CN112070803A - Unmanned ship path tracking method based on SSD neural network model - Google Patents
Unmanned ship path tracking method based on SSD neural network model Download PDFInfo
- Publication number
- CN112070803A CN112070803A CN202010911382.9A CN202010911382A CN112070803A CN 112070803 A CN112070803 A CN 112070803A CN 202010911382 A CN202010911382 A CN 202010911382A CN 112070803 A CN112070803 A CN 112070803A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- unmanned ship
- ssd neural
- ssd
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Geometry (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an unmanned ship path tracking method based on an SSD neural network model, which specifically comprises the following steps: s1, acquiring a video stream, wherein the video stream image comprises a target unmanned ship; s2, sequentially inputting the images in the video stream into an SSD neural network model, wherein the SSD neural network model outputs a target unmanned ship image containing a detection frame; s3, extracting the positions of key points in the target unmanned ship image in the detection frame, and storing the positions; s4, all the key point positions are output based on the time sequence, that is, the travel path of the target unmanned ship is generated. The method comprises the steps of obtaining the positions of key points in an image of the unmanned ship, displaying the positions of the key points based on a time comfort sequence to form a driving path of the unmanned ship, obtaining the driving path of the unmanned ship by an image carding method, reducing the monitoring cost of the path of the unmanned ship, and displaying the path of the unmanned ship visually.
Description
Technical Field
The invention belongs to the field of path tracking, and relates to an unmanned ship path tracking method based on an SSD neural network model.
Background
When the unmanned boat is sailed at sea, some obstacles, such as a ship sailing and a ship operating at sea, are inevitably met. Path tracking of marine vessels is also essential for safety when unmanned boats are sailing at sea. The path tracking of the existing ship is mostly based on a radar or a GPS positioning sensor to track the path of the unmanned ship, and the path tracking mode has the problems of high cost and non-intuitive path display.
Disclosure of Invention
The invention provides an unmanned ship path tracking method based on an SSD neural network model, which is used for tracking a path based on an image of an unmanned ship.
The invention is realized in such a way, and the unmanned ship path tracking method based on the SSD neural network model is characterized by comprising the following steps:
s1, acquiring a video stream, wherein the video stream image comprises a target unmanned ship;
s2, sequentially inputting the images in the video stream into an SSD neural network model, wherein the SSD neural network model outputs a target unmanned ship image containing a detection frame;
s3, extracting the positions of key points in the target unmanned ship image in the detection frame, and storing the positions;
s4, all the key point positions are output based on the time sequence, that is, the travel path of the target unmanned ship is generated.
Furthermore, the detection frame is a rectangular frame, and the key point is the central point of the detection frame.
Further, the Center point coordinate Center is calculated by the two vertex coordinates of the detection frame, and the calculation formula is as follows:
wherein the content of the first and second substances,
the coordinates of the Center point, (xx1, yy1) are the first vertex coordinates in the detection frame closest to the origin, and (xx2, yy2) are the first coordinates in the detection frame farthest from the origin.
Further, the SSD neural network model is improved, an L2 regularization penalty term is added to the seventh layer of the neural network of the improved SSD neural network model, and the characteristic sensitivity of the seventh layer is reduced.
Further, the L2 regularization is implemented based on the L2 norm, i.e.:
wherein C is a regularization term, C0Representing the error of the training sample of the regularization term, n is the number of the training samples, lambda is the coefficient of the regularization term, omegaiIs the weight of the ith neuron in the neural network.
Further, before step S1, the method further includes:
constructing a sample set of a ship target, wherein the sample set comprises a training sample set and a test sample set;
training the SSD neural network model pair based on the training sample set;
and testing the trained SSD neural network model based on the test sample set, and outputting the trained SSD neural network model when the identification accuracy of the trained SSD neural network model is greater than an accuracy threshold.
According to the unmanned ship route monitoring method, the positions of the key points in the image of the unmanned ship are obtained, the positions of the key points are displayed on the basis of the time comfort sequence, the unmanned ship route is formed, the unmanned ship route is obtained through the image combing method, the unmanned ship route monitoring cost is reduced, and the unmanned ship route can be visually displayed.
Drawings
Fig. 1 is a flowchart of an SSD neural network model-based unmanned ship path tracking method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of dynamic ship detection of an improved SSD neural network model according to an embodiment of the present invention, wherein (a) is a schematic diagram of dynamic ship detection one, and (b) is a schematic diagram of dynamic ship detection 2;
FIG. 3 is a center position labeling diagram provided in accordance with an embodiment of the present invention;
FIG. 4 is a flow chart of dynamic ship detection and path mapping according to an embodiment of the present invention;
fig. 5 is a diagram of dynamic ship detection and path tracking according to an embodiment of the present invention, in which (a), (b), (c), and (d) represent four consecutive frames of dynamic ship detection diagrams.
Detailed Description
The following detailed description of the embodiments of the present invention will be given in order to provide those skilled in the art with a more complete, accurate and thorough understanding of the inventive concept and technical solutions of the present invention.
Fig. 1 is a flowchart of an SSD neural network model-based unmanned ship path tracking method according to an embodiment of the present invention, which specifically includes the following steps:
s1, acquiring a video stream, wherein the video stream image comprises a target unmanned ship;
the method for acquiring the dynamic video stream comprises two methods, one is to acquire the dynamic video stream through a camera, but in order to ensure the definition, a network camera is generally adopted for acquisition, and the other is to acquire the recorded high-definition video directly from a network and other ways;
s2, sequentially inputting the images in the video stream into an SSD neural network model, wherein the SSD neural network model outputs unmanned ship images containing detection frames;
and inputting the obtained video stream into an SSD neural network model for detection, wherein the video stream consists of a plurality of frames of images with time identifications, each frame of image comprises the target unmanned ship, and each frame of image returns a corresponding detection frame and an unmanned ship image.
In the embodiment of the invention, the SSD neural network model is an improved SSD model, the improved SSD neural network model is characterized in that an L2 regularization punishment item is added to the seventh layer of the neural network to reduce the characteristic sensitivity of the seventh layer,
the neural network with the smaller weight matrix can simplify the model, and the purpose of regularization is to add a penalty term to the seventh layer, so that the value of the weight matrix of the seventh layer is reduced, and the characteristics of the seventh layer are reduced. In order to prevent the ship target recognition model from being particularly sensitive to the features of the seventh layer, L2 regularization is added to the seventh layer of the neural network, so that the expression capacities of all the features tend to be balanced, and even if noise with abnormal prominence exists on a certain feature, the accuracy of the output of the ship target recognition model is not influenced. The L2 regularization is also called weight decay because it forces the weights to decay toward 0 (but not 0), and the L2 regularization is implemented based on the L2 norm, i.e.:
wherein C is a regularization term, C0Representing the error of the training sample of the regularization term, n is the number of the training samples, lambda is the coefficient of the regularization term, omegaiWeighing regularization term and C for the weight of the ith neuron in the neural network0Specific gravity of term, ωiIn order to be the weight, the weight is,
i.e., the L2 regular term.
The neural network with the smaller weight matrix can simplify the model, the regularization aims to add a penalty term to the seventh layer, and when the ship target is trained and identified, due to the fact that characteristic noises of other obstacles exist in the image and enter the neurons, the identification and training of the image can be affected, so that the regularization is used, the value of the weight matrix of the seventh layer is reduced, and the obstacle characteristics of the seventh layer are reduced. If the eighth layer and the subsequent layers are regularized, the weight of the ship features is reduced, and therefore training and recognition effects are affected. And increasing regularization on a seventh layer of the SSD neural network model to reduce the value of a weight matrix of the seventh layer, so that the barrier characteristics of the seventh layer are reduced, and the accuracy and the recognition confidence of ship target recognition are improved.
And the SSD neural network model frames the recognition result in a rectangular frame mode, and if the recognition result cannot be recognized, the SSD neural network model is frameless. Fig. 2 shows two frames of the recognition result in which the ship target is arbitrarily framed.
S3, extracting the positions of key points in the detection frame in each frame of image, and storing the positions, wherein the key position points are determined as the central points in the detection frame and can also be understood as the middle points of the target unmanned ship image in the detection frame;
in the embodiment of the invention, after the video stream is sent into the improved SSD neural network model, each frame of picture of the video stream enters the improved SSD neural network model in sequence for identification, then the vertex coordinates of the detection frame and the detection frame are returned, and the Center coordinate Center is calculated through the two vertex coordinates of the detection frame:
wherein
As the coordinates of the Center point, (xx1, yy1) is the first vertex coordinate closest to the origin in the detection box, (xx2, yy2) is the first coordinate farthest from the origin in the detection box, and the python language used herein is used for system writing, and a reverse slice is used when outputting coordinates, so that the use of (xx1, yy1) and (xx2, yy2) are opposite, as shown in fig. 3;
s4, the key point positions of all frames are output based on the time sequence, that is, the travel path of the target unmanned ship is generated.
Fig. 4 is a schematic diagram of dynamic ship detection and path drawing, after a system acquires a video stream, the video stream is input into an improved SSD frame, each frame of a ship detection result including a detection frame is output by the frame, then, path key point extraction is performed on each frame of the image, so as to obtain path key point images in different time periods, and then, the key points are stored and sequentially output correspondingly, so as to generate a driving path of a ship.
The unmanned ship Video (Carver C43 Coupe Running Video) is acquired through the network, ship detection and path tracking are carried out on the Video stream scene from 00 hours 00 minutes 09.30 seconds to 00 hours 00 minutes 14.70 seconds, and the obtained experimental result is shown in fig. 5. When in detection, the detection frame shakes, so that the key points of the path are unstable, and the path shakes, but the local shaking does not influence the drawing of the real path, and when the path is lengthened, the influence of the local instability is reduced.
In the embodiment of the present invention, before step S1, the method further includes:
constructing a sample set of a ship target, wherein the sample set comprises a training sample set and a test sample set;
training the SSD neural network model based on the training sample set;
and testing the trained SSD neural network model based on the test sample set, and outputting the trained SSD neural network model when the identification accuracy of the trained SSD neural network model is greater than an accuracy threshold.
In the embodiment of the invention, the sample comprises the ship image of the ship target, the class and the position of the ship target and the background in the ship image are labeled to form a training sample, the training sample is put into a training sample set, the training sample set comprises a large number of labeled ship images, and the ship image comprising the ship target is taken as a test sample and put into a test sample set. The ship target is positioned on a course of the ship body, and the ship target is regarded as an obstacle on the course of the ship body.
According to the unmanned ship route monitoring method, the positions of the key points in the image of the unmanned ship are obtained, the positions of the key points are displayed on the basis of the time comfort sequence, the unmanned ship route is formed, the unmanned ship route is obtained through the image combing method, the unmanned ship route monitoring cost is reduced, and the unmanned ship route can be visually displayed.
The invention has been described above with reference to the accompanying drawings, it is obvious that the invention is not limited to the specific implementation in the above-described manner, and it is within the scope of the invention to apply the inventive concept and solution to other applications without substantial modification.
Claims (6)
1. An unmanned ship path tracking method based on an SSD neural network model is characterized by comprising the following steps:
s1, acquiring a video stream, wherein the video stream image comprises a target unmanned ship;
s2, sequentially inputting the images in the video stream into an SSD neural network model, wherein the SSD neural network model outputs a target unmanned ship image containing a detection frame;
s3, extracting the positions of key points in the target unmanned ship image in the detection frame, and storing the positions;
s4, all the key point positions are output based on the time sequence, that is, the travel path of the target unmanned ship is generated.
2. The SSD neural network model-based unmanned ship path tracking method as claimed in claim 1, wherein the detection frame is a rectangular frame, and the key point is a central point of the detection frame.
3. The method for unmanned ship path tracking based on SSD neural network model of claim 2, wherein the Center point coordinate Center is calculated by two vertex coordinates of the detection box, and the calculation formula is as follows:
wherein the content of the first and second substances,
the coordinates of the Center point, (xx1, yy1) are the first vertex coordinates in the detection frame closest to the origin, and (xx2, yy2) are the first coordinates in the detection frame farthest from the origin.
4. The unmanned ship path tracking method based on the SSD neural network model as claimed in claim 1 or 2, wherein the SSD neural network model is improved, and an L2 regularization penalty term is added to the improved SSD neural network model at a seventh layer of the neural network, so that the feature sensitivity of the seventh layer is reduced.
5. The SSD neural network model-based unmanned ship path tracking method of claim 4, wherein L2 regularization is implemented based on an L2 norm:
wherein C is a regularization term, C0Representing the error of the training sample of the regularization term, n is the number of the training samples, lambda is the coefficient of the regularization term, omegaiIs the weight of the ith neuron in the neural network.
6. The SSD neural network model-based unmanned ship path tracking method of claim 1, further comprising, before step S1:
constructing a sample set of a ship target, wherein the sample set comprises a training sample set and a test sample set;
training the SSD neural network model pair based on the training sample set;
and testing the trained SSD neural network model based on the test sample set, and outputting the trained SSD neural network model when the identification accuracy of the trained SSD neural network model is greater than an accuracy threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010911382.9A CN112070803A (en) | 2020-09-02 | 2020-09-02 | Unmanned ship path tracking method based on SSD neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010911382.9A CN112070803A (en) | 2020-09-02 | 2020-09-02 | Unmanned ship path tracking method based on SSD neural network model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112070803A true CN112070803A (en) | 2020-12-11 |
Family
ID=73666606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010911382.9A Pending CN112070803A (en) | 2020-09-02 | 2020-09-02 | Unmanned ship path tracking method based on SSD neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112070803A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114275134A (en) * | 2021-12-28 | 2022-04-05 | 上海海事大学 | Method for preventing underwater aquatic weed from winding on propeller of unmanned ship |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104376576A (en) * | 2014-09-04 | 2015-02-25 | 华为技术有限公司 | Target tracking method and device |
| CN108230353A (en) * | 2017-03-03 | 2018-06-29 | 北京市商汤科技开发有限公司 | Method for tracking target, system and electronic equipment |
| CN109190636A (en) * | 2018-07-30 | 2019-01-11 | 北京航空航天大学 | A kind of remote sensing images Ship Target information extracting method |
| CN110285793A (en) * | 2019-07-08 | 2019-09-27 | 中原工学院 | A kind of Vehicular intelligent survey track approach based on Binocular Stereo Vision System |
| CN111027445A (en) * | 2019-12-04 | 2020-04-17 | 安徽工程大学 | Target identification method for marine ship |
| CN111127509A (en) * | 2018-10-31 | 2020-05-08 | 杭州海康威视数字技术股份有限公司 | Target tracking method, device and computer readable storage medium |
| CN111314611A (en) * | 2020-02-26 | 2020-06-19 | 浙江大华技术股份有限公司 | Shooting method and device for multiple moving objects |
| CN111382708A (en) * | 2020-03-11 | 2020-07-07 | 广东工业大学 | Method and device for detecting burglary behavior of climbing stairs and turning windows in real time |
| CN111553474A (en) * | 2019-12-19 | 2020-08-18 | 珠海大横琴科技发展有限公司 | Ship detection model training method and ship tracking method based on unmanned aerial vehicle video |
-
2020
- 2020-09-02 CN CN202010911382.9A patent/CN112070803A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104376576A (en) * | 2014-09-04 | 2015-02-25 | 华为技术有限公司 | Target tracking method and device |
| CN108230353A (en) * | 2017-03-03 | 2018-06-29 | 北京市商汤科技开发有限公司 | Method for tracking target, system and electronic equipment |
| CN109190636A (en) * | 2018-07-30 | 2019-01-11 | 北京航空航天大学 | A kind of remote sensing images Ship Target information extracting method |
| CN111127509A (en) * | 2018-10-31 | 2020-05-08 | 杭州海康威视数字技术股份有限公司 | Target tracking method, device and computer readable storage medium |
| CN110285793A (en) * | 2019-07-08 | 2019-09-27 | 中原工学院 | A kind of Vehicular intelligent survey track approach based on Binocular Stereo Vision System |
| CN111027445A (en) * | 2019-12-04 | 2020-04-17 | 安徽工程大学 | Target identification method for marine ship |
| CN111553474A (en) * | 2019-12-19 | 2020-08-18 | 珠海大横琴科技发展有限公司 | Ship detection model training method and ship tracking method based on unmanned aerial vehicle video |
| CN111314611A (en) * | 2020-02-26 | 2020-06-19 | 浙江大华技术股份有限公司 | Shooting method and device for multiple moving objects |
| CN111382708A (en) * | 2020-03-11 | 2020-07-07 | 广东工业大学 | Method and device for detecting burglary behavior of climbing stairs and turning windows in real time |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114275134A (en) * | 2021-12-28 | 2022-04-05 | 上海海事大学 | Method for preventing underwater aquatic weed from winding on propeller of unmanned ship |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112308019B (en) | SAR ship target detection method based on network pruning and knowledge distillation | |
| Gasienica-Jozkowy et al. | An ensemble deep learning method with optimized weights for drone-based water rescue and surveillance | |
| CN111738112B (en) | Remote sensing ship image target detection method based on deep neural network and self-attention mechanism | |
| CN111967480A (en) | Multi-scale self-attention target detection method based on weight sharing | |
| CN111027445B (en) | Marine ship target identification method | |
| Yao et al. | ShorelineNet: An efficient deep learning approach for shoreline semantic segmentation for unmanned surface vehicles | |
| Zhang et al. | A object detection and tracking method for security in intelligence of unmanned surface vehicles | |
| CN114022697A (en) | Vehicle re-identification method and system based on multitask learning and knowledge distillation | |
| Yang et al. | An enhanced SiamMask network for coastal ship tracking | |
| CN113378675A (en) | Face recognition method for simultaneous detection and feature extraction | |
| CN112633274A (en) | Sonar image target detection method and device and electronic equipment | |
| CN113591592A (en) | Overwater target identification method and device, terminal equipment and storage medium | |
| CN112070803A (en) | Unmanned ship path tracking method based on SSD neural network model | |
| CN114596335A (en) | Unmanned ship target detection tracking method and system | |
| Žust et al. | Temporal context for robust maritime obstacle detection | |
| CN116958780A (en) | Cross-scale target detection method and system | |
| CN115797770A (en) | Continuous image target detection method, system and terminal considering relative movement of target | |
| CN111950586B (en) | Target detection method for introducing bidirectional attention | |
| CN115376023A (en) | Cultivation area detection method based on deformation convolutional network, unmanned aerial vehicle and medium | |
| CN115035429A (en) | Aerial photography target detection method based on composite backbone network and multiple measuring heads | |
| Wu et al. | SARFB: Strengthened asymmetric receptive field block for accurate infrared ship detection | |
| Li et al. | Research on ROI algorithm of ship image based on improved YOLO | |
| Cai et al. | FE-YOLO: YOLO ship detection algorithm based on feature fusion and feature enhancement | |
| Zhang et al. | Ship detection based on improved YOLO algorithm | |
| Qu et al. | Multi-Task Learning-Enabled Automatic Vessel Draft Reading for Intelligent Maritime Surveillance |
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 |