CN115346133A - Ship detection method and system based on optical satellite image - Google Patents
Ship detection method and system based on optical satellite image Download PDFInfo
- Publication number
- CN115346133A CN115346133A CN202210976576.6A CN202210976576A CN115346133A CN 115346133 A CN115346133 A CN 115346133A CN 202210976576 A CN202210976576 A CN 202210976576A CN 115346133 A CN115346133 A CN 115346133A
- Authority
- CN
- China
- Prior art keywords
- image
- full
- color image
- satellite
- ship
- 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 109
- 230000003287 optical effect Effects 0.000 title claims abstract description 43
- 238000000034 method Methods 0.000 claims abstract description 25
- 238000007781 pre-processing Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims description 25
- 238000002310 reflectometry Methods 0.000 claims description 22
- 238000012937 correction Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 description 5
- 238000003384 imaging method Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000007500 overflow downdraw method Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000009545 invasion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
Images
Classifications
-
- 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
-
- 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
-
- 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
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
-
- 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
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a ship detection method and system based on optical satellite images, wherein the method comprises the following steps: acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image; respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image; fusing the preprocessed full-color image and the multispectral image to obtain a fused image; and detecting the fused image by using a ship target detection model to obtain a ship target. Compared with the prior art, the method has the advantages that the quality of the satellite image is improved by adopting preprocessing operation, the interference which influences the subsequent ship detection is removed, then the ship target is detected by adopting the ship target detection model, the missing detection and the false detection can be prevented, and the accurate ship target is obtained.
Description
Technical Field
The invention relates to the technical field of ship detection, in particular to a ship detection method and system based on optical satellite images.
Background
In recent years, various illegal fishing boats have the behavior of entering a controlled sea area without authorization, and serious threat is brought to marine traffic safety in China. The method can discover, predict or early warn the invasion behavior of the suspicious ship as early as possible, and carry out real-time remote monitoring and accurate positioning on the suspicious ship, thereby having great significance on the marine safety of China.
However, due to the wide ocean area, it is difficult to effectively monitor the surface vessels by ground radar or airborne radar. In the face of possible illegal fishing boats, the quality of marine supervision is directly influenced by reliable early warning and prediction and effective reaction. With the rapid development of remote sensing satellite technology in China, a plurality of optical/radar remote sensing satellites are transmitted in sequence in China. The satellite for shooting the sea area in China can complete multiple times of shooting every day, and has wide coverage and high resolution. The remote sensing satellite has the characteristics of wide observation range, short observation period, strong data timeliness, high spatial resolution and the like, and has unique superiority in the aspect of illegal fishing boat detection. By analyzing and excavating the remote sensing image, an intelligent classification and identification model of the illegal fishing boat is established, so that the illegal fishing boat on the sea surface can be automatically identified and tracked quickly, and an effective way for solving the safety of the sea area in China is provided.
In recent years, earth observation technology is rapidly developed, satellite remote sensing enters an unprecedented development stage, a batch of imaging satellites with high spatial resolution and short revisit period emerge, and an extremely rich data source is provided for sea area reconnaissance and ship target monitoring. At present, most of ship detection research based on remote sensing images is algorithms developed around SAR images, and the algorithms are mature. The infrared imaging system can penetrate smoke, can work day and night, can identify a hot target, is often applied to uncovering the position of a hidden target, enhances the visual performance of the target under the condition of weak light, and is widely applied to night reconnaissance and monitoring. Compared with other sensor images, the infrared image resolution is low, the aerospace infrared remote sensing application is few, and ship target detection of most infrared images is based on ship-borne or low-altitude imaging. Therefore, the ship target detection method based on the infrared image has low precision of ship target detection due to the low resolution of the acquired infrared image.
Disclosure of Invention
In order to solve the above problems, embodiments of the present invention provide a method and a system for detecting ships based on optical satellite images.
A ship detection method based on optical satellite images comprises the following steps:
step 1: acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
and 2, step: respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
and step 3: fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
and 4, step 4: and detecting the fused image by using a ship target detection model to obtain a ship target.
Preferably, the step 2: respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image, wherein the preprocessing comprises the following steps:
step 2.1: carrying out radiometric calibration on the full-color image to obtain a calibrated full-color image;
step 2.2: performing orthorectification on the calibrated full-color image to obtain a preprocessed full-color image;
step 2.3: and sequentially carrying out radiometric calibration, atmospheric correction and orthometric correction on the multispectral image to obtain a preprocessed multispectral image.
Preferably, the step 2.1: radiometric calibration of the full-color image to obtain a calibrated full-color image, comprising:
step 2.1.1: calculating the radiance value and the apparent reflectivity of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
step 2.1.2: and carrying out radiometric calibration on the full-color image by utilizing the radiance value and the apparent reflectivity to obtain a calibrated full-color image.
Preferably, the step 4: utilize the naval vessel target detection model to detect the image after fusing and obtain the naval vessel target, include:
step 4.1: classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
and 4.2: inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
step 4.3: detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
step 4.4: and combining redundant rotation detection frames by using a soft rotation NMS method to obtain a ship target.
The invention also provides a ship detection system based on the optical satellite image, which comprises:
the satellite image acquisition module is used for acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
the preprocessing module is used for respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
the fusion module is used for fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
and the target detection module is used for detecting the fused image by using the ship target detection model to obtain a ship target.
Preferably, the preprocessing module includes:
the radiometric calibration unit is used for radiometric calibration of the full-color image to obtain a calibrated full-color image;
the orthorectification unit is used for orthorectifying the calibrated full-color image to obtain a preprocessed full-color image;
and the comprehensive processing unit is used for sequentially carrying out radiometric calibration, atmospheric correction and orthometric correction on the multispectral image to obtain a preprocessed multispectral image.
Preferably, the radiation scaling unit includes:
a scaling parameter calculating subunit for calculating a radiance value and an apparent reflectance of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
and the scaling subunit is used for carrying out radiometric scaling on the full-color image by utilizing the radiance value and the apparent reflectivity so as to obtain a scaled full-color image.
Preferably, the object detection module includes:
the training sample acquisition unit is used for classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
the training unit is used for inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
the target detection unit is used for detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
and the detection frame merging unit is used for merging the redundant rotation detection frames by utilizing a soft rotation NMS method to obtain the ship target.
The invention also provides an electronic device, which comprises a bus, a transceiver, a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the transceiver, the memory and the processor are connected through the bus, and the computer program is executed by the processor to realize the steps in the ship detection method based on the optical satellite image.
The present invention also provides a computer readable storage medium having stored thereon a computer program which, when being executed by a processor, implements the steps of the above-mentioned optical satellite image-based ship detection method.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention relates to a ship detection method and system based on optical satellite images, wherein the method comprises the following steps: acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image; respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image; fusing the preprocessed full-color image and the multispectral image to obtain a fused image; and detecting the fused image by using a ship target detection model to obtain a ship target. Compared with the prior art, the method has the advantages that the quality of the satellite image is improved by adopting preprocessing operation, the interference which influences the subsequent ship detection is removed, then the ship target is detected by adopting the ship target detection model, the missing detection and the false detection can be prevented, and the accurate ship target is obtained.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a ship detection method based on optical satellite images in an embodiment of the present invention;
fig. 2 is a schematic diagram of a ship detection method based on an optical satellite image in an embodiment provided by the present invention;
fig. 3 is a schematic diagram of a ship feature extraction in an embodiment of the present invention.
Detailed Description
In the description of the present invention, it is to be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", and the like, indicate orientations and positional relationships based on those shown in the drawings, and are used only for convenience of description and simplicity of description, and do not indicate or imply that the device or element being referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be considered as limiting the present invention.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
In the present invention, unless otherwise explicitly stated or limited, the terms "mounted," "connected," "fixed," and the like are to be construed broadly and may, for example, be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood according to specific situations by those of ordinary skill in the art.
The embodiment of the invention aims to provide a ship detection method and system based on an optical satellite image, which are used for improving the identification precision of a ship target.
Referring to fig. 1-2, a ship detection method based on optical satellite images includes:
step 1: acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
in practical application, the optical remote sensing satellite forms image data by collecting spectral band information of ground objects, and the image data generally comprises a panchromatic image and a multispectral image, wherein the panchromatic image is a gray image and has higher spatial resolution than the multispectral image, and the multispectral image has a plurality of band information and can be synthesized into a true color or a pseudo color image. And preprocessing and fusing the panchromatic image and the multispectral image to obtain the multispectral image with high resolution.
The invention uses panchromatic image and multispectral image to replace the traditional infrared image and has the following advantages:
(1) the optical remote sensing image data volume is large. Under the same condition, the higher the image resolution is, the larger the data volume of the remote sensing image is, the number of satellites is continuously increased, the revisit period is also continuously shortened, and the data volume of the optical remote sensing image is increased more and more.
(2) The optical remote sensing image is interfered by factors such as noise, cloud and fog due to the difference of the acquisition time, the imaging angle, the sea state and the sensor parameter of the image, so that the characteristics of the ship target expressed in the optical image are different.
(3) The remote sensing image has rich detail information. Compared with an infrared image, in the high-resolution optical remote sensing image, the shape of the ship target is clear and visible, the characteristics of the shape, texture and the like of the ground object in the image are obvious, and the image is easy to analyze and process.
(4) Compared with the background, the ship target on the sea surface has rich detail information and higher brightness information, and is easy to identify. The ship targets are mostly artificial rigid bodies, ideal large ships generally have axisymmetric structures, the heads of the ships are similar to V-shaped, the types of the ships can be different according to the difference of the purposes of the ships, and the appearance structure is correspondingly changed.
And 2, step: respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
the method firstly adopts preprocessing operation to improve the quality of an original image, can remove image noise, blurring, bright spots, cloud and fog and other interferences which influence the subsequent ship detection, then adopts a target detection algorithm to detect the ship target, eliminates missed detection and false detection, and can obtain a real ship target.
Further, the step 2 comprises:
step 2.1: carrying out radiometric calibration on the full-color image to obtain a calibrated full-color image;
radiometric calibration is a process of converting the gray value of remote sensing satellite image data into physical quantities such as radiance value, apparent reflectivity and the like so as to correct errors generated by the sensor.
Specifically, the step 2.1 includes:
step 2.1.1: calculating the radiance value and the apparent reflectivity of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
step 2.1.2: and carrying out radiometric calibration on the full-color image by utilizing the radiance value and the apparent reflectivity to obtain a calibrated full-color image.
Step 2.2: performing orthorectification on the calibrated full-color image to obtain a preprocessed full-color image;
it should be noted that in the ortho-rectification process of the invention, the remote sensing satellite data can be ortho-rectified by using the remote sensing satellite data self-contained ortho-rectification parameter file and by using the RPC model, which is briefly described as follows:
the RPC model relates the ground point geodetic coordinates D (Latitude, longitude, height) and its corresponding image point coordinates D (line, sample) by a ratio polynomial. To enhance the stability of the parametric solution, the ground and image coordinates are normalized to between-1 and 1. For an image, the following ratio polynomial is defined:
wherein: (P, L, H) is normalized ground coordinates calculated from ground point geodetic coordinates D (Latitude, longitude, height) and corresponding normalized conversion parameters; the (X, Y) is a regularized image coordinate calculated from the line (sample) and a corresponding regularization parameter. Num L 、Den L 、Num S 、Den S Is a cubic polynomial system of (P, L, H)And (4) counting.
Step 2.3: and sequentially carrying out radiometric calibration, atmospheric correction and orthorectification on the multispectral image to obtain a preprocessed multispectral image.
And 3, step 3: fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
in practical application, image fusion mainly refers to fusion of a panchromatic image (black and white) and a multispectral image, the panchromatic image and the multispectral image in an original satellite image are pure image data, the resolution of the panchromatic image of the data of the same satellite is higher than that of the multispectral image, for example, the resolution of a high-resolution 2 panchromatic image is 1 meter, the multispectral image is 4 meters, and the fused image data can reach color image data with the resolution of 1 meter. The image fusion method can adopt an IHS transformation fusion method, a Brovey transformation fusion method, a PCA transformation-based fusion method and a Panschrepen algorithm to fuse the images.
And 4, step 4: and detecting the fused image by using a ship target detection model to obtain a ship target.
Further, the step 4 comprises:
step 4.1: classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
and 4.2: inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
step 4.3: detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
step 4.4: and combining redundant rotation detection frames by using a soft rotation NMS method to obtain a ship target.
In practical application, the method carries out deep analysis from the basic theory of target detection to the practical problems faced by ship detection, firstly extracts common characteristics of various ships (ship characteristic extraction is to classify the ships according to the texture characteristics of the ships in the remote sensing satellite image, different ships have different texture characteristics, as shown in figure 3), generates a preset classification data set, and then inputs the existing classification data set into a deep convolution network for model training to obtain a ship target detection model. The satellite remote sensing image has a large scene and a small ship target, and ships sail or berth in any direction and are not beneficial to positioning detection of the ships, so the ship detection deep learning model based on key points is adopted for detecting and identifying the ships in the optical satellite image.
The ship detection based on the key points uses an SKNet model, wherein SKNet is a 'one-stage' anchor-free framework, and each non-directional ship target is modeled as a central key point in the SKNet, and the shape and the size (including the width, the height and the rotation angle) of the central key point are combined into a detection result. The entire detection process of SKNet can be described as follows:
i is the input image, ξ represents the backbone network composed of stacked CNNs, op represents the proposed new module called the orthogonal pool,
and
representing predictions of central key point and morphology size, respectively, eta represents a combination of predictions, f SR-NMS Indicating the proposed soft rotation NMS (nonmaximum compression).
The detection process of the SKNet comprises two stages: feature extraction and keypoint-based prediction. In the feature extraction stage, multi-level and rich visual features are extracted and integrated by an orthogonal pool method. Then, the feature map is supervised to predict central key points and morphology sizes. By combining the above predictions, the soft rotation NMS produces the final detection result. In addition, the distortion of the ship may hinder the prediction of the angle due to the imaging angle and the imaging method. In order to avoid ambiguity brought by single-angle prediction, the invention integrates the semantic characteristics of the multidirectional rotating ship. In particular, the invention is applied to the long sides and short sides of shipsThe same angle prediction is performed on the edges and the two predicted angles are averaged as:
θ l denotes the direction of the long side, [ theta ] s Indicates the direction of the short side, theta pred Representing the predicted angle of rotation. Finally, a new NMS method, soft rotation NMS, is used to merge the redundant rotation detection boxes.
The soft rotation NMS method of the present invention is further described below with reference to specific embodiments:
assuming that there are multiple objects (including ships and other similar objects) in a slice of a satellite image, and each object is in a candidate frame, and there may be regions where these frames overlap with each other, what is needed is to keep the optimal candidate frame. Assuming that there are N candidate boxes, the score calculated by the classifier for each box is Si,1< = i < = N.
1) Building a set H for storing candidate frames to be processed, and initializing the set H to include all N frames; building a set M for storing the optimal frames, and initializing the set M into an empty set;
2) Sorting the frames in all the sets H, selecting the frame M with the highest score, and moving the frame M from the set H to the set M;
3) Traversing the frames in the set H, respectively calculating an intersection-over-intersection (IoU) with the frame m, if the calculated intersection-over-intersection is higher than a certain threshold (generally 0-0.5), considering that the frame is overlapped with the frame m, and removing the frame from the set H;
4) Go back to step 2 iteration until set H is empty. The frame in the set M is the required ship target.
It should be noted that the flow of the feature extraction stage of the present invention is:
(1) Inputting satellite images
(2) Normalizing and correcting images
(3) Divide the image into several blocks
(4) Dividing each block into a number of cells
(5) Combining the feature descriptors of all cells in each block
(6) Connecting the feature descriptors of all blocks to form the feature descriptor of the current detection window
(7) Collecting feature descriptors of all detection windows, and classifying the feature descriptors
(8) Features are extracted according to the classification.
A feature descriptor is a representation of a picture or a block of pictures that simplifies the image by extracting useful information and throwing away superfluous information.
The invention adopts the SKNet ship target detection model to detect the ship target, and can prevent missed detection and false detection.
The invention also provides a ship detection system based on the optical satellite image, which comprises:
the satellite image acquisition module is used for acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
the preprocessing module is used for respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
the fusion module is used for fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
and the target detection module is used for detecting the fused image by using the ship target detection model to obtain a ship target.
Preferably, the preprocessing module includes:
the radiometric calibration unit is used for radiometric calibration of the full-color image to obtain a calibrated full-color image;
the orthorectification unit is used for orthorectifying the calibrated full-color image to obtain a preprocessed full-color image;
and the comprehensive processing unit is used for sequentially carrying out radiometric calibration, atmospheric correction and orthometric correction on the multispectral image to obtain a preprocessed multispectral image.
Preferably, the radiation scaling unit includes:
a scaling parameter calculating subunit for calculating a radiance value and an apparent reflectance of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
and the scaling subunit is used for carrying out radiometric scaling on the full-color image by utilizing the radiance value and the apparent reflectivity to obtain a scaled full-color image.
Preferably, the object detection module includes:
the training sample acquisition unit is used for classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
the training unit is used for inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
the target detection unit is used for detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
and the detection frame merging unit is used for merging the redundant rotation detection frames by using a soft rotation NMS method to obtain the ship target.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the method, the satellite image quality is improved by adopting preprocessing operation, interference which influences subsequent ship detection is removed, and then the ship target is detected by adopting the ship target detection model, so that missing detection and false detection can be prevented, and an accurate ship target is obtained.
The invention also provides an electronic device, which comprises a bus, a transceiver, a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the transceiver, the memory and the processor are connected through the bus, and the computer program is executed by the processor to realize the steps in the ship detection method based on the optical satellite image.
The present invention also provides a computer readable storage medium having stored thereon a computer program which, when being executed by a processor, implements the steps of the above-mentioned optical satellite image-based ship detection method.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of the technical scope of the present invention, and the technical scope of the present invention is covered by the modifications or alternatives. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A ship detection method based on optical satellite images is characterized by comprising the following steps:
step 1: acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
step 2: respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
and 3, step 3: fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
and 4, step 4: and detecting the fused image by using a ship target detection model to obtain a ship target.
2. The optical satellite image-based ship detection method according to claim 1, wherein the step 2: respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image, wherein the preprocessing comprises the following steps:
step 2.1: carrying out radiometric calibration on the full-color image to obtain a calibrated full-color image;
step 2.2: performing orthorectification on the calibrated full-color image to obtain a preprocessed full-color image;
step 2.3: and sequentially carrying out radiometric calibration, atmospheric correction and orthorectification on the multispectral image to obtain a preprocessed multispectral image.
3. The optical satellite image-based ship detection method according to claim 2, wherein the step 2.1: radiometric calibration of the full-color image to obtain a calibrated full-color image, comprising:
step 2.1.1: calculating the radiance value and the apparent reflectivity of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
step 2.1.2: and carrying out radiometric calibration on the full-color image by utilizing the radiance value and the apparent reflectivity to obtain a calibrated full-color image.
4. The optical satellite image-based ship detection method according to claim 3, wherein the step 4: utilize the naval vessel target detection model to detect the image after fusing and obtain the naval vessel target, include:
step 4.1: classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
and 4.2: inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
step 4.3: detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
step 4.4: and combining the redundant rotation detection frames by using a soft rotation NMS method to obtain the ship target.
5. A naval vessel detecting system based on optical satellite image, characterized by includes:
the satellite image acquisition module is used for acquiring an optical remote sensing satellite acquisition satellite image; the satellite image comprises a full-color image and a multispectral image;
the preprocessing module is used for respectively preprocessing the full-color image and the multispectral image to obtain a preprocessed full-color image and a preprocessed multispectral image;
the fusion module is used for fusing the preprocessed full-color image and the multispectral image to obtain a fused image;
and the target detection module is used for detecting the fused image by using the ship target detection model to obtain a ship target.
6. The optical satellite image-based ship detection system according to claim 5, wherein the preprocessing module comprises:
the radiometric calibration unit is used for radiometric calibration of the full-color image to obtain a calibrated full-color image;
the orthorectification unit is used for orthorectifying the calibrated full-color image to obtain a preprocessed full-color image;
and the comprehensive processing unit is used for sequentially carrying out radiometric calibration, atmospheric correction and orthometric correction on the multispectral image to obtain a preprocessed multispectral image.
7. The optical satellite image-based ship detection system according to claim 6, wherein the radiometric calibration unit comprises:
a scaling parameter calculating subunit for calculating a radiance value and an apparent reflectance of the full-color image; wherein, the calculation formula of the radiance value and the apparent reflectivity of the full-color image is as follows:
wherein L is a radiance value, gain is a calibration slope, DN is a satellite load observation value, bias is a calibration intercept, rho is an apparent reflectivity, ESUN is solar spectrum radiant quantity, d is a sun-earth distance parameter, and theta is a sun zenith angle;
and the scaling subunit is used for carrying out radiometric scaling on the full-color image by utilizing the radiance value and the apparent reflectivity so as to obtain a scaled full-color image.
8. The optical satellite image-based ship detection system according to claim 7, wherein the target detection module comprises:
the training sample acquisition unit is used for classifying the ships according to the texture features of the ships in the remote sensing satellite images to obtain training samples;
the training unit is used for inputting the training sample into a deep convolutional network for training to obtain a ship target detection model;
the target detection unit is used for detecting the fused image by using the ship target detection model to obtain a plurality of rotary detection frames;
and the detection frame merging unit is used for merging the redundant rotation detection frames by using a soft rotation NMS method to obtain the ship target.
9. An electronic device comprising a bus, a transceiver, a memory, a processor and a computer program stored on the memory and executable on the processor, the transceiver, the memory and the processor being connected via the bus, wherein the computer program when executed by the processor implements the steps of a method for optical satellite imagery based ship detection according to any one of claims 1 to 4.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of a method for optical satellite image-based ship detection according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210976576.6A CN115346133A (en) | 2022-08-15 | 2022-08-15 | Ship detection method and system based on optical satellite image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210976576.6A CN115346133A (en) | 2022-08-15 | 2022-08-15 | Ship detection method and system based on optical satellite image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115346133A true CN115346133A (en) | 2022-11-15 |
Family
ID=83951143
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210976576.6A Pending CN115346133A (en) | 2022-08-15 | 2022-08-15 | Ship detection method and system based on optical satellite image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115346133A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116580313A (en) * | 2023-03-31 | 2023-08-11 | 数字太空(北京)科技股份公司 | Abnormal ship identification method and device based on digital twin and remote sensing |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107564016A (en) * | 2017-08-28 | 2018-01-09 | 国家***第二海洋研究所 | A kind of Multi-Band Remote Sensing Images segmentation of integrally object light spectrum information and labeling method |
| CN109815916A (en) * | 2019-01-28 | 2019-05-28 | 成都蝉远科技有限公司 | A kind of recognition methods of vegetation planting area and system based on convolutional neural networks algorithm |
| CN111898534A (en) * | 2020-07-30 | 2020-11-06 | 上海交通大学 | Satellite-borne target fusion detection method based on rotation region convolutional neural network |
| CN112365540A (en) * | 2020-11-18 | 2021-02-12 | 北京观微科技有限公司 | Ship target positioning detection method and system suitable for multiple scales |
| CN112712049A (en) * | 2021-01-11 | 2021-04-27 | 中国电子科技集团公司第十五研究所 | Satellite image ship model identification method under small sample condition |
| CN114743015A (en) * | 2022-04-02 | 2022-07-12 | 中山大学 | Target detection method and system based on angle adaptive fusion |
-
2022
- 2022-08-15 CN CN202210976576.6A patent/CN115346133A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107564016A (en) * | 2017-08-28 | 2018-01-09 | 国家***第二海洋研究所 | A kind of Multi-Band Remote Sensing Images segmentation of integrally object light spectrum information and labeling method |
| CN109815916A (en) * | 2019-01-28 | 2019-05-28 | 成都蝉远科技有限公司 | A kind of recognition methods of vegetation planting area and system based on convolutional neural networks algorithm |
| CN111898534A (en) * | 2020-07-30 | 2020-11-06 | 上海交通大学 | Satellite-borne target fusion detection method based on rotation region convolutional neural network |
| CN112365540A (en) * | 2020-11-18 | 2021-02-12 | 北京观微科技有限公司 | Ship target positioning detection method and system suitable for multiple scales |
| CN112712049A (en) * | 2021-01-11 | 2021-04-27 | 中国电子科技集团公司第十五研究所 | Satellite image ship model identification method under small sample condition |
| CN114743015A (en) * | 2022-04-02 | 2022-07-12 | 中山大学 | Target detection method and system based on angle adaptive fusion |
Non-Patent Citations (1)
| Title |
|---|
| ZHENYU CUI等: "SKNet: Detecting Rotated Ships as Keypoints in Optical Remote Sensing Images", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSIN》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116580313A (en) * | 2023-03-31 | 2023-08-11 | 数字太空(北京)科技股份公司 | Abnormal ship identification method and device based on digital twin and remote sensing |
| CN116580313B (en) * | 2023-03-31 | 2023-11-14 | 数字太空(北京)科技股份公司 | Abnormal ship identification method and device based on digital twin and remote sensing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wei et al. | HRSID: A high-resolution SAR images dataset for ship detection and instance segmentation | |
| Chen et al. | Ship detection from coastal surveillance videos via an ensemble Canny-Gaussian-morphology framework | |
| CN101214851B (en) | Intelligent all-weather actively safety early warning system and early warning method thereof for ship running | |
| CN113627473B (en) | Multi-mode sensor-based water surface unmanned ship environment information fusion sensing method | |
| Mattyus | Near real-time automatic marine vessel detection on optical satellite images | |
| US20220024549A1 (en) | System and method for measuring the distance to an object in water | |
| CN111666854B (en) | High-resolution SAR image vehicle target detection method fusing statistical significance | |
| CN112347895A (en) | Ship remote sensing target detection method based on boundary optimization neural network | |
| CN112487912B (en) | Arbitrary direction ship detection method based on improved YOLOv3 | |
| CN113033315A (en) | Rare earth mining high-resolution image identification and positioning method | |
| Zhang et al. | Research on unmanned surface vehicles environment perception based on the fusion of vision and lidar | |
| CN114821358A (en) | Optical remote sensing image marine ship target extraction and identification method | |
| CN113486819A (en) | Ship target detection method based on YOLOv4 algorithm | |
| CN107169412B (en) | Remote sensing image harbor-berthing ship detection method based on mixed model decision | |
| CN115346133A (en) | Ship detection method and system based on optical satellite image | |
| Shi et al. | Obstacle type recognition in visual images via dilated convolutional neural network for unmanned surface vehicles | |
| Fu et al. | Real-time infrared horizon detection in maritime and land environments based on hyper-laplace filter and convolutional neural network | |
| CN113205139A (en) | Unmanned ship water sphere detection method based on density clustering | |
| Stets et al. | Comparing spectral bands for object detection at sea using convolutional neural networks | |
| Wang et al. | Scattering Information Fusion Network for Oriented Ship Detection in SAR Images | |
| CN113484864B (en) | Unmanned ship-oriented navigation radar and photoelectric pod collaborative environment sensing method | |
| CN114445572B (en) | DeeplabV3+ based method for immediately positioning and constructing map of obstacle in strange sea area | |
| CN113052117B (en) | AIS data-assisted Rayleigh CFAR detection method for SAR image | |
| CN111709308B (en) | Unmanned aerial vehicle-based maritime distress personnel detection and tracking method and system | |
| CN110827257B (en) | Visual navigation positioning method for embedded airborne infrared image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221115 |