CN114758119A - Sea surface recovery target detection method based on eagle eye-imitated vision and similarity - Google Patents
Sea surface recovery target detection method based on eagle eye-imitated vision and similarity Download PDFInfo
- Publication number
- CN114758119A CN114758119A CN202210414687.8A CN202210414687A CN114758119A CN 114758119 A CN114758119 A CN 114758119A CN 202210414687 A CN202210414687 A CN 202210414687A CN 114758119 A CN114758119 A CN 114758119A
- Authority
- CN
- China
- Prior art keywords
- color
- image
- eagle eye
- similarity
- window
- 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.)
- Granted
Links
- 238000011084 recovery Methods 0.000 title claims abstract description 50
- 238000001514 detection method Methods 0.000 title claims abstract description 45
- 230000004438 eyesight Effects 0.000 title claims abstract description 24
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 18
- 230000008569 process Effects 0.000 claims abstract description 11
- 238000009499 grossing Methods 0.000 claims abstract description 8
- 238000004088 simulation Methods 0.000 claims abstract description 5
- 230000007246 mechanism Effects 0.000 claims description 17
- 239000003086 colorant Substances 0.000 claims description 15
- 238000010606 normalization Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 6
- 241000272184 Falconiformes Species 0.000 description 47
- 230000010365 information processing Effects 0.000 description 8
- 210000001525 retina Anatomy 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 230000035945 sensitivity Effects 0.000 description 4
- 230000003042 antagnostic effect Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 108091008695 photoreceptors Proteins 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000001154 acute effect Effects 0.000 description 2
- 235000021466 carotenoid Nutrition 0.000 description 2
- 150000001747 carotenoids Chemical class 0.000 description 2
- 230000004456 color vision Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 244000062804 prey Species 0.000 description 1
- 238000012958 reprocessing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a sea surface recovery target detection method based on eagle eye simulation vision and similarity, which comprises the following steps: acquiring a similar property window image; step two: establishing an eagle eye imitating color space; step three: selecting an image with higher score for subsequent processing; step four: calculating a saliency value of the same color; step five: carrying out space smoothing treatment on the eagle eye-imitated color; step six: summing the similarity window significance results; step seven: and outputting a significance detection result of the offshore recovery device. According to the invention, the significance detection is carried out on the window result of the similarity detection, so that the algorithm complexity of the significance detection is greatly reduced; according to the method, the eagle eye-imitated color space is introduced into the process of detecting the saliency characteristics of the similarity window, and the distance relation between the pixels of the real image can be more closed in the process of evaluating the distance of the color space of the image.
Description
Technical Field
The invention relates to a sea surface recovered target significance detection research method based on biological vision, in particular to a sea surface recovered target detection method based on eagle eye simulation vision and similarity, and belongs to the field of computer vision.
Background
The unmanned aerial vehicle is a unmanned aerial vehicle which is operated by using a radio remote control device and a self-contained program control device, and is a product with high technical content in the information era.
With the three-dimensional and multi-level rapid development of modern sea warfare, small unmanned boats, unmanned ships, intelligent floating recovery devices and the like have the advantages of convenience, high flexibility, high autonomy and the like, and can realize approaching, surveying and reconnaissance of complex sea surface environments, dangerous sea areas and unknown island environments, so that the research and the application are more and more extensive. After the unmanned aerial vehicle completes all-weather autonomous patrol and monitoring tasks in a complex sea surface environment of a coastal zone, an important means for timely recovering the unmanned aerial vehicle completing the tasks to fully and repeatedly utilize the resources of the unmanned aerial vehicle is provided. Unmanned aerial vehicle retrieves to be an important performance of aassessment unmanned aerial vehicle performance, and the flexibility degree, the accuracy and the reliability of retrieving the mode are an important index of aassessment unmanned aerial vehicle combat power. Therefore, the research of the water-air cooperative sea surface patrol unmanned system has urgent needs on the technologies of unmanned aerial vehicle/unmanned ship target identification, environment perception, autonomous control and optimization, cooperative control and the like.
Traditional unmanned aerial vehicle retrieves the mode and retrieves, the parachute is retrieved, is hit net recovery, gasbag shock attenuation is retrieved, undercarriage pulley is retrieved, is hit the line and retrieves etc. including independently descending, nevertheless to the unmanned ship of recovery on the sea, do not have longer deck of taking off and land, can supply to wait to retrieve the region that unmanned aerial vehicle landed less, this accurate recovery descending of giving unmanned aerial vehicle has brought huge challenge. In addition, whether the unmanned ship to be recovered can accurately identify the unmanned ship target to be landed in the complex sea surface environment is a prerequisite for realizing recovery landing. In the unmanned aerial vehicle recovery process, a combined navigation system combining a GPS navigation system, an inertial navigation system and various navigation systems is usually adopted. Although the GPS navigation system can realize all-weather uninterrupted positioning and navigation, the GPS updating frequency is low, the precision is low, the real-time navigation of the unmanned aerial vehicle in the recovery process cannot be ensured, and the unmanned aerial vehicle is very easy to damage and lose efficacy in a special war period. The inertial navigation system does not receive and send information from the outside, has good concealment, strong anti-interference capability and higher updating frequency than a GPS, but can generate error accumulation along with the time. The combined navigation mode can make the navigation system have higher precision and more reliable by combining the advantages of various navigation modes, but the navigation mode has higher cost, and when one of the navigation modes fails, the defects of the GPS navigation mode or the inertial navigation mode can occur.
With the rapid development of computers and image processing, visual navigation is increasingly well known. The visual navigation has the advantages of simple structure, large information amount and low power consumption, and improves the autonomous level in the navigation process while improving the interference resistance in the navigation process. Compared with the traditional navigation mode, the visual navigation is more suitable for navigation in the processes of sea surface recovery target identification, tracking and landing in the recovery process of the sea surface unmanned aerial vehicle, and in the processes, high navigation precision is the key for determining whether recovery is successful or not.
The vision of eagle's eye is extremely acute, and eagle can accurately identify a prey on the ground from a high altitude of several kilometers and track the prey at a very fast speed until it is captured. The unique color space of eagle eye can carry out clear color classification. The acute vision of eagle is closely related to its double fovea, which is several times as dense as the photoreceptors of the fovea, and thus eagle eye has higher resolution. Combine together hawk eye vision and traditional vision navigation mode, full play hawk eye vision's advantage can be accurate, quick catch sea recovery unit target to the realization is to unmanned aerial vehicle's recovery task.
In conclusion, the invention provides a sea surface recovery target detection method based on eagle eye simulation vision and similarity, so as to solve the problem that a recovery device is difficult to detect on line in real time under a sea surface environment and effectively improve the success rate of recovery of a sea surface unmanned aerial vehicle.
Disclosure of Invention
The invention provides a sea surface recovery target detection method based on eagle eye simulation vision and similarity, and aims to provide a method for detecting a recovery device target in a real-time online complex sea surface background environment, so that the target detection level of the recovery device in the complex sea surface environment and the recovery efficiency of an unmanned aerial vehicle to be recovered are effectively improved.
The invention discloses a sea surface recovery target detection method based on eagle eye-imitated vision and similarity aiming at the problem of target identification and detection of a marine recovery device, a structural framework of the method is shown as figure 1, and the method comprises the following specific steps:
the method comprises the following steps: obtaining analogous windowed images
And (3) carrying out analog detection on the original input image I by using an analog detection method of binary gradient normalization BING, framing N analog window regions, and obtaining a score of each analog window. The scores are sorted from high to low, and the score condition is used for representing the probability omega that the window contains the target feature i＞ωj,i＜j,i,j＝1,2,…,N。
Step two: establishing an eagle eye-imitating color space
The color perception is increased in the visual world of hawks, helping hawks to distinguish environmental preys during the predation process. Colored oil drops exist in the distal ends of the insides of the cones of the eagle eye retinas, and the concentrations and types of the carotenoids contained in the oil drops are different. Thus, the retina of an eagle typically has a maximum of spectral sensitivity in the approximate regions of red, green, and blue, with a fourth spectrum of sensitivity in the ultraviolet or near ultraviolet range, and thus the color space of the eagle eye is a four-color system. The signals extract color information from photoreceptors (cone cells and rod cells) of the retina, and color signals with different color contrasts are generated through a series of color processing mechanisms, so that a colorful color space rich in the eagle eye is formed, wherein the vision processing mechanism of the eagle eye on the received color information is shown in fig. 2.
The hawk eye can identify 3 color signals of long wave, medium wave and short wave, and the 3 color signals are mutually antagonistic in the next layer of nerve processing mechanism. Defining the long wave, medium wave and short wave signals as red, green and blue, obtaining brightness channel by color processing mechanism, as following formula
Where r, g, b are red, green and blue, respectively, in the original input image signal. L, M, S, Y are the long wave, medium wave, short wave and brightness signal of eagle eye space, which respectively represent the color contrast signal after the color signal is processed by eagle eye.
Step three: selecting the image with higher score for subsequent processing
N similarity windows are separately extracted and saved as separate images I1，I2，……，IN. Selecting the top with higher score
Processing the M images in the fourth to fifth steps to obtain M primary saliency maps S (I)1),S(I2),…,S(IM)。
Step four: calculating saliency values for the same color
Window image IiMiddle pixel IikIs defined as a significance value of
Wherein D (I)ik,Iii) Is a pixel IikAnd a pixel IiiIn the measurement of color distance in the eagle eye-imitating color space, the formula (2) can be expanded into
S(Iik)＝D(Iik,Ii1)+D(Iik,Ii2)+…+D(Iik,Iim) (3)
Wherein m is a window image IiThe number of pixels of (c). As can be seen from equation (3), the pixels of the same color have the same color saliency value.
ciAs window image IiWill have the same color value c, is a selected color valueiThe saliency map of each color can be obtained by combining the pixels of (1) and (4).
Wherein, cijAs window image IiThe color value of each pixel in the image IiHas a color value of ciTotal number of pixels of fiIs a color value ciIn picture IiThe frequency of occurrence of (a).
Step five: eagle eye-imitating color space smoothing treatment
The RGB color space contains H-256 color values, but not all of the H colors are present in the test image. To improve the efficiency of the algorithm and to increase the distinguishability of the color information, the method reduces the number of colors to be considered.
Let image IiThe number of middle colors is PiQuantizing the colors of the R, G, B color channels to obtain QiA different value. At this time, image IiIn the existing color number of Qi 3And the complexity of the method is greatly reduced, and the efficiency is improved. Q is taken on the value of
Selecting an image IiMiddle frequency of occurrence fiMore than 90% of the pixels, the remaining 10% of the pixels are replaced by the median of their neighbors. To reduce the error that such operations introduce in the computation of saliency values, the present invention employs smoothing operations to improve the color of the replaced pixelSignificance value of (a).
The significance value S (c) of each replaced colori) With its weighted average of adjacent color saliency values S' (c)i) Instead, the adjacent color saliency values are obtained by distance measurement in an eagle eye-like color space. Selecting
Color of a neighbor pairiThe significance value of the replaced color is obtained preliminarily:
wherein, TiIs a color ciAnd q thereofiThe median of the distances of the neighbors can be expressed as
Will S' (c)i) Normalization is as follows:
wherein the normalization factor is
Wherein (T)i-D(ci,cij) Defined as a linearly varying smooth weight for giving the distance color c in the color feature spaceiThe closer color is assigned with a larger weight, and the effect of the linearly changing weight is better than that of the Gaussian weight changing with a larger change rate. Wherein D (c) i-cij)Is a color ciAnd color cijTi is calculated from the formula (7).
Step six: semblance Window saliency result summation
M primary saliency maps S (I)1),S(I2),…,S(IM) Restoring to the corresponding position in the original image I, setting the rest part in I as 0, and finally forming M saliency maps S (I)1′),S(I2′),…,S(IM') then the saliency map of image I may be represented as
S＝ω1S(I1′)+ω2S(I2′)+…+ωMS(IM′) (10)
Wherein, ω isiI is 1,2, …, and N is window IiContaining the probability of a significant object.
Step seven: outputting a sea recovery device significance detection result
Through the calculation of the first step to the sixth step, a final sea recovery device significance detection result is obtained, the feature points are selected for position estimation, then the relative position relation between the unmanned aerial vehicle to be recovered and the sea recovery device (such as an unmanned ship, an intelligent recovery device and the like) is estimated, and the unmanned aerial vehicle is finally recovered.
The invention provides a sea surface recovery target detection method based on eagle eye vision and similarity. The method comprises the steps of reprocessing window images of the similarity obtained by a similarity detection algorithm, and performing significance detection on each window image to further realize significance detection on a marine recovery target. The main advantages of the invention are mainly embodied in the following 2 aspects: 1) the significance detection is carried out on the window result of the similarity detection, so that the algorithm complexity of the significance detection is greatly reduced; 2) the eagle eye-imitating color space is introduced into the process of detecting the saliency characteristics of the similarity window, and the distance relation between the pixels of the real image can be more closely approached in the process of evaluating the color space distance of the image.
Drawings
FIG. 1 is a flow chart of the method of the present invention
FIG. 2 is a schematic view of the mechanism of the eagle eye-imitating vision processing
FIG. 3 is a graph showing the results of the detection of the similarity
FIG. 4 significance test result graph
FIG. 5 cooperative targets and feature points results plot
The reference numbers and symbols in the figures are as follows:
ω1probability that analog window 1 may contain a salient object
ωiProbability that the similarity window i may contain a significant object
ωMProbability that the anologue window M may contain a significant target
I1-likeness image 1 (window possibly containing salient objects) obtained by likeness detection
Ii-likeness image i (window possibly containing salient objects) obtained by likeness detection
IM-an likeness image M (a window possibly containing a salient object) obtained by likeness detection
RGB-Red, Green, blue color space
S(I1k) -color saliency value of likeness image 1 in RGB color space
S(Iik) -color saliency value of likeness image i in RGB color space
S(IMk) -color saliency value of likeness image M in RGB color space
f-frequency of occurrence of certain color pixel in detected window image
RY0-reference luminance signal of eagle eye color information processing mechanism
RS0Reference short wave signal of eagle eye color information processing mechanism
RM0Reference medium wave signal of eagle's eye color information processing mechanism
RL0Reference long wave signal of eagle's eye color information processing mechanism
RY-input luminance signal of eagle eye color information processing mechanism
RSInput short wave signal of eagle eye color information processing mechanism
RMInput medium wave signal of eagle eye color information processing mechanism
RLInput Long wave Signal of eagle eye color information processing mechanism
ΔSc-color contrast response of cone cells of each color type antagonistic to cone cells of other colors
P-output signal after brain processing
Y-is (satisfies the condition)
N-No (unsatisfied with condition)
R1～R4Feature points of cooperative targets
Detailed Description
The effectiveness of the method provided by the invention is verified by a specific target detection example of the offshore recovery unmanned ship. The experimental computer is configured with an Intel Core i7-4790 processor, 3.60Ghz dominant frequency, 4G memory, and software as MATLAB 2014a version. A sea surface recovery target detection method based on eagle eye vision and similarity comprises the following specific steps:
the method comprises the following steps: obtaining window images of the likeness
And (3) performing analog detection on the original input image I by using an analog detection method of binary gradient normalization BING, selecting N-12 analog window areas, and obtaining a score of each analog window. The scores are sorted from high to low, and the score condition is used for representing the probability omega that the window contains the target feature i＞ωjI < j, i, j ═ 1,2, …, N, the results of the analogous assays are shown in fig. 3.
Step two: establishing an eagle eye-imitating color space
The increased color perception in the visual world of hawks helps hawks to distinguish prey in the environment when not in composition. Colored oil drops exist in the distal ends of the insides of the cones of the eagle eye retinas, and the concentrations and types of the carotenoids contained in the oil drops are different. Thus, the retina of an eagle typically has a maximum of spectral sensitivity in the approximate regions of red, green, and blue, with a fourth spectrum of sensitivity in the ultraviolet or near ultraviolet range, and thus the color space of the eagle eye is four-color. The signals extract color information from photoreceptors (cone cells and rod cells) of the retina, and color signals with different color contrasts are generated through a series of color processing mechanisms, so that a colorful color space of the eagle eye is formed.
The hawk eye can identify 3 color signals of long wave, medium wave and short wave, and the 3 color signals are mutually antagonistic in the next layer of nerve processing mechanism. The long-wave, medium-wave and short-wave signals are defined as red, green and blue, and a brightness channel can be obtained through a color processing mechanism, as shown in the following formula (1).
Where r, g, b are red, green and blue, respectively, in the original input image signal. L, M, S, Y are the long wave, medium wave, short wave and brightness signal of eagle eye space, which respectively represent the color contrast signal after the color signal is processed by eagle eye.
Step three: selecting the image with higher score for post-processing
N similarity windows are extracted and stored separately as a single image I1，I2，……，IN. Selecting the top with higher score
And (4) sequentially carrying out the processing of the fourth step to the fifth step on the M images respectively, and finally obtaining 8 primary saliency maps.
Step four: calculating a saliency value of a same color
Window image IiMiddle pixel IikThe significance value of (a) is defined as formula (2). In the formula, D (I)ik,Iii) Is a pixel IikAnd a pixel IiiIn the measurement of the color distance in the eagle eye-imitating color space, the formula (2) can be expanded into the formula (3), wherein M is a window image IiThe number of pixels of (c). As can be seen from equation (3), the pixels of the same color have the same color saliency value.
Will have the same color value ciThe saliency map of each color can be obtained by combining the pixels of (1) and (4). In the formula, cijAs window image IiColor value of each pixel in the arrayN is the image IiHas a color value of ciTotal number of pixels of fiIs a color value ciIn picture IiThe frequency of occurrence of (a).
Step five: eagle eye-imitating color space smoothing treatment
The RGB color space contains H-256 color values, but not all of the H colors are present in the test image. To improve the efficiency of the algorithm and to increase the distinguishability of the color information, the method reduces the number of colors to be considered.
Let image IiThe number of middle colors is PiQuantizing the colors of the R, G, B color channels to obtain QiA different value. At this time, image IiThe existing color number is Qi 3And the complexity of the method is greatly reduced, and the efficiency is improved. The value of Q is defined as (5).
Selecting an image IiOf which more than 90% of the pixels occur with the remaining 10% replaced by the median of their neighbors. To reduce the error that such operations introduce to the saliency value calculation, the present invention employs a smoothing operation to improve the saliency value of the color of the replaced pixel.
The significance value S (c) of each replaced colori) With its weighted average of adjacent color saliency values S' (c)i) Instead, the adjacent color saliency values are obtained by distance measurement in an eagle eye-like color space. Selecting
Color of a neighbor pairiThe significance value of the substituted color is preliminarily obtained by improving the significance value of the substituted color as shown in the formula (6). In the formula, TiIs a color ciAnd q thereofiThe median of the distances of the neighbors can be expressed as equation (7).
Will S' (c)i) The normalization is shown as formula (8). In the formula, the normalization factor is shown in formula (6). Wherein (T-D (c)i,cij) ) is a linearly varying smooth weight for the distance color c in the color feature space iCloser colors are assigned greater weightsThe linearly changing weight value has the changing effect of the Gaussian weight value with a larger changing rate.
Step six: similarity window saliency result summation
The M saliency maps S (I) obtained in the step five1 S),S(I2 S),…,S(IM S) Restoring to the corresponding position in the original image I, setting the rest part in I as 0, and forming M saliency maps S (I)1′),S(I2′),…,S(IM') the saliency map of image I can be represented by equation (10). In the formula, ωiI is 1,2, …, and N is the detected window IiPossibly including the probability of the object.
Step seven: outputting a sea recovery device significance detection result
And obtaining a final sea recovery device significance detection result through the calculation from the first step to the sixth step. On the offshore recovery device, a cooperative target (such as a circular target in the center of the ship in fig. 5) with characteristic points is obtained, and vertexes (such as R in fig. five) of a circumscribed quadrangle (indicated by a white box in fig. 5) of the cooperative target are obtained1～R4) And (3) as feature points, matching the feature points, and estimating the relative position relationship between the unmanned aerial vehicle to be recovered and a marine recovery device (such as an unmanned ship, an intelligent recovery device and the like) by adopting an RPnP pose estimation algorithm to finally realize the recovery of the unmanned aerial vehicle. The detection output result of the significance of the offshore recovery target based on the method of the invention is shown in figure 4.
Claims (6)
1. A sea surface recovery target detection method based on eagle eye simulation vision and similarity is characterized by comprising the following steps: the method comprises the following steps:
the method comprises the following steps: obtaining window images of the likeness
Carrying out analog detection on an original input image I, selecting N analog window areas in a frame, and obtaining a score of each analog window; the scores are sorted from high to low, and the score condition is used for representing the probability omega that the window contains the target featurei＞ωj,i＜j,i,j＝1,2,…,N；
Step two: establishing an eagle eye-imitating color space
The eagle eye can identify 3 color signals of long wave, medium wave and short wave, the long wave, medium wave and short wave signals are defined as red, green and blue, and a brightness channel can be obtained through a color processing mechanism, as shown in the following formula
Wherein r, g, b are respectively red, green and blue in the original input image signal; l, M, S, Y are respectively long wave, medium wave, short wave and brightness signal of eagle eye space, respectively representing color contrast signal after eagle eye processing;
step three: selecting the image with higher score for subsequent processing
N similarity windows are separately extracted and saved as separate images I1，I2，……，IN(ii) a Selecting the top with higher score
Processing the M images in the fourth to fifth steps to obtain M primary saliency maps S (I) 1),S(I2),…,S(IM)；
Step four: calculating a saliency value of a same color
Window image IiMiddle pixel IikIs defined as a significance value of
Wherein D (I)ik,Iii) Is a pixel IikAnd a pixel IiiIn the measurement of the color distance in the eagle eye-imitated color space, the formula (2) can be expanded to
S(Iik)＝D(Iik,Ii1)+D(Iik,Ii2)+…+D(Iik,Iim) (3)
Wherein m is a window image IiThe number of pixels of (a); as can be seen from equation (3), pixels of the same color have the same color saliency value;
cifor a color value in the window image Ii, it will have the same color value ciThe pixels of (a) are combined together to obtain a saliency map of each color, as shown in formula (4);
wherein, cijAs window image IiThe color value of each pixel in the image IiHas a color value of ciTotal number of pixels of fiIs a color value ciIn picture IiThe frequency of occurrence of;
step five: eagle eye-imitating color space smoothing treatment
The RGB color space contains 256 color values, and the method reduces the number of colors to be considered;
step six: the similarity window significance results are summed.
2. The sea surface recovery target detection method based on the imitative eagle eye vision and the similarity, according to claim 1, is characterized in that: the method further comprises the following steps:
step seven: outputting a sea recovery device significance detection result
And calculating through the first step to the sixth step to obtain a final sea recovery device significance detection result, selecting the characteristic points to carry out position estimation, estimating the relative position relation between the unmanned aerial vehicle to be recovered and the sea recovery device, and finally realizing recovery of the unmanned aerial vehicle.
3. The sea surface recovery target detection method based on the eagle eye imitating vision and the similarity as claimed in claim 1 or 2, characterized in that: the process of reducing the number of colors to be considered as described in step five is as follows:
let image IiThe number of middle colors is PiQuantizing the colors of the R, G, B color channels to obtain QiA number of different values; at this time, image IiIn the existing color number of Qi 3Q takes the value as
Selecting an image IiMiddle frequency of occurrence fiMore than 90% of the pixels, the remaining 10% of the pixels are replaced by the median of their neighbors.
4. The sea surface recovery target detection method based on the imitative eagle eye vision and the similarity, according to claim 3, is characterized in that: to reduce the error in the computation of the saliency value due to the reduction of the number of colors to be considered, a smoothing operation is used to improve the saliency value of the color of the pixel being replaced.
5. The sea surface recovery target detection method based on the imitative eagle eye vision and the similarity, according to claim 4, is characterized in that: the method for improving the significance value of the color of the replaced pixel by adopting the smoothing operation comprises the following specific steps:
The significance value S (c) of each replaced colori) With its weighted average of the adjacent color saliency values S ″ (c)i) Instead, the adjacent color saliency values are obtained by distance measurement in an eagle eye-imitated color space; selecting
Color of one neighbor pairiThe significance value of the replaced color is preliminarily obtained by improving the significance value of the replaced color:
wherein Ti is the color ciAnd q thereofiThe median of the distances of the neighbors can be expressed as
Will S' (c)i) Normalization is as follows:
wherein the normalization factor is
Wherein (T)i-D(ci,cij) Defined as a linearly varying smooth weight for giving the distance color c in the color feature spaceiThe closer color is distributed with a larger weight, and the weight of the linear change is superior to the effect of the change of the Gaussian weight with a larger change rate; wherein D (ci-cij) is the distance between the color ci and the color cij, and Ti is calculated by the formula (7).
6. The sea surface recovery target detection method based on the eagle eye-imitated vision and the similarity is characterized in that: the concrete process of the step six is as follows:
m primary saliency maps S (I)1),S(I2),…,S(IM) Restoring to the corresponding position in the original image I, setting the rest part in I as 0, and finally forming M saliency maps S (I) 1′),S(I2′),…,S(IM') then the saliency map of image I may be represented as
S＝ω1S(I1′)+ω2S(I2′)+…+ωMS(IM′) (10)
Wherein, ω isiI is 1,2, …, and N is window IiContaining the probability of a significant object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210414687.8A CN114758119B (en) | 2022-04-20 | 2022-04-20 | Sea surface recovery target detection method based on eagle eye imitating vision and similar physical properties |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210414687.8A CN114758119B (en) | 2022-04-20 | 2022-04-20 | Sea surface recovery target detection method based on eagle eye imitating vision and similar physical properties |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114758119A true CN114758119A (en) | 2022-07-15 |
| CN114758119B CN114758119B (en) | 2024-06-07 |
Family
ID=82331203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210414687.8A Active CN114758119B (en) | 2022-04-20 | 2022-04-20 | Sea surface recovery target detection method based on eagle eye imitating vision and similar physical properties |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114758119B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI802514B (en) * | 2022-10-07 | 2023-05-11 | 國立中興大學 | Processing method of target identification for unmanned aerial vehicle (uav) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392963A (en) * | 2017-06-28 | 2017-11-24 | 北京航空航天大学 | A kind of imitative hawkeye moving target localization method for soft autonomous air refuelling |
| CN110084782A (en) * | 2019-03-27 | 2019-08-02 | 西安电子科技大学 | Full reference image quality appraisement method based on saliency detection |
| WO2020107717A1 (en) * | 2018-11-30 | 2020-06-04 | 长沙理工大学 | Visual saliency region detection method and apparatus |
| CN111950549A (en) * | 2020-08-12 | 2020-11-17 | 上海大学 | Sea surface obstacle detection method based on fusion of sea antennas and visual saliency |
| CN112101099A (en) * | 2020-08-04 | 2020-12-18 | 北京航空航天大学 | Eagle eye self-adaptive mechanism-simulated unmanned aerial vehicle sea surface small target identification method |
| CN112232181A (en) * | 2020-10-14 | 2021-01-15 | 北京航空航天大学 | Eagle eye color cognitive antagonism mechanism-simulated unmanned aerial vehicle marine target detection method |
| CN114200439A (en) * | 2021-12-09 | 2022-03-18 | 哈尔滨工程大学 | Multi-mode airspace target tracking method based on Doppler blind area |
-
2022
- 2022-04-20 CN CN202210414687.8A patent/CN114758119B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392963A (en) * | 2017-06-28 | 2017-11-24 | 北京航空航天大学 | A kind of imitative hawkeye moving target localization method for soft autonomous air refuelling |
| WO2020107717A1 (en) * | 2018-11-30 | 2020-06-04 | 长沙理工大学 | Visual saliency region detection method and apparatus |
| CN110084782A (en) * | 2019-03-27 | 2019-08-02 | 西安电子科技大学 | Full reference image quality appraisement method based on saliency detection |
| CN112101099A (en) * | 2020-08-04 | 2020-12-18 | 北京航空航天大学 | Eagle eye self-adaptive mechanism-simulated unmanned aerial vehicle sea surface small target identification method |
| CN111950549A (en) * | 2020-08-12 | 2020-11-17 | 上海大学 | Sea surface obstacle detection method based on fusion of sea antennas and visual saliency |
| CN112232181A (en) * | 2020-10-14 | 2021-01-15 | 北京航空航天大学 | Eagle eye color cognitive antagonism mechanism-simulated unmanned aerial vehicle marine target detection method |
| CN114200439A (en) * | 2021-12-09 | 2022-03-18 | 哈尔滨工程大学 | Multi-mode airspace target tracking method based on Doppler blind area |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI802514B (en) * | 2022-10-07 | 2023-05-11 | 國立中興大學 | Processing method of target identification for unmanned aerial vehicle (uav) |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114758119B (en) | 2024-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220197281A1 (en) | Intelligent decision-making method and system for unmanned surface vehicle | |
| CN114202696B (en) | SAR target detection method and device based on context vision and storage medium | |
| CN107818326B (en) | A kind of ship detection method and system based on scene multidimensional characteristic | |
| CN106384344B (en) | A kind of remote sensing image surface vessel target detection and extracting method | |
| CN104933680B (en) | A kind of intelligent quick sea fog minimizing technology of unmanned boat vision system video | |
| CN110189304B (en) | Optical remote sensing image target on-line rapid detection method based on artificial intelligence | |
| CN109271856A (en) | Remote sensing image object detection method based on expansion residual error convolution | |
| CN111626290A (en) | Infrared ship target detection and identification method under complex sea surface environment | |
| CN111985274B (en) | Remote sensing image segmentation method based on convolutional neural network | |
| CN108460341A (en) | Remote sensing image object detection method based on integrated depth convolutional network | |
| CN106023257A (en) | Target tracking method based on rotor UAV platform | |
| CN109344878B (en) | Eagle brain-like feature integration small target recognition method based on ResNet | |
| CN111462131A (en) | Method and equipment for attention-driven image segmentation | |
| Dos Santos et al. | Matching color aerial images and underwater sonar images using deep learning for underwater localization | |
| CN107967474A (en) | A kind of sea-surface target conspicuousness detection method based on convolutional neural networks | |
| CN106295503A (en) | The high-resolution remote sensing image Ship Target extracting method of region convolutional neural networks | |
| CN113495575B (en) | Unmanned aerial vehicle autonomous landing visual guidance method based on attention mechanism | |
| He et al. | Ship detection without sea-land segmentation for large-scale high-resolution optical satellite images | |
| CN116486243A (en) | DP-ViT-based sonar image target detection method | |
| CN114758119A (en) | Sea surface recovery target detection method based on eagle eye-imitated vision and similarity | |
| CN108765439A (en) | A kind of sea horizon detection method based on unmanned water surface ship | |
| Du et al. | SARNas: A hardware-aware SAR target detection algorithm via multiobjective neural architecture search | |
| CN116246139A (en) | Target identification method based on multi-sensor fusion for unmanned ship navigation environment | |
| CN116030298A (en) | Scene complexity classification method, storage medium and device for ship navigation image | |
| CN115100516A (en) | Relation learning-based remote sensing image target detection method |
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 |