CN112529896A - Infrared small target detection method and system based on dark channel prior - Google Patents
Infrared small target detection method and system based on dark channel prior Download PDFInfo
- Publication number
- CN112529896A CN112529896A CN202011548661.XA CN202011548661A CN112529896A CN 112529896 A CN112529896 A CN 112529896A CN 202011548661 A CN202011548661 A CN 202011548661A CN 112529896 A CN112529896 A CN 112529896A
- Authority
- CN
- China
- Prior art keywords
- dark channel
- significance
- map
- scale
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 39
- 238000013507 mapping Methods 0.000 claims abstract description 43
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 9
- 238000001228 spectrum Methods 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 12
- 238000002834 transmittance Methods 0.000 claims description 9
- 230000002708 enhancing effect Effects 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 5
- 239000003086 colorant Substances 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 8
- 230000004927 fusion Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 5
- 238000001914 filtration Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 230000002146 bilateral effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000001629 suppression 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
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/94—Dynamic range modification of images or parts thereof based on local image properties, e.g. for local contrast enhancement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- 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)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an infrared small target detection method and system based on dark channel prior, belonging to the technical field of infrared target detection.A dark channel prior method is adopted to preprocess an image to be detected to generate a significance mapping chart; secondly, performing feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map; and finally, selecting the optimal scale of the multi-scale saliency map by using the gray difference and the maximum value of the defined local information entropy as judgment standards, and identifying the small target. The method uses a dark channel prior method to preprocess the picture, so that a single significant target can be detected more accurately, and a plurality of significant targets can also be detected more accurately; the significance of the preprocessed image is enhanced, then, according to a special feature matching algorithm, an accurately matched multi-scale significant image is obtained, and a reliable, stable and accurate infrared small target detection result is obtained.
Description
Technical Field
The invention relates to the technical field of infrared target detection, in particular to an infrared small target detection method and system based on dark channel prior.
Background
In the infrared guidance system technology, infrared small target detection has always been a challenging problem. Small targets tend to be drowned out in complex backgrounds of low signal-to-noise ratio, low contrast. In addition, the infrared small target is often not obvious in characteristic, uncertain in brightness and weak in strength due to the fact that the imaging distance is long in the atmosphere.
At present, the traditional infrared small target detection method can be divided into two types, namely single-frame detection and sequential detection. Such as interframe difference, an optical flow method, three-dimensional direction filtering, Bayesian theory and other sequence detection methods, the effect is better under the condition that the target has prior knowledge of the shape and the position of the adjacent frames. Typical single-frame image detection methods, such as maximum mean and maximum median filtering, two-dimensional least mean square filtering, background regression estimation, morphological methods, bilateral filtering, and the like, can effectively detect targets in a simple background, and are fast and short in initialization time. However, when a small target is submerged in an infrared scene with a highly heterogeneous background, the above single-frame detection and sequential detection algorithms cannot obtain a reliable, stable and accurate small target detection result.
In recent years, the matrix decomposition and feature detection method has obvious advantages in single-frame infrared image detection. For a typical matrix decomposition method, a robust principal component analysis method based on convex optimization is adopted, and a foreground target matrix and a background matrix are accurately separated from an original infrared image. The infrared speckle pattern extracted patch-image (IPI) model summarizes the traditional image model into a new patch-image model constructed based on local patch. The core idea of the IPI model is to divide an infrared image into a patch-image model, separate a foreground target matrix from a background matrix through stable principal component analysis, and finally reconstruct the image. However, due to the defect of sparsity measure based on l1 norm in the IPI model, the target still retains part of background residual edge in strong noise environment.
Disclosure of Invention
The invention aims to provide an infrared small target detection method and system based on dark channel prior, which can accurately detect single or multiple salient targets and obtain an accurately matched multi-scale salient map, so as to solve at least one technical problem in the background technology.
In order to achieve the purpose, the invention adopts the following technical scheme:
on one hand, the invention provides a dark channel prior-based infrared small target detection method, which comprises the following steps:
step S110: preprocessing an image to be detected by adopting a dark channel prior method to generate a significance mapping chart;
step S120: carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map;
step S130: and (3) selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value of the defined local information entropy as a judgment standard, and identifying the small target.
Preferably, the step S110 specifically includes:
generating an initial saliency mapping map for the image to be detected according to the color saliency and the depth saliency of the RGB-D image; generating a central dark channel mapping map from the image to be detected by combining the central bias significance and dark channel prior; and fusing the initial saliency map and the central dark channel map to generate a final saliency map.
Preferably, the generating of the initial saliency map specifically includes:
dividing the image into a plurality of areas according to colors by a K-means algorithm;
determining the color significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the color space and the color space weighting item of the region to be mapped;
determining the depth significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the depth space and the depth space weighting item of the region to be mapped;
and distributing weights of center bias and depth for the color significance and the depth significance, and combining Gaussian normalization to obtain an initial significance mapping map of the region to be mapped.
Preferably, the generating the central dark channel map specifically includes:
integrating each pixel into the image based on the background by combining a global color distinguishing matrix and a spatial distance matrix constructed based on the clustering boundary seeds based on the significance of the distance from each pixel to the center of the image to be detected to obtain a center bias significance map;
carrying out dark channel prior on an image to be detected to obtain a dark channel prior transmittance map;
and distinguishing a significant object from the background of the central bias significance map by combining different transmittances of the foreground and the background in the image, and performing significance detection to obtain a central dark channel mapping map.
Preferably, generating the final saliency map comprises:
enhancing the projection degree of the front edge area by using a negation operation to enhance the depth cue;
based on the combination of the central significance and the dark channel prior, enhancing the central dark channel prior;
and fusing the depth clue and the central dark channel prior by using the initial significance value, and updating to obtain a final significance value.
Preferably, the step S120 specifically includes:
calculating the information entropy of the small region of interest, determining an optimal scale selection mechanism based on the local information entropy, and obtaining significance mapping of different scales by using Gaussian kernel functions of different scales;
generating a single-parameter smooth spectrum family by taking the linear scale space representation as reference based on significance mapping of different scales; giving a target foreground matrix, and determining a logarithmic spectrum and an original phase spectrum of the target foreground matrix; convolving the logarithmic amplitude spectrum with a series of Gaussian kernel functions to obtain a smooth logarithmic amplitude spectrum;
and combining the obtained smooth logarithmic amplitude spectrum with the original phase spectrum to calculate Fourier inverse transformation to obtain significance mapping.
Preferably, the step S130 specifically includes:
projecting pixel values of local areas near the pixel points to a plurality of intervals according to the information entropy of the pixel points in the image, determining the interval with the highest probability of the pixel values, and obtaining a scale chart;
when the salient region is selected, firstly, traversing the maximum pixel points in all the scale maps, defining a plurality of adjacent points by taking the maximum pixel points as the center, calculating the mean value of the pixel values of the adjacent points to obtain the optimal scale map, and completing the optimal scale selection.
In a second aspect, the present invention provides a dark channel prior-based infrared small target detection system, which includes:
the preprocessing module is used for preprocessing the image to be detected by adopting a dark channel prior method to generate a significance mapping chart;
the matching module is used for carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map;
and the identification module is used for selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value which define the local information entropy as judgment standards, and identifying the small target.
In a third aspect, the invention provides a computer apparatus comprising a memory and a processor, the processor and the memory being in communication with each other, the memory storing program instructions executable by the processor, the processor invoking the program instructions to perform the method as described above.
In a fourth aspect, the invention provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements a method as described above.
The invention has the beneficial effects that: the method for detecting the saliency of the image comprises the steps that a dark channel prior method is used for preprocessing the image, so that a single saliency target can be detected more accurately, and a plurality of saliency targets can also be detected more accurately; the significance of the preprocessed image is enhanced, then, according to a special feature matching algorithm, an accurately matched multi-scale significant image is obtained, and a reliable, stable and accurate infrared small target detection result is obtained.
Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic flow chart of a dark channel prior method according to an embodiment of the present invention.
Fig. 2 is a schematic flow chart of a feature matching method according to an embodiment of the present invention.
Fig. 3 is a schematic flow chart of an optimal scale selection method according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below by way of the drawings are illustrative only and are not to be construed as limiting the invention.
It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
For the purpose of facilitating an understanding of the present invention, the present invention will be further explained by way of specific embodiments with reference to the accompanying drawings, which are not intended to limit the present invention.
It should be understood by those skilled in the art that the drawings are merely schematic representations of embodiments and that the elements shown in the drawings are not necessarily required to practice the invention.
Example 1
The embodiment 1 of the invention provides an infrared small target detection system based on dark channel prior, which comprises: the preprocessing module is used for preprocessing the image to be detected by adopting a dark channel prior method to generate a significance mapping chart; the matching module is used for carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map; and the identification module is used for selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value which define the local information entropy as judgment standards, and identifying the small target.
In this embodiment 1, based on the above infrared small target detection system based on dark channel prior, an infrared small target detection method based on dark channel prior is implemented, and the implementation of the detection method includes the following steps: step S110: preprocessing an image to be detected by adopting a dark channel prior method to generate a significance mapping chart; step S120: carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map; step S130: and (3) selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value of the defined local information entropy as a judgment standard, and identifying the small target.
In this embodiment 1, the step S110 specifically includes:
generating an initial saliency mapping map for the image to be detected according to the color saliency and the depth saliency of the RGB-D image; generating a central dark channel mapping map from the image to be detected by combining the central bias significance and dark channel prior; and fusing the initial saliency map and the central dark channel map to generate a final saliency map.
Wherein generating the initial saliency map specifically comprises:
dividing the image into a plurality of areas according to colors by a K-means algorithm;
determining the color significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the color space and the color space weighting item of the region to be mapped;
determining the depth significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the depth space and the depth space weighting item of the region to be mapped;
and distributing weights of center bias and depth for the color significance and the depth significance, and combining Gaussian normalization to obtain an initial significance mapping map of the region to be mapped.
The generating of the central dark channel map specifically includes:
integrating each pixel into the image based on the background by combining a global color distinguishing matrix and a spatial distance matrix constructed based on the clustering boundary seeds based on the significance of the distance from each pixel to the center of the image to be detected to obtain a center bias significance map;
carrying out dark channel prior on an image to be detected to obtain a dark channel prior transmittance map;
and distinguishing a significant object from the background of the central bias significance map by combining different transmittances of the foreground and the background in the image, and performing significance detection to obtain a central dark channel mapping map.
Generating the final saliency map includes:
enhancing the projection degree of the front edge area by using a negation operation to enhance the depth cue;
based on the combination of the central significance and the dark channel prior, enhancing the central dark channel prior;
and fusing the depth clue and the central dark channel prior by using the initial significance value, and updating to obtain a final significance value.
In this embodiment 1, the step S120 specifically includes:
calculating the information entropy of the small region of interest, determining an optimal scale selection mechanism based on the local information entropy, and obtaining significance mapping of different scales by using Gaussian kernel functions of different scales;
generating a single-parameter smooth spectrum family by taking the linear scale space representation as reference based on significance mapping of different scales; giving a target foreground matrix, and determining a logarithmic spectrum and an original phase spectrum of the target foreground matrix; convolving the logarithmic amplitude spectrum with a series of Gaussian kernel functions to obtain a smooth logarithmic amplitude spectrum;
and combining the obtained smooth logarithmic amplitude spectrum with the original phase spectrum to calculate Fourier inverse transformation to obtain significance mapping.
In this embodiment 1, the step S130 specifically includes:
projecting pixel values of local areas near the pixel points to a plurality of intervals according to the information entropy of the pixel points in the image, determining the interval with the highest probability of the pixel values, and obtaining a scale chart;
when the salient region is selected, firstly, traversing the maximum pixel points in all the scale maps, defining a plurality of adjacent points by taking the maximum pixel points as the center, calculating the mean value of the pixel values of the adjacent points to obtain the optimal scale map, and completing the optimal scale selection.
Example 2
The embodiment 2 of the invention provides a dark channel prior-based infrared small target detection method. First, a background complex picture is preprocessed using a dark channel a priori, including: generating an initial saliency map from a color saliency map and a depth saliency map of an RGB-D image; generating a center-dark channel map according to the center saliency and the dark channel prior; the initial saliency map is fused with the center dark channel map to generate a final saliency map. Then, performing feature matching on the preprocessed significant picture, wherein the feature matching comprises the following steps: and obtaining the accurately matched multi-scale saliency map according to a scale division strategy. Finally, the optimal dimension selection is carried out, and the optimal dimension selection comprises the following steps: and identifying the small target by using the gray difference value and the maximum value which define the local information entropy as a judgment standard for selecting the optimal scale.
As shown in fig. 1, in this embodiment 2, the dark channel prior method specifically includes the following steps:
step (1): an initial saliency map is generated from a color saliency map and a depth saliency map of an RGB-D image. Step (2): a center-dark channel map is generated from the center saliency and the dark channel prior. And (3): the initial saliency map is fused with the center dark channel map to generate a final saliency map.
In this embodiment 2, the specific steps of producing the initial saliency map from the color saliency map and the depth saliency map of the RGB-D image include:
step (1-1): and (3) initializing a saliency map: color and depth features are extracted from the original image and the depth map image, respectively, to initialize the saliency map.
Step (2-1): a center-dark channel map is generated from the center saliency and the dark channel prior.
Step (3-1): significance refinement algorithm based on update fusion based on initialized significance mapping, we refine the significance mapping with updated fusion. The updated fusion includes depth cues, central dark channel priors, and an updated enhancement-based fusion.
Wherein, the step (1-1) specifically comprises the following steps:
and (1-1-1) dividing the image into K regions according to colors by a K-means algorithm, wherein the formula is as follows:
wherein S isc(rk) Represents the color significance of a region k, k ∈ [1, k ]]。rkAnd riRepresenting regions k and i, respectively. Dc(rk,ri) Representing the euclidean distance between regions k and i in Lab color space. PiDenotes the area ratio of the region to the hologram, Wd(rk) Representing the spatial weighting term of the region k. As shown in the following formula:
Similar to color saliency, depth saliency is defined as:
wherein S isd(rk) Is depth significance. Dd(rk,ri) Is the euclidean distance of region k from i in depth space.
Step (1-1-2): in most cases, the protruding object is always located at or near the center of the image. Therefore, in this embodiment 2, weights of center offset and depth are assigned to both the color and depth images. The weight of the region k is:
in the formula, G () is gaussian normalized, and | | · | | represents an euclidean distance. PkIs the position of the k region, PoIs the center position of the map. N is a radical ofkIndicates the number of pixels in the k region, DW (d)k) For depth weighting, the following is shown: DW (d)k)＝(max{d}-dk)u(ii) a Where max { d } represents the maximum depth of the image. The depth value of region k is a fixed depth map determined by u:
where min { d } represents the minimum depth of the image.
Then, the initial saliency value for region k is calculated by:
S1(rk)＝G(Sc(rk)Wcd(rk)+Sd(rk)Wcd(rk))。
the step (2-1) specifically comprises the following steps:
step (2-1-1): central prior: according to cognitive neuroscience, human eyes use the fovea to locate objects so that they are clearly visible. Therefore, most cameras capture images that always locate a salient object near the center. The saliency map based on the distance of each pixel to the center of the image can better predict salient objects than many previous saliency models. And constructing a global color distinguishing matrix and a spatial distance matrix based on the clustering boundary seeds, and integrating the global color distinguishing matrix and the spatial distance matrix into a background-based map. Thereby eliminating the influence of image edges and improving the precision of the central target. The resulting significance of this center bias is shown as Scsp。
Step (2-1-2): dark channel prior: dark channel priors are commonly used priors and are widely applied to the field of image defogging. It is based on statistics of outdoor fog-free images. The dark channel can detect the most blurred opaque areas, improving the estimation of atmospheric light.
Step (2-1-3): inspired by the dark channel a priori, the foreground and background have different transmittances, and therefore, a significant object can be distinguished from the background. The theory is applied to the field of significance detection, and the prior transmittance graph of a dark channel is shown as Sdcp. The minimum intensity value of a typical image should be low.
Formally, for image J, define:
wherein the content of the first and second substances,
Jcis one color channel of J, Ω (x) is the local center block of x, JdarJ, processing a general image according to the dark channel of the experimental dark channel, where the dark channel image is often zero, as follows:
calculating an estimated transmission value by the equation;
wherein IcColor channel I, image with fog. A. thecIs a global light, constant for most images. The transmittance of the image is the inverse of the dark channel result. Because objects close to the camera have less photographic fog, the foreground part will be the haze _ free part and the dark channel result tends towards zero.
The step (3-1) specifically comprises the following steps:
step (3-1-1): enhancement based on depth cues. Depth cues highlight salient objects:
Step (3-1-2): enhancement based on a central dark channel prior. The final significance results were enhanced in combination with central significance and dark channel prior as follows:
Scdcp(rk)＝Norm(Scsp(rk))Norm(Sdcp(rk))。
step (3-1-3): the updated fusion is enhanced. Depth cues and central dark channel prior-based enhancement are fused with the initial saliency values as follows:
wherein, S (r)k) For fusion of enhanced significance values, S1(rk) Is the initial significance value.
To refine the significance results, the fusion enhanced significance value is updated to the following formula:
Sf(rk) Refers to the final significance value.
As shown in fig. 2, in this embodiment 2, the method for performing feature matching on a preprocessed saliency picture specifically includes the following steps:
step (1): and obtaining significance mapping of different scales by using Gaussian kernel functions of different scales. Aiming at the optimal scale factor selection problem of the Gaussian kernel function, an optimal scale selection mechanism based on local information entropy is provided. The mechanism meets the characteristics of infrared small targets, limits the calculation of information entropy to interested small areas, and does not need to traverse all complete scale maps. The gaussian kernel is the only kernel that can produce a multi-scale space.
Step (2): with reference to the linear scale space representation, a single-parameter smooth spectral family is generated, the parameters of which depend on the scale of the gaussian kernel. The specific process of the ISSS (improved spectral scale Space) algorithm is as follows. Given a target foreground matrix I (x, y), its log-amplitude spectrum IA(u, v) and phase Spectrum IP(u, v) is represented as follows:
IA(u,v)＝log|fft(I(x,y))|
IP(u,v)＝angle(fft(I(x,y)))
defining the scale space phi (u, v; k) as IA(u, v) convolution with a series of Gaussian kernel functions:
Φ(u,v；k)＝g(u,v；δ)*IA(u,v)
where g (u, v; δ) is a Gaussian kernel whose standard deviation σ is related to the scale factor q:
and setting the step value of the scale parameter of the Gaussian kernel function as an irregular value. When the q value is small, the standard deviation self-adaption change of the Gaussian kernel function is slow, and when the q value is large, the radial spectrum change is fast. The fine scale division strategy is beneficial to selecting proper and accurate Gaussian kernels for small targets. Different types of saliency areas require different filter sizes. Large background areas with a uniform pattern require appropriate dimensions to smooth the amplitude for suppression. Too many small-scale or large-scale selections may result in the background region being insufficiently suppressed, or only the edges of the highlighted region being highlighted.
When a small nucleus is used, a large area is prominent. Long-range or texture-rich targets are detected using large-scale kernels. The infrared small target flying in the sky is generally classified as a remote target, and the number of pixel points is small; therefore, selecting the optimal saliency map requires a fine-scale partitioning strategy. The obtained smooth logarithmic amplitude spectrum phi (u, v; k) and the original phase spectrum IP(u, v) calculating the inverse Fourier transform in combination to obtain significance mapping Sk(x,y)：Sk(x,y)＝ifft{exp(Φ(u,v；k)+i·Ip(u,v))}。
As shown in fig. 3, in this embodiment 2, when selecting the optimal scale, identifying a small target by using the gray scale difference value and the maximum value that define the local information entropy as the judgment criteria for selecting the optimal scale specifically includes the following steps:
step (1): information entropy is often used as a quantitative indicator of the amount of system information. Therefore, it can be further used as a criterion for system equation optimization or parameter selection. In a relatively simple background, a highlighted object may change the information entropy of the entire image. In contrast, for a weak infrared target, its contribution to the overall image information entropy is insignificant. In a suitable scleral image, the region of interest is highlighted, while the other portions are maximally suppressed. For saliency detection of large objects, the minimum image information entropy may well select the optimal saliency map, but is not suitable for infrared objects of very small size. The small target has little influence on the information entropy evaluation of the local salient region. The information entropy is a local concept, and for a pixel point in an image, the information entropy H (x, y) is defined as follows:
wherein ^ (x, y) represents a local region in the vicinity of a pixel point (x, y), and the pixel values of the local region are projected onto K intervals, Pb(x, y) represents the probability that the pixel value is within the b interval. The entropy value of the local information is high, which indicates that the information of the region is rich and the probability of containing small targets is high. In the optimal scale map, the target significance is better than the background clutter, and the background often shows certain spatial similarity.
Step (2): when selecting the salient region, firstly traversing the maximum pixel point L in all the scale mapsk：Lk＝max(Sk(x,y)) k＝1,2...,K
Where K is set to 16 in the 288 × 384 original image. Centering on this point, define 8 adjacent points as bkCalculate the mean m of the pixel values of its 8 neighborhoodskThe following were used:
mk＝argmax{H(Vk)}Vk＝Bk+Lk；
wherein L iskRepresenting the largest pixel point, BkRepresenting adjacent 8 pixels.
Example 4
The embodiment 4 of the present invention provides a computer device, including a memory and a processor, where the processor and the memory are in communication with each other, the memory stores a program instruction executable by the processor, and the processor calls the program instruction to execute a dark channel prior-based infrared small target detection method, where the method includes:
step S110: preprocessing an image to be detected by adopting a dark channel prior method to generate a significance mapping chart; step S120: carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map; step S130: and (3) selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value of the defined local information entropy as a judgment standard, and identifying the small target.
Example 5
Embodiment 5 of the present invention provides a computer-readable storage medium, in which a computer program is stored, where the computer program, when executed by a processor, implements a dark channel prior-based infrared small target detection method, where the method includes:
step S110: preprocessing an image to be detected by adopting a dark channel prior method to generate a significance mapping chart; step S120: carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map; step S130: and (3) selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value of the defined local information entropy as a judgment standard, and identifying the small target.
In summary, the embodiments of the present invention are described in the following
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the present disclosure has been described with reference to the specific embodiments shown in the drawings, it is not intended to limit the scope of the present disclosure, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive faculty based on the technical solutions disclosed in the present disclosure.
Claims (10)
1. A small infrared target detection method based on dark channel prior is characterized by comprising the following steps:
step S110: preprocessing an image to be detected by adopting a dark channel prior method to generate a significance mapping chart;
step S120: carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map;
step S130: and (3) selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value of the defined local information entropy as a judgment standard, and identifying the small target.
2. The method for detecting infrared small targets based on dark channel prior according to claim 1, wherein the step S110 specifically includes:
generating an initial saliency mapping map for the image to be detected according to the color saliency and the depth saliency of the RGB-D image; generating a central dark channel mapping map from the image to be detected by combining the central bias significance and dark channel prior; and fusing the initial saliency map and the central dark channel map to generate a final saliency map.
3. The dark channel prior-based infrared small target detection method according to claim 2, wherein generating an initial saliency map specifically comprises:
dividing the image into a plurality of areas according to colors by a K-means algorithm;
determining the color significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the color space and the color space weighting item of the region to be mapped;
determining the depth significance of the region to be mapped according to the Euclidean distance between the region to be mapped and another region in the depth space and the depth space weighting item of the region to be mapped;
and distributing weights of center bias and depth for the color significance and the depth significance, and combining Gaussian normalization to obtain an initial significance mapping map of the region to be mapped.
4. The infrared small target detection method based on dark channel prior according to claim 3, wherein generating the central dark channel map specifically comprises:
integrating each pixel into the image based on the background by combining a global color distinguishing matrix and a spatial distance matrix constructed based on the clustering boundary seeds based on the significance of the distance from each pixel to the center of the image to be detected to obtain a center bias significance map;
carrying out dark channel prior on an image to be detected to obtain a dark channel prior transmittance map;
and distinguishing a significant object from the background of the central bias significance map by combining different transmittances of the foreground and the background in the image, and performing significance detection to obtain a central dark channel mapping map.
5. The dark channel prior-based infrared small target detection method of claim 4, wherein generating a final saliency map comprises:
enhancing the projection degree of the front edge area by using a negation operation to enhance the depth cue;
based on the combination of the central significance and the dark channel prior, enhancing the central dark channel prior;
and fusing the depth clue and the central dark channel prior by using the initial significance value, and updating to obtain a final significance value.
6. The method for detecting infrared small targets based on dark channel prior according to claim 1, wherein the step S120 specifically includes:
calculating the information entropy of the small region of interest, determining an optimal scale selection mechanism based on the local information entropy, and obtaining significance mapping of different scales by using Gaussian kernel functions of different scales;
generating a single-parameter smooth spectrum family by taking the linear scale space representation as reference based on significance mapping of different scales; giving a target foreground matrix, and determining a logarithmic spectrum and an original phase spectrum of the target foreground matrix; convolving the logarithmic amplitude spectrum with a series of Gaussian kernel functions to obtain a smooth logarithmic amplitude spectrum;
and combining the obtained smooth logarithmic amplitude spectrum with the original phase spectrum to calculate Fourier inverse transformation to obtain significance mapping.
7. The method for detecting infrared small targets based on dark channel prior according to claim 1, wherein the step S130 specifically includes:
projecting pixel values of local areas near the pixel points to a plurality of intervals according to the information entropy of the pixel points in the image, determining the interval with the highest probability of the pixel values, and obtaining a scale chart;
when the salient region is selected, firstly, traversing the maximum pixel points in all the scale maps, defining a plurality of adjacent points by taking the maximum pixel points as the center, calculating the mean value of the pixel values of the adjacent points to obtain the optimal scale map, and completing the optimal scale selection.
8. A small infrared target detection system based on dark channel prior is characterized by comprising:
the preprocessing module is used for preprocessing the image to be detected by adopting a dark channel prior method to generate a significance mapping chart;
the matching module is used for carrying out feature matching on the preprocessed saliency mapping map according to a scale division strategy to obtain a multi-scale saliency map;
and the identification module is used for selecting the optimal scale of the multi-scale saliency map by using the gray difference value and the maximum value which define the local information entropy as judgment standards, and identifying the small target.
9. A computer device comprising a memory and a processor, the processor and the memory in communication with each other, the memory storing program instructions executable by the processor, characterized in that: the processor calls the program instructions to perform the method of any one of claims 1-7.
10. A computer-readable storage medium storing a computer program, characterized in that: the computer program, when executed by a processor, implements the method of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548661.XA CN112529896A (en) | 2020-12-24 | 2020-12-24 | Infrared small target detection method and system based on dark channel prior |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548661.XA CN112529896A (en) | 2020-12-24 | 2020-12-24 | Infrared small target detection method and system based on dark channel prior |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112529896A true CN112529896A (en) | 2021-03-19 |
Family
ID=74976192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011548661.XA Pending CN112529896A (en) | 2020-12-24 | 2020-12-24 | Infrared small target detection method and system based on dark channel prior |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112529896A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113409082A (en) * | 2021-06-18 | 2021-09-17 | 湖南快乐阳光互动娱乐传媒有限公司 | Interactive advertisement putting method, system, server and client |
| CN115393579A (en) * | 2022-10-27 | 2022-11-25 | 长春理工大学 | Infrared small target detection method based on weighted block contrast |
| CN115861988A (en) * | 2023-03-01 | 2023-03-28 | 四川省农业机械研究设计院 | Tea leaf picking method and system based on RGB discrimination |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292318A (en) * | 2017-07-21 | 2017-10-24 | 北京大学深圳研究生院 | Image significance object detection method based on center dark channel prior information |
| CN109741293A (en) * | 2018-11-20 | 2019-05-10 | 武汉科技大学 | Conspicuousness detection method and device |
| CN109816641A (en) * | 2019-01-08 | 2019-05-28 | 西安电子科技大学 | Weighted local entropy infrared small target detection method based on Multiscale Morphological Fusion |
| CN111260624A (en) * | 2020-01-15 | 2020-06-09 | 陕西师范大学 | Multi-scale infrared small target detection method based on feature matching and scale selection strategy |
-
2020
- 2020-12-24 CN CN202011548661.XA patent/CN112529896A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107292318A (en) * | 2017-07-21 | 2017-10-24 | 北京大学深圳研究生院 | Image significance object detection method based on center dark channel prior information |
| CN109741293A (en) * | 2018-11-20 | 2019-05-10 | 武汉科技大学 | Conspicuousness detection method and device |
| CN109816641A (en) * | 2019-01-08 | 2019-05-28 | 西安电子科技大学 | Weighted local entropy infrared small target detection method based on Multiscale Morphological Fusion |
| CN111260624A (en) * | 2020-01-15 | 2020-06-09 | 陕西师范大学 | Multi-scale infrared small target detection method based on feature matching and scale selection strategy |
Non-Patent Citations (3)
| Title |
|---|
| CHUNBIAO ZHU ETC.: ""An Innovative Salient Object Detection Using Center-Dark Channel Prior"", 《PROCEEDINGS OF THE IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
| KAIMING HE ETC.: ""Single Image Haze Removal Using Dark Channel Prior"", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACNINE INTELLIGENCE》 * |
| YAO QIN ETC.: ""Saliency detection via Cellular Automata"", 《IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113409082A (en) * | 2021-06-18 | 2021-09-17 | 湖南快乐阳光互动娱乐传媒有限公司 | Interactive advertisement putting method, system, server and client |
| CN115393579A (en) * | 2022-10-27 | 2022-11-25 | 长春理工大学 | Infrared small target detection method based on weighted block contrast |
| CN115393579B (en) * | 2022-10-27 | 2023-02-10 | 长春理工大学 | Infrared small target detection method based on weighted block contrast |
| CN115861988A (en) * | 2023-03-01 | 2023-03-28 | 四川省农业机械研究设计院 | Tea leaf picking method and system based on RGB discrimination |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US12020474B2 (en) | Image processing apparatus, image processing method, and non-transitory computer-readable storage medium | |
| Zeng et al. | Deep surface normal estimation with hierarchical rgb-d fusion | |
| Rahtu et al. | Segmenting salient objects from images and videos | |
| CN112529896A (en) | Infrared small target detection method and system based on dark channel prior | |
| US9042662B2 (en) | Method and system for segmenting an image | |
| Lindeberg | Scale selection | |
| CN111079556A (en) | Multi-temporal unmanned aerial vehicle video image change area detection and classification method | |
| CN112052831B (en) | Method, device and computer storage medium for face detection | |
| CN112926410A (en) | Target tracking method and device, storage medium and intelligent video system | |
| JP7142420B2 (en) | Image processing device, learning method, trained model, image processing method | |
| Hasinoff | Saturation (imaging) | |
| Tangsakul et al. | Single image haze removal using deep cellular automata learning | |
| Huang et al. | Removing reflection from a single image with ghosting effect | |
| Radolko et al. | Dataset on underwater change detection | |
| Tan | Specularity, specular reflectance | |
| Fisher | Subpixel estimation | |
| CN112132753B (en) | Infrared image super-resolution method and system for multi-scale structure guide image | |
| AU2018202801A1 (en) | Method, apparatus and system for producing a foreground map | |
| Faraji et al. | Simplified active calibration | |
| CN114283087A (en) | Image denoising method and related equipment | |
| Cao et al. | Depth image vibration filtering and shadow detection based on fusion and fractional differential | |
| Tunali et al. | Fast semi-automatic target initialization based on visual saliency for airborne thermal imagery | |
| Sliti et al. | Efficient visual tracking via sparse representation and back-projection histogram | |
| Varalakshmamma et al. | Detection and restoration of image from multi-color fence occlusions | |
| Schneider | Shape from silhouette |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210319 |