CN117115436A - Ship attitude detection method and device, electronic equipment and storage medium - Google Patents
Ship attitude detection method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117115436A CN117115436A CN202310839541.2A CN202310839541A CN117115436A CN 117115436 A CN117115436 A CN 117115436A CN 202310839541 A CN202310839541 A CN 202310839541A CN 117115436 A CN117115436 A CN 117115436A
- Authority
- CN
- China
- Prior art keywords
- ship
- image
- target
- wake
- semantic segmentation
- 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 106
- 230000011218 segmentation Effects 0.000 claims abstract description 139
- 238000000034 method Methods 0.000 claims abstract description 42
- 238000012706 support-vector machine Methods 0.000 claims description 71
- 230000006870 function Effects 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 238000010845 search algorithm Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 230000000007 visual effect Effects 0.000 abstract description 2
- 230000003044 adaptive effect Effects 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 11
- 238000009826 distribution Methods 0.000 description 9
- 230000000877 morphologic effect Effects 0.000 description 8
- 238000000605 extraction Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 239000000523 sample Substances 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 230000006978 adaptation Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000005286 illumination Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 101000872083 Danio rerio Delta-like protein C Proteins 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 239000013074 reference sample Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Remote Sensing (AREA)
- Astronomy & Astrophysics (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a ship attitude detection method, a device, electronic equipment and a storage medium, and relates to the technical field of visual detection, wherein the method comprises the following steps: based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to a wake of the target ship; determining a first center of a ship semantic segmentation area and a second center of a wake semantic segmentation area; determining a first attitude direction of the target ship based on the first center and the second center; determining the long axis direction and the short axis direction of the ship semantic segmentation area, determining the second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship. The invention can effectively realize the rapid and accurate detection of the marine ship attitude.
Description
Technical Field
The present invention relates to the field of visual detection technologies, and in particular, to a method and apparatus for detecting a ship gesture, an electronic device, and a storage medium.
Background
The automatic detection and tracking of the marine ship targets have important significance for guaranteeing the safety of deep sea routes, implementing sea area supervision, fighting offshore crimes and the like. The satellite remote sensing image has the characteristics of visualization and high precision, and has a key effect on obtaining the state of a ship target at a specific moment. With the development of the technology of 'satellite chain', the shooting cost of satellites is obviously reduced, so that satellite remote sensing plays an increasingly important role in the detection of marine ship targets.
However, the ship target often has only a small area ratio in a wide space-based remote sensing image, and the interference caused by cloud layers, islands and noise points is large, so that false alarms or missed detection are easily caused, the detection precision is influenced, and besides ship position information detection, the information of ship heading (hereinafter collectively referred to as ship attitude) is very important for the track analysis, track prediction and the like of the ship target.
In the related art, the method is only often oriented to the position detection task of the ship, and the detection of the ship posture is not concerned. Therefore, how to accurately detect the attitude of the marine ship becomes a problem to be solved in the industry.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a ship attitude detection method, a device, electronic equipment and a storage medium.
In a first aspect, the present invention provides a ship attitude detection method, including:
based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship;
determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
determining a first attitude direction of the target ship based on the first center and the second center;
determining a long axis direction and a short axis direction of the ship semantic segmentation area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
Optionally, according to the ship attitude detection method provided by the present invention, the obtaining, based on a global remote sensing image including a target ship, a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to a wake of the target ship includes:
based on a YOLO detection network, extracting a ship region image in which a target ship is located and a wake region image in which a wake of the target ship is located from the global remote sensing image containing the target ship;
And respectively carrying out semantic segmentation processing on the ship area image and the wake area image to obtain the ship semantic segmentation area and the wake semantic segmentation area.
Optionally, according to the ship gesture detection method provided by the present invention, the semantic segmentation processing is performed on the ship region image and the wake region image respectively, to obtain the ship semantic segmentation region and the wake semantic segmentation region, including:
inputting the ship area image and the global remote sensing image into a pre-constructed Support Vector Machine (SVM) semantic classifier, obtaining a first classification result output by the SVM semantic classifier, and determining the ship semantic segmentation area based on the first classification result;
inputting the wake area image and the global remote sensing image into the SVM semantic classifier to obtain a second classification result output by the SVM semantic classifier, and determining the wake semantic segmentation area based on the second classification result.
Optionally, according to the ship gesture detection method provided by the present invention, before inputting the ship region image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the wake region image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further includes:
Initializing a target feature recognition threshold;
taking the target feature recognition threshold value as a bias parameter in an optimal interface judgment function of the SVM semantic classifier, and optimizing the target feature recognition threshold value based on a heuristic search algorithm to obtain the optimized target feature recognition threshold value;
substituting the optimized target feature recognition threshold value into an optimal interface judgment function of the SVM semantic classifier to obtain the optimized optimal interface judgment function;
and constructing the SVM semantic classifier based on the optimized optimal interface judgment function.
Optionally, according to the ship gesture detection method provided by the present invention, before inputting the ship region image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the wake region image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further includes:
based on Min-Max standardization thought, the ship area image and the wake area image are respectively subjected to linear calibration pretreatment.
Optionally, according to the ship attitude detection method provided by the invention, based on Min-Max standardization thought, the linear calibration preprocessing is performed on the ship area image and the wake area image respectively, and the method comprises the following steps:
And respectively carrying out linear calibration pretreatment on the ship area image and the wake area image based on the following formula:
where x represents a feature component of the target image,representing the characteristic component, x, of the target image after calibration 0.05 0.05 quantile, x representing a feature component of the target image 0.95 0.95 quantiles, x representing characteristic components of the target image 0.05 (ref) 0.05 minutes, x representing the characteristic component of the reference image 0.95 (ref) represents a 0.95 quantile of a feature component of the reference image, the target image being the ship region image or the wake region image.
Optionally, according to the ship attitude detection method provided by the present invention, the determining the major axis direction and the minor axis direction of the ship semantic division area includes:
and determining the major axis direction and the minor axis direction of the ship semantic segmentation area based on principal component analysis PCA and singular value decomposition SVD algorithm.
In a second aspect, the present invention also provides a ship attitude detection apparatus, including:
the acquisition module is used for acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship based on a global remote sensing image containing the target ship;
The first determining module is used for determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
the second determining module is used for determining a first gesture direction of the target ship based on the first center and the second center;
and the third determining module is used for determining the long axis direction and the short axis direction of the ship semantic segmentation area, determining the second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
In a third aspect, the present invention also provides an electronic device, including a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the ship attitude detection method according to the first aspect when executing the program.
In a fourth aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the ship attitude detection method according to the first aspect.
In a fifth aspect, the invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the ship attitude detection method according to the first aspect.
According to the ship attitude detection method, the device, the electronic equipment and the storage medium, the ship semantic segmentation area corresponding to the target ship and the wake semantic segmentation area corresponding to the wake of the target ship are obtained based on the global remote sensing image containing the target ship, then the first attitude direction of the target ship is determined based on the first center of the ship semantic segmentation area and the second center of the wake semantic segmentation area, further the long axis direction and the short axis direction of the ship semantic segmentation area are determined, the second attitude direction of the target ship is determined based on the long axis direction, the short axis direction and the first attitude direction, and the second attitude direction is used as an attitude detection result of the target ship; the method is simple in algorithm and easy to realize, and can be used for effectively realizing quick and accurate detection of the marine ship attitude.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a ship attitude detection method provided by the invention;
FIG. 2 is a schematic view of the distribution of pixel features in space under different illumination and photographing conditions provided by the present invention;
FIG. 3 is a flow chart of the adaptive semantic segmentation method provided by the invention;
FIG. 4 is a schematic structural view of the ship attitude detection device provided by the invention;
fig. 5 is a schematic diagram of the physical structure of the electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, in the description of the present invention, the terms "first," "second," and the like are used for distinguishing between similar objects and not for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged, as appropriate, such that embodiments of the present invention may be implemented in sequences other than those illustrated or described herein, and that the objects identified by "first," "second," etc. are generally of a type, and are not limited to the number of objects, such as the first object may be one or more.
The ship attitude detection method, the ship attitude detection device, the electronic equipment and the storage medium provided by the invention are exemplarily introduced below with reference to the accompanying drawings.
Fig. 1 is a schematic flow chart of a ship attitude detection method provided by the invention, and as shown in fig. 1, the method comprises the following steps:
step 100, acquiring a ship semantic segmentation area corresponding to a target ship and a wake semantic segmentation area corresponding to the wake of the target ship based on a global remote sensing image containing the target ship;
step 110, determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
step 120, determining a first gesture direction of the target ship based on the first center and the second center;
and 130, determining a long axis direction and a short axis direction of the ship semantic division area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
It should be noted that, the execution body of the ship gesture detection method provided by the embodiment of the invention may be an electronic device, a component, an integrated circuit or a chip in the electronic device. The electronic device may be a mobile electronic device or a non-mobile electronic device. Illustratively, the mobile electronic device may be a mobile phone, a tablet computer, a notebook computer, a palm computer, a wearable device, an Ultra mobile personal computer (Ultra-mobile Personal Computer, UMPC), a netbook or a personal digital assistant (Personal Digital Assistant, PDA), etc., and the non-mobile electronic device may be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (Personal Computer, PC), a Television (Television, TV), a teller machine or a self-service machine, etc., which is not particularly limited by the embodiments of the present invention.
The technical scheme of the embodiment of the invention is described in detail below by taking a computer to execute the ship attitude detection method provided by the invention as an example.
Specifically, in order to overcome the defect that the prior art only faces the position detection task of the ship and does not pay attention to the detection of the ship gesture, the invention obtains a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship based on a global remote sensing image containing the target ship, then determines the first gesture direction of the target ship based on the first center of the ship semantic segmentation area and the second center of the wake semantic segmentation area, further determines the long axis direction and the short axis direction of the ship semantic segmentation area, determines the second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and takes the second gesture direction as the gesture detection result of the target ship; the method is simple in algorithm and easy to realize, and can be used for effectively realizing quick and accurate detection of the marine ship attitude.
Alternatively, a global remote sensing image containing the target ship can be acquired first, and then a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship can be obtained based on the acquired global remote sensing image.
Optionally, after the ship semantic division region and the wake semantic division region are obtained, a first center of the ship semantic division region and a second center of the wake semantic division region may be determined, and the first gesture direction of the target ship may be roughly determined based on the first center of the ship semantic division region and the second center of the wake semantic division region.
Alternatively, after the ship semantic division region is obtained, the long axis direction and the short axis direction of the ship semantic division region may be determined.
Optionally, after the long axis direction and the short axis direction of the ship semantic division area and the first gesture direction of the target ship are obtained, a second gesture direction of the target ship can be determined based on the long axis direction and the short axis direction of the ship semantic division area and the first gesture direction of the target ship, and the second gesture direction is used as a gesture detection result of the target ship.
According to the ship attitude detection method, the ship semantic segmentation area corresponding to the target ship and the wake semantic segmentation area corresponding to the wake of the target ship are obtained based on the global remote sensing image containing the target ship, then the first attitude direction of the target ship is determined based on the first center of the ship semantic segmentation area and the second center of the wake semantic segmentation area, the long axis direction and the short axis direction of the ship semantic segmentation area are further determined, the second attitude direction of the target ship is determined based on the long axis direction, the short axis direction and the first attitude direction, and the second attitude direction is used as an attitude detection result of the target ship; the method is simple in algorithm and easy to realize, and can be used for effectively realizing quick and accurate detection of the marine ship attitude.
Optionally, the obtaining, based on the global remote sensing image including the target ship, a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship includes:
extracting a ship area image where the target ship is located and a wake area image where the wake of the target ship is located from the global remote sensing image containing the target ship based on a YOLO (You Only Look Once) detection network;
and respectively carrying out semantic segmentation processing on the ship area image and the wake area image to obtain the ship semantic segmentation area and the wake semantic segmentation area.
Specifically, in the embodiment of the invention, in order to obtain a ship semantic segmentation area corresponding to a target ship and a wake semantic segmentation area corresponding to a wake of the target ship based on a global remote sensing image containing the target ship, a ship area image in which the target ship is located and a wake area image in which the wake of the target ship is located can be extracted from the global remote sensing image containing the target ship based on a YOLO detection network, and then semantic segmentation processing is performed on the extracted ship area image and wake area image respectively to obtain the ship semantic segmentation area and the wake semantic segmentation area.
It should be noted that, the threshold segmentation method in the conventional image segmentation technology is to define the category of the pixel point based on the pixel point and the neighborhood feature thereof, so that when the distribution of the feature of the pixel point in the region in space is more concentrated and the distance between the feature of the pixel point in the region and the feature of other regions is further, the region is easier to be extracted accurately, i.e. the robust image region extraction should have the following characteristics: the pixels in the region have a more concentrated and isolated distribution in the feature space.
When the feature space of the image is selected as the gray space, the following area bandwidth Ar is defined bw : the interval width of the middle 90% gray value in the region accounts for the interval width ratio of the middle 90% gray value in the complete image, as shown in formula (1):
wherein Ar is 0.05 (i),Ar 0.95 (i) The gray scale quantile of 0.05 and the gray scale quantile of 0.95 of the region i are respectively; ar (Ar) 0.05 ,Ar 0.95 The 0.05 gray scale quantile and the 0.95 gray scale quantile of the whole image area are respectively defined, and the definition of p quantiles in statistics is used here.
In statistics, kurtosis (Kurtosis) statistics are generally used to describe the number of Kurtosis features distributed at the mean, and an unbiased sample Kurtosis calculation method shown in formulas (2) and (3) is used:
Wherein Kurt 0 Represents the kurtosis of unbiased samples, n is the number of samples,the sample values and their average values, respectively.
Under the definition of the formulas (2) and (3), the lower kurtosis limit is 1, the kurtosis of the standard normal distribution is 3, the kurtosis of the uniform distribution is 1.8, and under the same standard deviation, the larger kurtosis represents the sharper and steeper shape of the sample distribution. Therefore, a robust image region should have a smaller region bandwidth Ar bw Smaller sample standard deviation sigma and larger sample Kurt kurtosis 0 。
When the original feature space is selected as the high-dimensional space, the correlation index in the dimension-reduction feature space and the concentration parameter kappa in the fitted von Mises-fern distribution (von Mises-Fisher distribution) are adopted for evaluation, and the probability distribution function is shown as a formula (4):
f p (x|μ,κ)＝C p (κ)exp(κμ T x) (4)
wherein C is p (κ) is a normalized constant generated by a modified Bessel function, x represents a vector of pixel point eigenvalues within the image, and μ represents a mean vector of pixels within the eigenvalue space.
Note that YOLO networks can be used for real-time target detection and have a lower false-positive (false-positive) detection rate in the background environment. Original YOLO networks are not suitable for detecting small target objects, YOLOv3 and YOLOv4 are designed successively to improve target detection performance, and YOLOv3 and YOLOv4 obviously improve detection performance and can better realize small target detection in images without sacrificing calculation speed.
Optionally, in the embodiment of the present invention, a YOLOv4 detection network is used to respectively implement rapid extraction of a ship area image where the target ship is located and a wake area image where the wake of the target ship is located from a global remote sensing image including the target ship.
It should be noted that, because the ship area image and the wake area image extracted based on the YOLO detection network further include more noise points (for example, including a background), the embodiment of the invention performs semantic segmentation processing on the extracted ship area image and wake area image respectively to obtain a ship semantic segmentation area and a wake semantic segmentation area, and then performs detection on the target ship gesture based on the ship semantic segmentation area and the wake semantic segmentation area, thereby being beneficial to improving the detection precision of the target ship gesture.
Optionally, the performing semantic segmentation processing on the ship area image and the wake area image to obtain the ship semantic segmentation area and the wake semantic segmentation area respectively includes:
inputting the ship area image and the global remote sensing image into a pre-constructed support vector machine (Support Vector Machine, SVM) semantic classifier, obtaining a first classification result output by the SVM semantic classifier, and determining the ship semantic segmentation area based on the first classification result;
Inputting the wake area image and the global remote sensing image into the SVM semantic classifier to obtain a second classification result output by the SVM semantic classifier, and determining the wake semantic segmentation area based on the second classification result.
Specifically, in the embodiment of the invention, in order to realize the semantic segmentation processing on the ship area image and the wake area image respectively to obtain a ship semantic segmentation area and a wake semantic segmentation area, an SVM semantic classifier can be pre-constructed, further the ship area image and the global remote sensing image are input into the pre-constructed SVM semantic classifier to obtain a first classification result output by the SVM semantic classifier, and the ship semantic segmentation area is determined based on the first classification result; similarly, inputting the wake semantic segmentation region and the global remote sensing image into a pre-constructed SVM semantic classifier, obtaining a second classification result output by the SVM semantic classifier, and determining the wake semantic segmentation region based on the second classification result.
In practical applications, the extraction accuracy of the image segmentation method based on the fixed threshold is lowered due to the influence of illumination and photographing conditions. The difference between the ship area and the neighborhood background area in the ship area image is not obvious enough and has interference of cloud layers and the like, so that the adaptation to different shooting conditions under a single threshold value cannot be realized by the methods of automatic brightness adjustment, white balance, image brightness calibration and the like.
Therefore, the invention provides a self-adaptive primitive semantic segmentation method to improve the adaptability of the ship attitude detection method to shooting environments.
The basic idea of the self-adaptive primitive semantic segmentation method provided by the invention is as follows: and (3) taking regional pixel point characteristics (ship regional image and wake regional image) and global image characteristics (global remote sensing image) as inputs of an algorithm, constructing an adaptive classifier, and realizing identification of primitive regions in the image.
It should be noted that, in the embodiment of the present invention, the primitive area includes the characteristics of the ship area and the wake area.
Alternatively, the RGB image of the original ship area and the RGB image of the wake area may be first converted into the Lab color space, and the Lab values of the pixels are used as basic feature inputs, so as to implement the adjustment and balance of the tone scale and the brightness, respectively.
For example, fig. 2 is a schematic diagram of the distribution of pixel characteristics in space under different illumination and photographing conditions provided by the present invention, as shown in fig. 2. For the classification problem shown in fig. 2, the following confusion duty cycle function is defined as shown in equation (5):
s.t.y i ＝1，y k ＝-1，||x i -x k ||≤ε (5)
wherein { y } i ＝1，y k = -1} is the label set of the region class to which the pixel belongs, ε is the small relaxation threshold, x i Representing the characteristics of pixel points belonging to the ith class, x k Features belonging to the kth class of pixels are represented.
In the classification problem shown in FIG. 2, K of each class miss Greater than 20% and wake area and background areaK between domains miss More than 80%, characterizing that for identification problems between wake and background areas, more than 80% of the data features in one pattern class are identical to the other pattern class, so global pattern identification is an unidentifiable problem; similarly, global identification will produce at least 20% aliasing errors for ship regions and wake regions. This illustrates the necessity of performing algorithm adaptation and using global features.
It should be noted that, as known from the regional feature distribution in the specific environment, the primitive extraction problem in the specific environment is a learning problem, and can be converted into a support vector machine (support vector machine, SVM) form as shown in the formula (6):
s.t.y i [w T f(x i )+b]≥1-ξ i ，ξ i ≥0 (6)
wherein w is T f(x i ) +b is the optimal interface judgment function; f (x) i ) As a kernel function, here chosen as a linear kernel for facilitating subsequent adaptive regression; c epsilon R + Is a penalty factor; zeta type toy i To classify erroneous samples.
The class judgment function of the recognition primitive is as shown in equations (7) and (8):
wherein w is T f(x i )+b 11 ，w T f(x i )+b 21 Interfaces formed by support vectors belonging to the class respectively; w (w) T f(xi)+b 12 ，w T f(x i )+b 22 Respectively the partitions formed by boundary points belonging to the classAn interface; the solving process is shown in the formulas (9) and (10):
therefore, the self-adaptive semantic segmentation problem of the primitive can be constructed as a regression problem of primitive interface functional parameters under different shooting environments. The adaptation process being by global information I of the image g The following 5 adaptive functions were obtained:
w＝g 1 (I g )，b 11 ＝g 2 (I g )，b 12 ＝g 3 (I g )，b 21 ＝g 4 (I g )，b 22 ＝g 5 (I g )；
the learning process of the adaptive function comprises the following steps (1) to (5):
(1) Performing manual threshold division on the image recognition primitives under different shooting conditions to generate a training set for training;
(2) SVM learning is carried out on each image based on the training set, and corresponding { w } T }；
(3) By characteristic mean of images(e.g. luminance average) as input, for { w T Linear regression to obtain adaptive projection function g 1 ；
(4) According to g 1 Generating boundary functions { b } 11 ，b 12 ，b 21 ，b 22 }；
(5) By characteristic mean valueFor { b 11 ，b 12 ，b 21 ，b 22 Respectively carry out linear backObtaining adaptive boundary function g 2 ，g 3 ，g 4 ，g 5 。
It can be understood that by constructing the adaptive SVM semantic classifier and using the ship region image, the global remote sensing image, the wake region image and the global remote sensing image as the input of the SVM semantic classifier for semantic segmentation, the ship gesture detection method provided by the invention is not influenced by shooting environment or conditions, and has stronger adaptability.
Optionally, before inputting the ship area image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the trail area image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further comprises:
based on Min-Max standardization thought, the ship area image and the wake area image are respectively subjected to linear calibration pretreatment.
Specifically, in the embodiment of the invention, in order to improve the stability of an algorithm and ensure that images are uniformly transformed at bright and dark boundaries, before semantic segmentation processing is respectively carried out on a ship region image and a wake region image, namely before the ship region image and the wake region image are respectively input into an SVM semantic classifier, linear calibration preprocessing can be respectively carried out on the ship region image and the wake region image based on Min-Max standardization thought.
Optionally, based on the Min-Max standardization concept, the linear calibration preprocessing is performed on the ship area image and the wake area image respectively, and the method comprises the following steps:
and respectively carrying out linear calibration pretreatment on the ship area image and the wake area image based on the following formula:
Where x represents a feature component of the target image,representing the characteristic component, x, of the target image after calibration 0.05 0.05 quantile, x representing a feature component of the target image 0.95 0.95 quantiles, x representing characteristic components of the target image 0.05 (ref) 0.05 minutes, x representing the characteristic component of the reference image 0.95 (ref) represents a 0.95 quantile of a feature component of the reference image, the target image being the ship region image or the wake region image.
Note that x=x in formula (11) i (k) K=1, 2,3, k=1, 2,3 denote R, G, B three channels of the image.
Optionally, before inputting the ship area image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the trail area image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further comprises:
initializing a target feature recognition threshold;
taking the target feature recognition threshold value as a bias parameter in an optimal interface judgment function of the SVM semantic classifier, and optimizing the target feature recognition threshold value based on a heuristic search algorithm to obtain the optimized target feature recognition threshold value;
Substituting the optimized target feature recognition threshold value into an optimal interface judgment function of the SVM semantic classifier to obtain the optimized optimal interface judgment function;
and constructing the SVM semantic classifier based on the optimized optimal interface judgment function.
Specifically, in the embodiment of the invention, before the SVM semantic classifier is utilized to carry out semantic segmentation processing on the ship area image and the wake area image, the SVM semantic classifier needs to be built, wherein the method for building the SVM semantic classifier comprises the following steps: firstly initializing a target feature recognition threshold, then taking the target feature recognition threshold as a bias parameter in an optimal interface judgment function of the SVM semantic classifier, optimizing the target feature recognition threshold based on a heuristic search algorithm to obtain an optimized target feature recognition threshold, substituting the optimized target feature recognition threshold into the optimal interface judgment function of the SVM semantic classifier to obtain the optimized optimal interface judgment function, and finally constructing the SVM semantic classifier based on the optimized optimal interface judgment function.
It should be noted that, although the adaptive SVM semantic classifier constructed as described above can adapt to most of shooting environments, the image feature mean value is as follows The global characterization of the image is imperfect and the errors of the linear regression make it possible in some extreme cases to not extract the correct recognition primitive. For this reason, the embodiment of the present invention proposes that the following heuristic search algorithm implements adaptation to such a situation, including the following steps (1) to (4):
(1) Initializing recognition threshold (target feature recognition threshold) b=b of primitive 0 ；
(2) Judging function w according to optimal interface of SVM semantic classifier T f(x i ) The output of +b divides the identification area, and carries on morphological calculation to the identification area;
(3) Updating the recognition threshold of the primitive by the morphological calculation result;
(4) Repeating operation with the updated recognition threshold until the morphological calculation result meets the preset primitive recognition requirement.
Alternatively, the initial search value of the recognition threshold in the heuristic search algorithm may be calculated by using a fuzzy algorithm based on membership functions as shown in equation (12):
wherein θ εR m For a fuzzy set of reference thresholds in different shooting environments, m is the number of reference sample images used,for the corresponding membership function vector, the membership function takes the form of gaussian, exponential functions as shown in equations (13), (14) and (15):
Wherein a is 1 ,b 1 ,a m ,b m ,a k ,b k Is the shape parameter of the corresponding membership function.
The updating process of the recognition threshold is as shown in formulas (16) and (17):
b′＝b-kΔb (17)
b, b' are respectively identification thresholds before and after updating; a is that 1 ,A 0 The area of the region extracted during morphological analysis and the area of the reference primitive region are respectively; sgn () is a sign function, the selection of the initial sign is determined according to the change relation between the recognition threshold and the area of the area under the criterion, and the principle is that positive delta b will generate positive area change delta A; a epsilon R + Is a constant coefficient; k is the basic search step where k is 0.1.
It can be appreciated that the classification accuracy of the SVM semantic classifier can be improved by optimizing the target feature recognition threshold based on the heuristic search algorithm.
Optionally, fig. 3 is a schematic flow chart of the adaptive semantic segmentation method provided by the invention, as shown in fig. 3, calibration pretreatment is performed on a ship area image, a wake area image and a global remote sensing image, then adaptive primitive recognition is performed based on an SVM classifier, further morphological calculation is performed on the recognized primitives, whether a morphological calculation result meets primitive criteria (preset primitive recognition requirement) is judged, if not, a heuristic search algorithm is used for updating a threshold value and re-performing morphological calculation until the primitive criteria are met, and finally required primitive areas (ship semantic segmentation area and wake semantic segmentation area) can be obtained.
Optionally, the determining the major axis direction and the minor axis direction of the ship semantic division area includes:
the long axis direction and the short axis direction of the ship semantic segmentation region are determined based on principal component analysis (Principal Component Analysis, PCA) and singular value decomposition (Singular Value Decomposition, SVD) algorithms.
Specifically, in the embodiment of the invention, the long axis direction and the short axis direction of the ship semantic segmentation area can be determined based on a principal component analysis PCA and a singular value decomposition SVD algorithm.
It should be noted that, because the target ship has a large aspect ratio, the gesture description of the target ship can be performed through the long axis and the short axis orthogonal to the ship semantic division area, so the problem of ship gesture detection is equivalent to projecting the ship semantic division area data, and searching and retaining the projection directions of the maximum and minimum information amounts. In the embodiment of the invention, the singular value decomposition algorithm is applied to effectively solve by a principal component analysis method, as shown in formulas (18) and (19):
/>
wherein S is cov To analyze dataA difference matrix; lambda (lambda) i ,v i Is S cov Characteristic values arranged from large to small and corresponding characteristic vectors; d is the original dimension of each datum; k is the dimension of the reduced data. For the present problem, k=d=2, v at this time 1 ,v 2 Respectively correspond to the long axis direction and the short axis direction of the ship semantic division area.
The sole bow orientation is determined by the ship's wake, defined as wake area C 0 Directional ship detection area C 1 Ship attitude direction defined by recognition elementThe calculation process is shown in formulas (20) and (21):
optionally, the calculation flow of the robust ship attitude detection algorithm may include the following steps (1) to (4):
(1) Obtaining a ship area image Img 0 And wake area image Img 1 ；
(2) For ship area image Img 0 And wake area image Img 1 Global binarization processing is respectively carried out, and a ship semantic segmentation area Ar is obtained according to morphological calculation results (the structural size of a target ship and the analysis result of a connected domain) 0 And wake semantic segmentation region Ar 1 ；
(3) From Ar 0 And Ar is a group 1 Calculating the wake and the area center C of the ship 0 ,C 1 And based on C 0 ,C 1 Calculating rough attitude direction of ship
(4) For Ar 1 Performing PCA operation, obtaining the long axis direction v 1 Short axis direction v 2 Based on PCA resultsCalculating ship attitude direction +.>
Optionally, also according to the image resolution I coff And calculating the ship position and combining the ship attitude detection algorithm to obtain the complete ship attitude state.
Analyzing the error of the ship position, wherein the generated position error delta T is shown as a formula (22):
ΔT 1 ＝I coff ΔC 1 (22)
Wherein DeltaC 1 Is the detection error of the center of the ship area.
The covariance matrix of the image region at this time is shown in formula (23):
wherein,/>
the errors mainly originate from the extraction of the ship area, and have smaller variance and uncertainty, so that the errors have smaller attitude detection errors and uncertainty.
It can be appreciated that the embodiment of the invention provides a ship attitude detection method, which comprises the steps of firstly preprocessing an original image based on defined image area robustness; then, respectively extracting a ship area and a wake area by using a YOLO depth detection network; and finally, calculating the ship pose by adopting a characteristic analysis method. The algorithm realizes the rapid and accurate detection of the target pose of the ship in the sea area through the extraction of the ship identification primitives, the self-adaptive semantic segmentation of the primitives and the detection of the ship pose based on the regional analysis; by combining the advantages of deep learning and feature analysis, the problem of detecting the attitude of the sea area ship target in the wide remote sensing image can be effectively solved.
It should be noted that, in the related art, the (Convolutional Neural Network, CNN) method is generally used to directly perform the target state estimation, so as to implement the target detection and estimation that is completely driven by data. Some of these studies focus on the state estimation of objects from three-dimensional (3D) point cloud data, and other researchers consider using multi-view images of objects to achieve classification and state estimation of objects. These methods are generally directed to the general objective estimation problem that does not require consideration of algorithm accuracy and generally require a large amount of training data; the embodiment of the invention aims to establish a set of marine ship pose detection method based on remote sensing images, and combines the region extraction capability of deep learning and the robustness of defining characteristics to realize accurate, rapid and reliable detection of the ship target pose.
According to the ship attitude detection method, the ship semantic segmentation area corresponding to the target ship and the wake semantic segmentation area corresponding to the wake of the target ship are obtained based on the global remote sensing image containing the target ship, then the first attitude direction of the target ship is determined based on the first center of the ship semantic segmentation area and the second center of the wake semantic segmentation area, the long axis direction and the short axis direction of the ship semantic segmentation area are further determined, the second attitude direction of the target ship is determined based on the long axis direction, the short axis direction and the first attitude direction, and the second attitude direction is used as an attitude detection result of the target ship; the method is simple in algorithm and easy to realize, and can be used for effectively realizing quick and accurate detection of the marine ship attitude.
The ship attitude detection device provided by the invention is described below, and the ship attitude detection device described below and the ship attitude detection method described above can be correspondingly referred to each other.
Fig. 4 is a schematic structural diagram of a ship attitude detection device provided by the present invention, as shown in fig. 4, the device includes: the obtaining module 410, the first determining module 420, the second determining module 430, and the third determining module 440; wherein:
The obtaining module 410 is configured to obtain a ship semantic segmentation area corresponding to a target ship and a wake semantic segmentation area corresponding to a wake of the target ship based on a global remote sensing image including the target ship;
the first determining module 420 is configured to determine a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
the second determining module 430 is configured to determine a first attitude direction of the target ship based on the first center and the second center;
the third determining module 440 is configured to determine a major axis direction and a minor axis direction of the ship semantic division area, determine a second gesture direction of the target ship based on the major axis direction, the minor axis direction and the first gesture direction, and use the second gesture direction as a gesture detection result of the target ship.
According to the ship attitude detection device, the ship semantic segmentation area corresponding to the target ship and the wake semantic segmentation area corresponding to the wake of the target ship are obtained based on the global remote sensing image containing the target ship, then the first attitude direction of the target ship is determined based on the first center of the ship semantic segmentation area and the second center of the wake semantic segmentation area, the long axis direction and the short axis direction of the ship semantic segmentation area are further determined, the second attitude direction of the target ship is determined based on the long axis direction, the short axis direction and the first attitude direction, and the second attitude direction is used as an attitude detection result of the target ship; the method is simple in algorithm and easy to realize, and can be used for effectively realizing quick and accurate detection of the marine ship attitude.
It should be noted that, the ship gesture detection device provided by the embodiment of the present invention can implement all the method steps implemented by the ship gesture detection method embodiment, and can achieve the same technical effects, and specific details of the same parts and beneficial effects as those of the method embodiment in the embodiment are not repeated here.
Fig. 5 is a schematic physical structure of an electronic device according to the present invention, as shown in fig. 5, the electronic device may include: processor 510, communication interface (Communications Interface) 520, memory 530, and communication bus 540, wherein processor 510, communication interface 520, memory 530 complete communication with each other through communication bus 540. Processor 510 may invoke logic instructions in memory 530 to perform the ship attitude detection method provided by the methods described above, the method comprising:
based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship;
determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
Determining a first attitude direction of the target ship based on the first center and the second center;
determining a long axis direction and a short axis direction of the ship semantic segmentation area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
Further, the logic instructions in the memory 530 described above may be implemented in the form of software functional units and may be stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform the method of vessel attitude detection provided by the methods described above, the method comprising:
based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship;
determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
determining a first attitude direction of the target ship based on the first center and the second center;
determining a long axis direction and a short axis direction of the ship semantic segmentation area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform the above-provided ship attitude detection methods, the method comprising:
Based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship;
determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
determining a first attitude direction of the target ship based on the first center and the second center;
determining a long axis direction and a short axis direction of the ship semantic segmentation area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The ship attitude detection method is characterized by comprising the following steps of:
based on a global remote sensing image containing a target ship, acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship;
determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
determining a first attitude direction of the target ship based on the first center and the second center;
determining a long axis direction and a short axis direction of the ship semantic segmentation area, determining a second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
2. The ship attitude detection method according to claim 1, wherein the obtaining, based on a global remote sensing image including a target ship, a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to a wake of the target ship includes:
based on a YOLO detection network, extracting a ship region image in which a target ship is located and a wake region image in which a wake of the target ship is located from the global remote sensing image containing the target ship;
And respectively carrying out semantic segmentation processing on the ship area image and the wake area image to obtain the ship semantic segmentation area and the wake semantic segmentation area.
3. The ship attitude detection method according to claim 2, wherein the performing semantic segmentation processing on the ship region image and the wake region image, respectively, to obtain the ship semantic segmentation region and the wake semantic segmentation region, comprises:
inputting the ship area image and the global remote sensing image into a pre-constructed Support Vector Machine (SVM) semantic classifier, obtaining a first classification result output by the SVM semantic classifier, and determining the ship semantic segmentation area based on the first classification result;
inputting the wake area image and the global remote sensing image into the SVM semantic classifier to obtain a second classification result output by the SVM semantic classifier, and determining the wake semantic segmentation area based on the second classification result.
4. The ship attitude detection method according to claim 3, characterized in that before inputting the ship region image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the trail region image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further comprises:
Initializing a target feature recognition threshold;
taking the target feature recognition threshold value as a bias parameter in an optimal interface judgment function of the SVM semantic classifier, and optimizing the target feature recognition threshold value based on a heuristic search algorithm to obtain the optimized target feature recognition threshold value;
substituting the optimized target feature recognition threshold value into an optimal interface judgment function of the SVM semantic classifier to obtain the optimized optimal interface judgment function;
and constructing the SVM semantic classifier based on the optimized optimal interface judgment function.
5. The ship attitude detection method according to claim 3, characterized in that before inputting the ship region image and the global remote sensing image into a pre-constructed support vector machine SVM semantic classifier, and inputting the trail region image and the global remote sensing image into the support vector machine SVM semantic classifier, the method further comprises:
based on Min-Max standardization thought, the ship area image and the wake area image are respectively subjected to linear calibration pretreatment.
6. The ship attitude detection method according to claim 5, wherein the linear calibration preprocessing is performed on the ship area image and the wake area image based on Min-Max standardization thought, respectively, comprising:
And respectively carrying out linear calibration pretreatment on the ship area image and the wake area image based on the following formula:
where x represents a feature component of the target image,representing the characteristic component, x, of the target image after calibration 0.05 0.05 quantile, x representing a feature component of the target image 0.95 0.95 quantiles, x representing characteristic components of the target image 0.05 (ref) 0.05 minutes, x representing the characteristic component of the reference image 0.95 (ref) represents a 0.95 quantile of a feature component of the reference image, the target image being the ship region image or the wake region image.
7. The ship attitude detection method according to claim 1, wherein the determining the major axis direction and the minor axis direction of the ship semantic division area includes:
and determining the major axis direction and the minor axis direction of the ship semantic segmentation area based on principal component analysis PCA and singular value decomposition SVD algorithm.
8. The utility model provides a naval vessel gesture detection device which characterized in that includes:
the acquisition module is used for acquiring a ship semantic segmentation area corresponding to the target ship and a wake semantic segmentation area corresponding to the wake of the target ship based on a global remote sensing image containing the target ship;
The first determining module is used for determining a first center of the ship semantic segmentation area and a second center of the wake semantic segmentation area;
the second determining module is used for determining a first gesture direction of the target ship based on the first center and the second center;
and the third determining module is used for determining the long axis direction and the short axis direction of the ship semantic segmentation area, determining the second gesture direction of the target ship based on the long axis direction, the short axis direction and the first gesture direction, and taking the second gesture direction as a gesture detection result of the target ship.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the ship attitude detection method according to any one of claims 1 to 7 when executing the program.
10. A non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor implements the ship attitude detection method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310839541.2A CN117115436A (en) | 2023-07-10 | 2023-07-10 | Ship attitude detection method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310839541.2A CN117115436A (en) | 2023-07-10 | 2023-07-10 | Ship attitude detection method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117115436A true CN117115436A (en) | 2023-11-24 |
Family
ID=88799100
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310839541.2A Pending CN117115436A (en) | 2023-07-10 | 2023-07-10 | Ship attitude detection method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117115436A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117788808A (en) * | 2024-02-28 | 2024-03-29 | 南京航空航天大学 | Positioning detection method for separating wake of weak and small target mobile ship from ship |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020975A (en) * | 2012-12-29 | 2013-04-03 | 北方工业大学 | Wharf and ship segmentation method combining multi-source remote sensing image characteristics |
| US20160305782A1 (en) * | 2015-04-14 | 2016-10-20 | Invensense Incorporated | System and method for estimating heading misalignment |
-
2023
- 2023-07-10 CN CN202310839541.2A patent/CN117115436A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020975A (en) * | 2012-12-29 | 2013-04-03 | 北方工业大学 | Wharf and ship segmentation method combining multi-source remote sensing image characteristics |
| US20160305782A1 (en) * | 2015-04-14 | 2016-10-20 | Invensense Incorporated | System and method for estimating heading misalignment |
Non-Patent Citations (3)
| Title |
|---|
| YEJIAN ZHOU 等: ""Attitude Estimation and Geometry Reconstruction of Satellite Targets Based on ISAR Image Sequence Interpretation"", 《IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS 》, vol. 55, no. 4, 11 October 2018 (2018-10-11), pages 1698 - 1711, XP011738834, DOI: 10.1109/TAES.2018.2875503 * |
| 乔腾飞: ""遥感卫星在轨船舰目标检测关键技术研究"", 乔腾飞, 15 February 2023 (2023-02-15), pages 1 - 75 * |
| 舒服: ""高分辨率光学遥感影像舰船尾迹检测及运动参数估计方法研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》, 15 January 2023 (2023-01-15), pages 1 - 113 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117788808A (en) * | 2024-02-28 | 2024-03-29 | 南京航空航天大学 | Positioning detection method for separating wake of weak and small target mobile ship from ship |
| CN117788808B (en) * | 2024-02-28 | 2024-05-03 | 南京航空航天大学 | Positioning detection method for separating wake of weak and small target mobile ship from ship |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | New hierarchical saliency filtering for fast ship detection in high-resolution SAR images | |
| Guo et al. | Scattering enhanced attention pyramid network for aircraft detection in SAR images | |
| CN108765458B (en) | Sea surface target scale self-adaptive tracking method of high-sea-condition unmanned ship based on correlation filtering | |
| CN108830879A (en) | A kind of unmanned boat sea correlation filtering method for tracking target suitable for blocking scene | |
| CN109101897A (en) | Object detection method, system and the relevant device of underwater robot | |
| Feng et al. | Multiphase SAR image segmentation with $ G^{0} $-statistical-model-based active contours | |
| CN109816051B (en) | Hazardous chemical cargo feature point matching method and system | |
| CN110334703B (en) | Ship detection and identification method in day and night image | |
| Yu et al. | Object detection-tracking algorithm for unmanned surface vehicles based on a radar-photoelectric system | |
| CN110633727A (en) | Deep neural network ship target fine-grained identification method based on selective search | |
| CN113705375A (en) | Visual perception device and method for ship navigation environment | |
| Tueller et al. | Target detection using features for sonar images | |
| CN117115436A (en) | Ship attitude detection method and device, electronic equipment and storage medium | |
| Wang et al. | License plate recognition system | |
| CN114821358A (en) | Optical remote sensing image marine ship target extraction and identification method | |
| CN114764801A (en) | Weak and small ship target fusion detection method and device based on multi-vision significant features | |
| Li et al. | Outlier-robust superpixel-level CFAR detector with truncated clutter for single look complex SAR images | |
| Gebhardt et al. | Hunting for naval mines with deep neural networks | |
| CN110472607A (en) | A kind of ship tracking method and system | |
| CN113128518B (en) | Sift mismatch detection method based on twin convolution network and feature mixing | |
| CN109887004A (en) | A kind of unmanned boat sea area method for tracking target based on TLD algorithm | |
| Ye et al. | Mobilenetv3-yolov4-sonar: Object detection model based on lightweight network for forward-looking sonar image | |
| CN117689995A (en) | Unknown spacecraft level detection method based on monocular image | |
| CN117036740A (en) | Anti-occlusion tracking method for moving target | |
| CN111681266A (en) | Ship tracking method, system, equipment and storage medium |
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 |