se;
step C.3.2: and (3) performing closed operation on the image result h by using a rectangular structural element se to obtain a time stationarity target response diagram T, wherein the formula is as follows:
the closed operation of the rectangular structural element se is used for filling the gap, and the rectangle with the size of 3 multiplied by 3 is selected to achieve a better effect, meanwhile, the operation amount is not too large, and excessive adhesion can be caused if the rectangular structural element se is too large.
Step D: b, forming a plane-gray three-dimensional space by taking the gray value as an intensity dimension, mapping the result in the step A to the plane-gray three-dimensional space, and performing average down-sampling to obtain a down-sampling result; the specific steps of the step D are as follows:
step D.1: mapping the current frame image to a plane-gray scale three-dimensional space to form a three-dimensional homogeneous vector (wi, w), wherein the formula is as follows:
i(X，Y，I)＝fn(X，Y)
w(X，Y，I)＝δ(I-fn(X，Y))
step D.2: calculating the minimum pixel value I in the current frame imagemin＝min(fi(X，Y))；
Step D.3: coordinate (X, Y, f)n(X, Y)) to a down-sampled form (X, Y, ζ):
wherein [. ]]For the rounding operator, ssThe sampling rate for the spatial dimension is taken to be 350, srTaking the sampling rate of the intensity dimension as 0.05, and taking zeta as the intensity dimension coordinate after down sampling;
step D.4: the average down-sampled result of the vector (wi, w) is set as the vector (w)↓i↓，w↓) Will vector (w)↓i↓，w↓) Initialized to a zero vector and then paired with the vector (w)↓i↓，w↓) Updating:
(w↓i↓(x，y，ζ)，w↓(x，y，ζ))＝(wi(X，Y，I)，w(X，Y，I))
wherein the coordinate relationship of (X, Y, ζ) to (X, Y, I) is given by the formula in step d.3.
Step E: d, performing three-dimensional Gaussian filtering on the down-sampling result in the step D, and performing linear up-sampling on the result obtained after the three-dimensional Gaussian filtering to obtain an up-sampling result; the concrete steps of the step E are as follows:
step E.1: vector (w)↓i↓，w↓) And three-dimensional Gaussian nucleus
Performing convolution to obtain a vector
The formula is as follows:
wherein, in
Middle, sigmasTake 350, σrTaking 0.05, establishing a coordinate system by taking the geometric center of the Gaussian kernel as an origin, wherein the coordinate is the value of an element at the position of (m, n, epsilon), and the formula is as follows:
step E.2: for vector
Linear interpolation is performed to obtain an upsampled original size result vector (W)bIb，Wb)。
Step F: carrying out normalization processing on the up-sampling result, and then carrying out threshold segmentation to obtain a space stability target response graph; the specific steps of the step F are as follows:
step F.1: for the result vector (W)bIb，Wb) Normalized to obtain IbThen, the result graph f is obtainedbThe formula is as follows:
Ib(X，Y，I)＝WbIb(X，Y，I)/Wb(X，Y，I)
fb(X，Y)＝Ib(X，Y，I)
step F.2: for the result chart fbPerforming threshold segmentation:
step F.2.1: graph f of calculation resultsbInner maximum pixel value mub＝max(fb(X，Y))；
Step F.2.2: to fbPerforming threshold segmentation to obtain a threshold-segmented image gb：
Wherein beta is a constant, taken as 0.72, mubRepresenting a difference image fbThe largest pixel value.
Step G: performing AND operation on the time stationarity target response graph and the space stationarity target response graph to obtain a space-time stationarity target response graph; the specific steps of the step G are as follows:
step G.1: for the image g and the image gbAnd operation is carried out to obtain a space-time stationarity target response graph G, which specifically comprises the following steps:
step H: and analyzing the connected domain of the space-time stationarity target response graph G, and taking the obtained centroid position of each connected region as the centroid position of the candidate target to obtain a detection result.
The above are merely representative examples of the many specific applications of the present invention, and do not limit the scope of the invention in any way. All the technical solutions formed by the transformation or the equivalent substitution fall within the protection scope of the present invention.
Claims (8)
1. An infrared small target detection method based on space-time stationarity is characterized by comprising the following steps:
step A: acquiring an image of the infrared small target detection, reading an nth frame of image, preprocessing the nth frame of image, and graying the nth frame of image to obtain a processing result;
and B: reading k frame images before an nth frame, calculating the time average of the k frame images before the nth frame, and carrying out difference calculation on the calculated time average result and the processing result in the step A to obtain a difference result;
and C: b, threshold segmentation is carried out on the difference result in the step B, impulse noise is filtered out, a filtering result is obtained, then morphological closed operation is carried out on the filtering result, and the obtained morphological closed operation result is used as a time stationarity target response diagram;
step D: b, forming a plane-gray three-dimensional space by taking the gray value as an intensity dimension, mapping the processing result in the step A to the plane-gray three-dimensional space, and performing average down-sampling to obtain a down-sampling result;
and E, step E: d, performing three-dimensional Gaussian filtering on the down-sampling result in the step D, and performing linear up-sampling on the result obtained after the three-dimensional Gaussian filtering to obtain an up-sampling result;
step F: carrying out normalization processing on the up-sampling result, and then carrying out threshold segmentation to obtain a space stability target response graph;
g: performing AND operation on the time stationarity target response diagram obtained in the step C and the space stationarity target response diagram obtained in the step F to obtain a space-time stationarity target response diagram;
step H: and carrying out connected domain analysis on the space-time stationarity target response graph to obtain a target centroid position.
2. A method for detecting infrared small targets based on space-time stationarity according to claim 1, wherein the step B comprises:
step B.1: defining the gray value of the pixel point positioned in (X, Y) in the nth frame as fn(X, Y), calculating the average value of the gray values of the pixel points of the k frame images before the nth frame, wherein the pixel points are located at (X, Y), and the formula is as follows:
step B.2: then, calculating pixel value difference image Diff of the mean value of the current nth frame image and the previous k frame image, wherein the formula is as follows:
3. a method for detecting small infrared targets based on space-time stationarity according to claim 2, characterized in that the specific steps of step C are:
step C.1: and carrying out threshold segmentation on the difference image Diff:
step C.1.1: calculating the maximum pixel value μ in the difference image Diff as max (Diff (X, Y));
step C.1.2: then, threshold segmentation is carried out on the Diff of the differential image, and the formula is as follows:
where α is a constant and μ represents the largest pixel value in the difference image Diff; finally, obtaining an image g subjected to threshold segmentation;
step C.2: then, the obtained image g is subjected to impulse noise filtering:
step C.2.1: designating a window with the size of c X c, wherein the value range of c is 3-10, and the window is used for performing sliding window calculation from top to bottom and from left to right on the image g, then obtaining a sliding window with a certain pixel point in the image g as a center through the value g (X, Y) of the pixel point, then sequencing the pixel point values in the sliding window range, replacing the value g (X, Y) of the pixel point with the sequenced median value of the pixel point, and finally obtaining an image result h after replacing the value of each pixel point in the image g according to the replacement method;
step C.3: and performing morphological closed operation on the image result h:
step C.3.1: selecting a rectangular structural element with the size of 3 multiplied by 3 as se, and setting
In order to etch the operation symbol,
for the sign of the dilation operation, the image result h is eroded and dilated by se as follows:
wherein D ishAnd DseThe definition domains are respectively an image result h and a rectangular structural element se;
step C.3.2: and (3) performing closed operation on the image result h by using a rectangular structural element se to obtain a time stationarity target response diagram T, wherein the formula is as follows:
4. a method for detecting a small infrared target based on space-time stationarity as claimed in claim 1, wherein said step D comprises the following steps:
step D.1: mapping the current frame image to a plane (gray scale three-dimensional space) to form a three-dimensional homogeneous vector (wi, w), wherein the formula is as follows:
i(X，Y，I)＝fn(X，Y)
w(X，Y，I)＝δ(I-fn(X，Y))
step D.2: calculating the minimum pixel value I in the current frame imagemin＝min(fi(X，Y))；
Step D.3: coordinate (X, Y, f)n(X, Y)) to a down-sampled form (X, Y, ζ):
wherein [. ]]For the rounding operator, ssSampling rate, s, for spatial dimensionsrZeta is the intensity dimension coordinate after down-sampling;
step D.4: the average down-sampled result of the vector (wi, w) is set as the vector (w)↓i↓，w↓) Will vector (w)↓i↓，w↓) Initialized to a zero vector and then paired with the vector (w)↓i↓，w↓) And (3) updating:
(w↓i↓(x，y，ζ)，w↓(x，y，ζ))＝(wi(X，Y，I)，w(X，Y，I))
where the coordinate relationship of (X, Y, ζ) to (X, Y, I) is given by the formula in step d.3.
5. A space-time stationarity-based infrared small target detection method according to claim 4, characterized in that the specific steps of step E are:
step E.1: vector (w)↓i↓，w↓) And three-dimensional Gaussian nucleus
Performing convolution to obtain a vector
The formula is as follows:
wherein, in
In the method, a coordinate system is established by taking a geometric center of a Gaussian kernel as an origin, and the coordinate is the value of an element at the position of (m, n, epsilon), as shown in a formula:
step E.2: for vector
Linear interpolation is performed to obtain an upsampled original size result vector (W)bIb，Wb)。
6. A method for detecting a small infrared target based on space-time stationarity according to claim 5, wherein the specific steps of the step F are specifically:
step F.1: for the result vector (W)bIb，Wb) Normalized to obtain IbThen, the result graph f is obtainedbThe formula is as follows:
Ib(X，Y，I)＝WbIb(X，Y，I)/Wb(X，Y，I)
fb(X，Y)＝Ib(X，Y，I)
step F.2: for the result chart fbPerforming threshold segmentation:
step F.2.1: graph f of calculation resultsbInner maximum pixel value mub＝max(fb(X，Y))；
Step F.2.2: to fbPerforming threshold segmentation, thenObtaining a thresholded image gb：
Wherein beta is a constant, mubRepresenting a difference image fbThe largest pixel value.
7. A method for detecting small infrared targets based on space-time stationarity according to claim 6, wherein the specific steps of the step G are as follows:
step G.1: for the image g and the image gbAnd operation is carried out to obtain a space-time stationarity target response graph G, which specifically comprises the following steps:
8. a method for detecting small infrared targets based on space-time stationarity according to claim 7, wherein the specific steps of the step H are as follows: and analyzing the connected domain of the space-time stationarity target response graph G, and taking the obtained centroid position of each connected region as the centroid position of the candidate target to obtain a detection result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110376498.1A CN113034533B (en) | 2021-04-06 | 2021-04-06 | Infrared small target detection method based on space-time stationarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110376498.1A CN113034533B (en) | 2021-04-06 | 2021-04-06 | Infrared small target detection method based on space-time stationarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113034533A CN113034533A (en) | 2021-06-25 |
| CN113034533B true CN113034533B (en) | 2022-05-20 |
Family
ID=76454144
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110376498.1A Active CN113034533B (en) | 2021-04-06 | 2021-04-06 | Infrared small target detection method based on space-time stationarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113034533B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834915A (en) * | 2015-05-15 | 2015-08-12 | 中国科学院武汉物理与数学研究所 | Small infrared object detection method in complex cloud sky background |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102819740B (en) * | 2012-07-18 | 2016-04-20 | 西北工业大学 | A kind of Single Infrared Image Frame Dim targets detection and localization method |
| CN106251344B (en) * | 2016-07-26 | 2019-02-01 | 北京理工大学 | A kind of multiple dimensioned infrared target self-adapting detecting method of view-based access control model receptive field |
| CN106570889A (en) * | 2016-11-10 | 2017-04-19 | 河海大学 | Detecting method for weak target in infrared video |
| CN109461168B (en) * | 2018-10-15 | 2021-03-16 | 腾讯科技（深圳）有限公司 | Target object identification method and device, storage medium and electronic device |
| CN111222502B (en) * | 2019-12-28 | 2023-05-12 | 中国船舶重工集团公司第七一七研究所 | Infrared small target image labeling method and system |
-
2021
- 2021-04-06 CN CN202110376498.1A patent/CN113034533B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104834915A (en) * | 2015-05-15 | 2015-08-12 | 中国科学院武汉物理与数学研究所 | Small infrared object detection method in complex cloud sky background |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113034533A (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110728697A (en) | Infrared dim target detection tracking method based on convolutional neural network | |
| CN109961506A (en) | A kind of fusion improves the local scene three-dimensional reconstruction method of Census figure | |
| CN110555868A (en) | method for detecting small moving target under complex ground background | |
| CN109859247B (en) | Near-ground scene infrared small target detection method | |
| JP2008523454A (en) | How to model background and foreground regions | |
| CN111145121B (en) | Confidence term filter target tracking method for strengthening multi-feature fusion | |
| CN112364865B (en) | Method for detecting small moving target in complex scene | |
| CN111369570B (en) | Multi-target detection tracking method for video image | |
| CN111709968B (en) | Low-altitude target detection tracking method based on image processing | |
| CN111311644B (en) | Moving target detection method based on video SAR | |
| Cvejic et al. | The effect of pixel-level fusion on object tracking in multi-sensor surveillance video | |
| CN111539980B (en) | Multi-target tracking method based on visible light | |
| Benedek et al. | Moving target analysis in ISAR image sequences with a multiframe marked point process model | |
| CN116229359A (en) | Smoke identification method based on improved classical optical flow method model | |
| CN111161308A (en) | Dual-band fusion target extraction method based on key point matching | |
| CN114648547A (en) | Weak and small target detection method and device for anti-unmanned aerial vehicle infrared detection system | |
| CN112288780B (en) | Multi-feature dynamically weighted target tracking algorithm | |
| CN113205494A (en) | Infrared small target detection method and system based on adaptive scale image block weighting difference measurement | |
| CN113034533B (en) | Infrared small target detection method based on space-time stationarity | |
| CN112613456A (en) | Small target detection method based on multi-frame differential image accumulation | |
| CN116883897A (en) | Low-resolution target identification method | |
| CN109087334B (en) | Target tracking method based on significant context sparse representation | |
| CN114820718A (en) | Visual dynamic positioning and tracking algorithm | |
| CN108389219B (en) | Weak and small target tracking loss re-detection method based on multi-peak judgment | |
| Qi et al. | Fast detection of small infrared objects in maritime scenes using local minimum patterns |
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 |