CN114022473B - Horizon detection method based on infrared image - Google Patents
Horizon detection method based on infrared image Download PDFInfo
- Publication number
- CN114022473B CN114022473B CN202111374988.4A CN202111374988A CN114022473B CN 114022473 B CN114022473 B CN 114022473B CN 202111374988 A CN202111374988 A CN 202111374988A CN 114022473 B CN114022473 B CN 114022473B
- Authority
- CN
- China
- Prior art keywords
- horizon
- preliminary
- detection result
- detection
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 105
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 20
- 238000003708 edge detection Methods 0.000 claims abstract description 17
- 238000009499 grossing Methods 0.000 claims abstract description 11
- 238000012216 screening Methods 0.000 claims abstract description 3
- 238000000034 method Methods 0.000 claims description 28
- 230000002159 abnormal effect Effects 0.000 claims description 21
- 238000001914 filtration Methods 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 6
- 230000000694 effects Effects 0.000 abstract description 8
- 230000011218 segmentation Effects 0.000 description 13
- 238000003384 imaging method Methods 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 230000035772 mutation Effects 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 230000008439 repair process Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 238000004043 dyeing Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 238000011426 transformation method Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000013082 photovoltaic technology Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
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 relates to a horizon detection method based on infrared images, which comprises the following steps: performing edge-preserving smoothing treatment on the infrared image to be detected by using an edge-preserving filter to obtain an output image; performing edge detection on the output image by using a Canny edge detection algorithm, and screening out candidate horizon point sets corresponding to each column; searching a demarcation point in each candidate horizon point set, so that the gray average value of a sky column area taking the demarcation point as a demarcation point is larger than the gray average value of a ground object column area, the sum of the gray variance of the sky column area and the gray variance of the ground object column area is minimum, taking the demarcation point as a preliminary optimal horizon point of a corresponding column, and forming a horizon preliminary detection result by the preliminary optimal horizon points of all columns; and repairing the preliminary detection result of the horizon to obtain a repaired horizon detection result. The invention obtains good horizon detection effect, and can keep better robustness and instantaneity for complex situations of sky having cloud layers or ground building interlacing and the like.
Description
Technical Field
The invention relates to the technical field of horizon detection, in particular to a horizon detection method based on infrared images.
Background
The horizon detection technology is a key link in the fields of infrared early warning and target searching, and particularly when an infrared early warning system detects a head-up surface in a long-distance manner, an observed scene is a composite image containing sky and ground objects. Because the sky background and the ground background have larger difference, the sky background is cleaner, the ground object background is more complex and changeable, ground clutter, noise interference and the like often cause the increase of false alarm rate and false detection rate, and the accurate detection and identification of infrared targets face challenges. In order to avoid the influence of ground clutter on infrared target detection precision when an infrared early warning system detects in a far distance on a head-up surface, accurate horizon detection on a world composite image is necessary. Through the effective detection of the horizon, the target searching range can be reduced, the region of interest is divided, the influence of ground features is removed, and a powerful guarantee is provided for the precision and the efficiency of target detection of an infrared early warning system.
At present, the horizon detection problem of infrared images is relatively few in domestic research. At present, most of the detection algorithms mainly used are referenced to the sea-sky-line detection method, and mainly comprise a Hough transformation method, a region segmentation method and the like.
For example, the invention CN109978869a provides a sea-sky-line detection method based on gray level co-occurrence matrix and Hough transform. The image quality of the collected original visible light video monitoring image is enhanced; converting the preprocessed image into a gray scale image by using a gray scale conversion formula after dyeing; calculating gray level co-occurrence matrixes under different angle values for the obtained gray level map; determining a possible area of the sea-sky line according to the texture change rate; carrying out gray morphological corrosion and expansion operation on the region by adopting a mathematical morphology method; removing noise by using improved weighted Gaussian blur based on histogram optimization to obtain an edge image; and finally fitting to the finally detected sea-antenna.
The invention patent CN108022214A discloses a horizon detection method suitable for unmanned aerial vehicle foggy day flight. And carrying out dark primary color defogging treatment on the RGB image obtained by reading in and collecting to obtain a clear defogging image and a dark primary color image improved by a soft matting method for feature extraction. And after dyeing, sequentially carrying out image segmentation, morphological processing and edge detection on the improved dark primary color image. And finally, carrying out linear detection by using Hough transformation, and carrying out least square method by using a linear detection result to accurately obtain horizon information.
The invention patent CN105644785A discloses an unmanned aerial vehicle landing method based on an optical flow method and horizon detection, which comprises the steps of firstly, preprocessing images of videos shot by cameras fixed at the bottom of an unmanned aerial vehicle in the flight process; then, carrying out straight line detection of Hough transformation on each image to obtain horizon information in the image; calculating horizon information to obtain the current flight attitude of the unmanned aerial vehicle, and detecting the attitude information of the unmanned aerial vehicle by adopting an optical flow method; and finally, combining an unmanned aerial vehicle motion model, filtering the unmanned aerial vehicle gesture detected by an optical flow method and a horizon by adopting an extended Kalman filtering method, and selecting correct horizon information to realize an autonomous landing process based on the unmanned aerial vehicle.
Yang Shanmin et al propose a horizon detection algorithm under low visibility (study of horizon detection algorithm under low visibility [ J ]. Computer engineering and design 2012 (01): 238-242) for characteristics of low infrared image contrast and large noise. The algorithm firstly carries out gray level transformation on the image to enhance contrast, defines the energy of each pixel according to the attention and binding two stages in the human visual system identification process and the related theory of edge detection, finds candidate horizon demarcation points by using a dynamic programming method, and finally obtains a real horizon through Hough transformation.
Liang Lei et al propose an unmanned aerial vehicle horizon detection method based on sky segmentation (unmanned aerial vehicle horizon detection based on sky segmentation [ J ]. Modern computer 2021, 1:73-77), which performs color space conversion on an image to be detected; performing sky segmentation operation on the converted image; then, edge detection is carried out on the segmented image by using a Canny operator; finally, the recognition and detection of the horizon are performed by using Hough transformation.
Sun Yuxin et al propose an infrared image horizon detection algorithm based on semantic segmentation (research of infrared image horizon detection algorithm based on semantic segmentation [ J ]. Photovoltaic technology application, 2020,35 (6): 55-57,78), which introduces a deep semantic segmentation model into horizon detection tasks, and utilizes Deeplab-v3+ algorithm as semantic segmentation algorithm to realize effective segmentation of sky and ground.
Gao Haifeng et al propose a horizon detection method based on infrared images (research on horizon detection method based on infrared images [ J ]. Electro-optic and control, 2016,23 (7): 20-23), which firstly carries out smoothing treatment on infrared images, then detects the edges of the images by Sobel gradient operator, and finds the edges by comparing Harr linear features so as to determine the sky area and the position of the quasi horizon; the preliminary detection of the horizon position is completed through the optimal extraction of boundary energy; and carrying out rationality analysis and repair on the horizon by using a related technology of numerical analysis.
However, the existing horizon detection method of the infrared image based on semantic segmentation introduces a depth semantic segmentation model into a horizon detection task, so that the calculation complexity of an algorithm is high, and the real-time requirement cannot be met for an early warning system with a high frame frequency or a high resolution. The existing unmanned aerial vehicle horizon detection method based on sky segmentation is based on color space extraction and is not suitable for infrared images. Other existing horizon detection methods have the problems of large calculation amount and limited precision.
In addition, most detection algorithms are based on sea-sky-line detection methods, and mainly depend on Hough transformation methods, region segmentation methods and the like. However, these methods are mainly directed to imaging characteristics of sea-sky lines, such as singleness of sea-sky background, straight line characteristics of sea-sky boundary positions, and the like. Whereas the horizon of the infrared image is different from the imaging properties of the sea-sky-line of the visible image. The infrared image has irregular ground object, the ground object texture is more complex, and the horizon is not always a simple straight line in the strict sense. Thus, when such a method is applied to horizon detection, there is a limit, and the detection result is not satisfactory. Therefore, the horizon real-time detection method for the infrared image has very important significance.
Disclosure of Invention
In view of the above problems, the invention provides a horizon detection method based on infrared images, which achieves good horizon detection effect, and can maintain good robustness and real-time even if there are cloud layers in the sky or complicated conditions such as staggering of ground building houses.
In order to achieve the above purpose, the invention adopts the following technical scheme:
A horizon detection method based on infrared images comprises the following steps:
step 1: performing edge-preserving smoothing treatment on the infrared image to be detected by using an edge-preserving filter to obtain an output image after edge-preserving smoothing filtering;
step2: performing edge detection on the output image by using a Canny edge detection algorithm, and screening out candidate horizon point sets corresponding to each column of the output image;
Step 3: searching a demarcation point in each candidate horizon point set, so that the sum of the sky column region gray variance and the ground object column region gray variance, which are bounded by the demarcation points, is minimum, the sky column region gray average is larger than the ground object column region gray average, and taking the demarcation points as preliminary optimal horizon points of corresponding columns, wherein the preliminary optimal horizon points of all columns form a horizon preliminary detection result;
step 4: and repairing the preliminary detection result of the horizon to obtain a repaired horizon detection result.
Firstly, performing edge-preserving smoothing treatment on an infrared image to be detected by using an edge-preserving filter; then extracting the edge contour of the filtered image by using a Canny edge detection algorithm; considering that the horizon of the infrared image has edge characteristics and the sky area with the horizon as a boundary has respectively consistent imaging characteristics, then taking the image column as a unit, taking the edge characteristics and gray average values of the sky column area and the ground object column area as constraint conditions, and establishing an energy function based on the gray variances of the sky column area and the ground object column area and the minimum energy function, so as to finish the preliminary detection of the horizon position; and finally, removing abnormal points in the horizon by using a numerical analysis method based on gradient information mutation, and reasonably repairing the horizon, so that an accurate and complete horizon is obtained.
Compared with the prior art, the invention has the following beneficial effects:
(1) Compared with the traditional sea-sky-line-based detection method, the method fully considers the imaging characteristic of the infrared image and the nonlinear edge characteristic of the horizon line, and has good applicability to horizon line detection of the infrared image;
(2) Compared with the existing infrared image horizon detection method, the method provided by the invention has the advantages that the candidate horizon point set corresponding to each row is screened out based on the edge protection filter and the Canny edge detection algorithm, then the horizon primary detection result is extracted by taking the constraint condition that the sum of the sky row region gray variance and the ground object row region gray variance is minimum and the sky row region gray average is larger than the ground object row region gray average, and the horizon primary detection result is repaired, so that the repaired horizon detection result is finally obtained, a good horizon detection effect is obtained, and better robustness and instantaneity can be maintained even if cloud layers exist in the sky or complicated conditions such as ground object building house interlacing are achieved.
Drawings
FIG. 1 is a flow chart of a horizon detection method based on infrared images according to the present invention;
FIG. 2 is a graph showing the preliminary detection result of the horizon of an infrared image of a scene in the present invention;
Fig. 3 is a repaired horizon detection result obtained by repairing the horizon primary detection result shown in fig. 2;
fig. 4 shows the results of the horizon detection after the partial repair of the infrared images of different scenes.
Detailed Description
The technical scheme of the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments.
The overall flow of the horizon detection method based on the infrared image, which is provided by the invention, is shown in figure 1, and the method comprises four steps, namely the following specific steps:
Step 1: image preprocessing
In the step, the infrared image to be detected is subjected to edge-preserving smoothing treatment by using an edge-preserving filter so as to remove the complex texture structure of the ground object and protect the edge at the same time, avoid influencing the detection result, and an output image after the edge-preserving smoothing filtration is obtained after the treatment.
The edge protection filter utilized in the image preprocessing of the step is fast guiding filter
When the edge protection filter in the embodiment of the invention adopts the rapid guiding filtering, the image preprocessing process specifically comprises the following steps:
Firstly, performing small structure removal on an infrared image by using a Gaussian filter, and taking an output image G of the Gaussian filter as a guiding image of the next step, wherein the formula is as follows:
G＝Gussian(I，σr) (1)
Wherein I is an input infrared image, sigma r is the size of a Gaussian kernel, namely the radius of a filtering window, and the filter effect of different degrees can be realized by changing the value of the sigma r. The image after being output through Gaussian filtering is used as a guiding image of the next step.
Next, considering the real-time performance of the algorithm, the invention uses the fast guiding filtering to perform the edge protection operation, and the formula is as follows:
Iout＝FastGuidedFilte r(G,I,σr,σs) (2)
Wherein G is a guide graph, sigma s is a regularization parameter, and I out is an output image after edge-preserving smoothing filtering.
Step 2: edge detection
In general, the horizon of the infrared image has a boundary between the sky and the ground object, has obvious gradient change and accords with edge characteristics, so that the step adopts a Canny edge detection algorithm to carry out edge detection on the output image obtained in the step 1, and each row of candidate horizon point sets corresponding to the output image are screened out. In the j-th column, the set of edge points detected by the Canny edge detection algorithm, i.e., candidate horizon points, is denoted as P Cj in units of columns of the output image.
Step 3: horizon point extraction
Consider that in an infrared image, a world region bounded by an optimal horizon has the following two imaging characteristics:
Texture information of the ground object area is more complex than texture information of the sky area;
the sky area is brighter in imaging effect than the ground object area.
The present invention thus summarizes the corresponding mathematical characterization that corresponds to the imaging characteristics above:
1) The sum of the sky area gray variance and the ground object area gray variance is minimum;
2) The gray average value of the sky area is larger than that of the ground object area.
Therefore, taking imaging characteristics of the infrared image and edge characteristics of the horizon as constraint conditions, taking the columns as units, establishing gray variance and minimum function based on the upper column region (namely the sky column region) and the lower column region (namely the ground column region), and taking the gray average value of the sky column region as constraint conditions which are larger than that of the ground column region, so as to finish preliminary detection of the position of the horizon point.
Specifically, a demarcation point P j in the candidate horizon point set corresponding to each column is searched, so that the sum of the sky column region gray variance and the ground column region gray variance, which are bounded by the demarcation points, is minimized, and the condition that the sky column region gray mean is larger than the ground column region gray mean is satisfied. Suppose that at column j, demarcation point P j satisfies:
Wherein J (P j) is the sum of the gray variances of the upper and lower regions, which is the energy value when P j is taken as the demarcation point, I U (I, J) and I D (I, J) respectively represent the gray values of the pixels of the ith row and jth column when P j is taken as the demarcation point, H and W respectively represent the height and width of the image, And/>The gray average values of the sky row area and the ground object row area when the P j is taken as a demarcation point are respectively represented, and the calculation formula is as follows:
In the formula, N U and N D respectively represent the number of pixels in the sky line area and the ground line area when P j is taken as a demarcation point.
The preliminary optimal horizon point of the j-th column can be solved according to formulas (3) - (6)The preliminary optimal horizon points of all columns constitute the horizon preliminary detection result.
Still referring to fig. 1, step 3 specifically includes the following cyclic process:
Step 31: in the j-th column of the output image, taking one candidate horizon point as a demarcation point, and dividing the j-th column of the output image into a sky column area and a ground object column area;
Step 32: then judging whether the gray average value of the sky line area is larger than the gray average value of the ground object line area and the sum of the gray variance of the sky line area and the gray variance of the ground object line area is minimum, if yes, executing step 33, otherwise discarding the candidate horizon point and returning to step 31, and reselecting one candidate horizon point in a j-th column as a demarcation point;
step 33: taking the candidate horizon point as a preliminary optimal horizon point of a j-th column;
step 34: judging whether the j-th column is the last column of the output image, if so, executing the step 4, otherwise, returning to the step 31 after j++.
Step 4: and repairing the preliminary detection result of the horizon to obtain a repaired horizon detection result.
The general horizon can be detected through the steps 1 to 3, but the detection result often has some abnormal point segments, so that the abnormal point segments need to be removed, and the horizon is reasonably repaired.
According to the initially detected horizon trend characteristic, the regional characteristic of the abnormal point segment is statistically analyzed, the abnormal point in the initial horizon detection result is eliminated by using a numerical analysis method based on gradient information mutation, and then the horizon is reasonably repaired, and the method specifically comprises the following steps:
Step 41: calculating an absolute gradient value of a preliminary horizon detection result;
Knowing that the set of preliminary optimal horizon points for all columns of the output image is { P 1 best…Pj best,Pj+1 best … }, the absolute gradient value of the horizon point corresponding to the j-th column is:
The absolute gradient value of the initial horizon detection result is { G 1,G2…Gj,Gj+1 … }.
Step 42: searching for an abnormal point segment in the preliminary horizon detection result according to the absolute gradient value calculated in the step 41 and a preset gradient abrupt change threshold;
Assume that the preset gradient abrupt threshold is th, for example, th takes 10 in this embodiment. If G m is the first gradient abrupt change value, i.e., G m > th, then the corresponding preliminary optimal horizon point P m best is considered to be the horizon abrupt change point. If G n is the next gradient abrupt change value, i.e., G n > th, and m < n is satisfied, then the corresponding preliminary optimal horizon point P n best is considered to be the horizon abrupt change point as well. If |m-n|ε [0,l ], where l is the maximum length of the outlier segment, e.g., l takes 4 in this embodiment. The outlier segment is considered as { P m best…Pn best }, i.e., the outlier segment is a segment composed of the preliminary optimal horizon points corresponding to any two adjacent gradient abrupt change values and all the preliminary optimal horizon points between them. It should be noted that the gradient mutation threshold value used when searching the abnormal point segment and the maximum length of the abnormal point segment are not limited to 10 and 4 given in the embodiment, and other reasonable values can be adopted to achieve the effect of the invention.
Step 43: and removing the abnormal point segment from the preliminary detection result of the horizon, and replacing the abnormal point segment with the previous preliminary optimal horizon point adjacent to the abnormal point segment to obtain the repaired horizon detection result.
Knowing that the abnormal point segment is { P m best…Pn best }, removing the abnormal point segment { P m best…Pn best } from the preliminary detection result of the horizon and replacing the abnormal point segment by the following formula to reasonably repair the horizon:
{Pm best…Pn best}＝Pm-1 best (8)
And obtaining a final repaired horizon line detection result after substitution.
Firstly, performing edge-preserving smoothing treatment on an infrared image to be detected by using an edge-preserving filter; then extracting the edge contour of the filtered image by using a Canny edge detection algorithm; considering that the horizon of the infrared image has edge characteristics and the sky area with the horizon as a boundary has respectively consistent imaging characteristics, then taking the image column as a unit, taking the edge characteristics and gray average values of the sky column area and the ground object column area as constraint conditions, and establishing an energy function based on the gray variances of the sky column area and the ground object column area and the minimum energy function, so as to finish the preliminary detection of the horizon position; and finally, removing abnormal points in the horizon by using a numerical analysis method based on gradient information mutation, and reasonably repairing the horizon, so that an accurate and complete horizon is obtained.
Compared with the prior art, the invention has the following beneficial effects:
(1) Compared with the traditional sea-sky-line-based detection method, the method fully considers the imaging characteristic of the infrared image and the nonlinear edge characteristic of the horizon line, and has good applicability to horizon line detection of the infrared image;
(2) Compared with the existing infrared image horizon detection method, the method provided by the invention has the advantages that the candidate horizon point set corresponding to each row is screened out based on the edge protection filter and the Canny edge detection algorithm, then the horizon primary detection result is extracted by taking the constraint condition that the sum of the sky row region gray variance and the ground object row region gray variance is minimum and the sky row region gray average is larger than the ground object row region gray average, and the horizon primary detection result is repaired, so that the repaired horizon detection result is finally obtained, a good horizon detection effect is obtained, and better robustness and instantaneity can be maintained even if cloud layers exist in the sky or complicated conditions such as ground object building house interlacing are achieved.
Fig. 2 shows a preliminary detection result of the horizon of the infrared image in a certain scene obtained through steps 1 to 3, and fig. 3 shows a repaired horizon detection result obtained through repairing the preliminary detection result of the horizon shown in fig. 2. As can be seen by comparing fig. 2 and fig. 3, through reasonable repair of the preliminary detection result of the horizon, certain abnormal point segments are removed, so that the repaired horizon detection result is closer to the real horizon.
In order to verify the robustness and accuracy of the invention, the invention also carries out experiments of horizon detection on infrared images of different scenes under complex conditions, the detection steps are all as described in the steps 1 to 4, and partial horizon detection results are shown in fig. 4. The detection result shows that the invention has good horizon detection effect, and can keep good robustness and real-time even if the sky has cloud layers or the ground building is staggered and other complex conditions.
The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples illustrate only a few embodiments of the invention, which are described in detail and are not to be construed as limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the invention, which are all within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.
Claims (3)
1. The horizon detection method based on the infrared image is characterized by comprising the following steps of:
step 1: performing edge-preserving smoothing treatment on the infrared image to be detected by using an edge-preserving filter to obtain an output image after edge-preserving smoothing filtering; the edge protection filter is a fast guiding filter, and the step 1 comprises the following steps:
and performing small structure removal on the infrared image by using a Gaussian filter, and taking the image output by the Gaussian filter as a guiding image of the next step, wherein the formula is as follows:
G＝Gussian(I,σr) (1)
Wherein G is a guide image, I is an input infrared image, and sigma r is a Gaussian kernel size;
the edge protection operation is carried out by utilizing the fast guiding filtering, and the formula is as follows:
Iout＝FastGuidedFilter(G,I,σr,σs) (2)
wherein sigma s is regularization parameter, and I out is output image after edge-preserving smooth filtering;
step2: performing edge detection on the output image by using a Canny edge detection algorithm, and screening out candidate horizon point sets corresponding to each column of the output image;
Step 3: searching a demarcation point in each candidate horizon point set, so that the sum of the sky column region gray variance and the ground object column region gray variance, which are bounded by the demarcation points, is minimum, the sky column region gray average is larger than the ground object column region gray average, and taking the demarcation points as preliminary optimal horizon points of corresponding columns, wherein the preliminary optimal horizon points of all columns form a horizon preliminary detection result;
Step 4: repairing the preliminary detection result of the horizon to obtain a repaired horizon detection result; the process for repairing the preliminary detection result of the horizon comprises the following steps:
step 41: calculating an absolute gradient value of the horizon primary detection result;
Step 42: searching for an abnormal point segment in the preliminary horizon detection result according to the absolute gradient value and a preset gradient abrupt change threshold;
step 43: and removing the abnormal point segment from the preliminary detection result of the horizon, and replacing the abnormal point segment with the previous preliminary optimal horizon point adjacent to the abnormal point segment to obtain the repaired horizon detection result.
2. The method for detecting the horizon based on the infrared image according to claim 1, wherein the outlier segment is a point segment consisting of preliminary optimal horizon points corresponding to any two adjacent gradient abrupt change values and all preliminary optimal horizon points therebetween.
3. The method for detecting the horizon based on the infrared image according to claim 1 or 2, wherein the gradient abrupt change threshold is 10, and the maximum length of the abnormal point segment is 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111374988.4A CN114022473B (en) | 2021-11-19 | 2021-11-19 | Horizon detection method based on infrared image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111374988.4A CN114022473B (en) | 2021-11-19 | 2021-11-19 | Horizon detection method based on infrared image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114022473A CN114022473A (en) | 2022-02-08 |
| CN114022473B true CN114022473B (en) | 2024-04-26 |
Family
ID=80065236
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111374988.4A Active CN114022473B (en) | 2021-11-19 | 2021-11-19 | Horizon detection method based on infrared image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114022473B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205160655U (en) * | 2012-12-21 | 2016-04-13 | 菲力尔系统公司 | A infrared imaging system for vehicle |
| CN109242870A (en) * | 2018-07-13 | 2019-01-18 | 上海大学 | A kind of sea horizon detection method divided based on image with textural characteristics |
| CN109493353A (en) * | 2018-10-06 | 2019-03-19 | 中国航空工业集团公司洛阳电光设备研究所 | A kind of sky and ocean background is plunged into the commercial sea the method for detecting in day line of demarcation |
| WO2020007307A1 (en) * | 2018-07-03 | 2020-01-09 | 深圳岚锋创视网络科技有限公司 | Sky filter method for panoramic images and portable terminal |
| CN113239733A (en) * | 2021-04-14 | 2021-08-10 | 重庆利龙科技产业(集团)有限公司 | Multi-lane line detection method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013009697A1 (en) * | 2011-07-08 | 2013-01-17 | Bendix Commercial Vehicle Systems Llc | Image-based vehicle detection and distance measuring method and apparatus |
-
2021
- 2021-11-19 CN CN202111374988.4A patent/CN114022473B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN205160655U (en) * | 2012-12-21 | 2016-04-13 | 菲力尔系统公司 | A infrared imaging system for vehicle |
| WO2020007307A1 (en) * | 2018-07-03 | 2020-01-09 | 深圳岚锋创视网络科技有限公司 | Sky filter method for panoramic images and portable terminal |
| CN109242870A (en) * | 2018-07-13 | 2019-01-18 | 上海大学 | A kind of sea horizon detection method divided based on image with textural characteristics |
| CN109493353A (en) * | 2018-10-06 | 2019-03-19 | 中国航空工业集团公司洛阳电光设备研究所 | A kind of sky and ocean background is plunged into the commercial sea the method for detecting in day line of demarcation |
| CN113239733A (en) * | 2021-04-14 | 2021-08-10 | 重庆利龙科技产业(集团)有限公司 | Multi-lane line detection method |
Non-Patent Citations (3)
| Title |
|---|
| 一种天地复杂背景下的红外弱小目标检测方法;宋敏敏;王爽;吕弢;袁瑜键;;红外技术;20181018(第10期);全文 * |
| 基于红外图像的SF_6检漏定位系统;毕国玲;赵建;隋龙;张艳超;;光电工程;20110315(第03期);全文 * |
| 基于红外图像的天地线检测方法研究;高海峰;张生伟;聂青凤;;电光与控制;20160601(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114022473A (en) | 2022-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107316031B (en) | Image feature extraction method for pedestrian re-identification | |
| CN106651872B (en) | Pavement crack identification method and system based on Prewitt operator | |
| CN108230264B (en) | Single image defogging method based on ResNet neural network | |
| CN111179232A (en) | Steel bar size detection system and method based on image processing | |
| CN111027446B (en) | Coastline automatic extraction method of high-resolution image | |
| CN104616290A (en) | Target detection algorithm in combination of statistical matrix model and adaptive threshold | |
| CN109685045A (en) | A kind of Moving Targets Based on Video Streams tracking and system | |
| CN110070545B (en) | Method for automatically extracting urban built-up area by urban texture feature density | |
| CN104182976B (en) | Field moving object fining extraction method | |
| CN115272306B (en) | Solar cell panel grid line enhancement method utilizing gradient operation | |
| CN110969164A (en) | Low-illumination imaging license plate recognition method and device based on deep learning end-to-end | |
| Liu et al. | Robust image-based crack detection in concrete structure using multi-scale enhancement and visual features | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| Du et al. | Research on an efficient method of license plate location | |
| CN118297837A (en) | Infrared simulator virtual image enhancement system based on image processing | |
| CN114519694A (en) | Seven-segment digital tube liquid crystal display screen identification method and system based on deep learning | |
| CN110047041A (en) | A kind of empty-frequency-domain combined Traffic Surveillance Video rain removing method | |
| CN110276260B (en) | Commodity detection method based on depth camera | |
| CN114022473B (en) | Horizon detection method based on infrared image | |
| CN114758139B (en) | Method for detecting accumulated water in foundation pit | |
| CN114119634B (en) | Building automatic extraction method and system combining vegetation rejection and image feature consistency constraint | |
| Aung et al. | Study for license plate detection | |
| CN108416768B (en) | Binary-based foreground image similarity evaluation method | |
| Lv et al. | Machine learning based road detection from high resolution imagery | |
| Idder et al. | Unsupervised multi-level thresholding method for weather satellite cloud segmentation |
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 |