CN117671504A - An offshore wind power identification method and system based on yolo algorithm - Google Patents
An offshore wind power identification method and system based on yolo algorithm Download PDFInfo
- Publication number
- CN117671504A CN117671504A CN202311697617.9A CN202311697617A CN117671504A CN 117671504 A CN117671504 A CN 117671504A CN 202311697617 A CN202311697617 A CN 202311697617A CN 117671504 A CN117671504 A CN 117671504A
- Authority
- CN
- China
- Prior art keywords
- offshore wind
- wind power
- module
- model
- power identification
- 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
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 47
- 230000007246 mechanism Effects 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims description 26
- 238000012360 testing method Methods 0.000 claims description 18
- 238000012795 verification Methods 0.000 claims description 16
- 238000004519 manufacturing process Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000002372 labelling Methods 0.000 claims description 4
- 230000000873 masking effect Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 55
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000010801 machine learning Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 230000002123 temporal effect Effects 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 239000000306 component Substances 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007667 floating Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
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
-
- 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
- G06V10/77—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
- G06V10/77—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
- G06V10/776—Validation; Performance evaluation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Wind Motors (AREA)
Abstract
本发明公开了一种基于yolo算法的海上风电识别方法及系统,涉及机器学习技术领域。该方法包括:采集多个地区的海上风电场遥感图像,并制作海上风电样本集;基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型;利用海上风电样本集对初始海上风电识别模型进行训练和验证,以得到目标海上风电识别模型;获取实时海上风电影像,利用目标海上风电识别模型对海上风电目标进行识别。本发明可对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。
The invention discloses an offshore wind power identification method and system based on the yolo algorithm, and relates to the field of machine learning technology. The method includes: collecting remote sensing images of offshore wind farms in multiple areas and producing an offshore wind power sample set; building an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module; using the offshore wind power sample set to The initial offshore wind power recognition model is trained and verified to obtain the target offshore wind power recognition model; real-time offshore wind film images are obtained, and the target offshore wind power recognition model is used to identify offshore wind power targets. The invention can accurately and stably detect offshore wind power, effectively improving detection efficiency and detection accuracy.
Description
技术领域Technical field
本发明涉及机器学习技术领域,具体而言,涉及一种基于yolo算法的海上风电识别方法及系统。The present invention relates to the field of machine learning technology, and specifically to an offshore wind power identification method and system based on the yolo algorithm.
背景技术Background technique
现有的深度学习检测算法主要基于以R-CNN(Girshick,2015;Girshick et al.,2014;He et al.,2017;Ren et al.,2017)、Fast R-CNN(Girshick,2015)、Faster R-CNN(Ren et al.,2017)、SPP-NET(He et al.,2014)等为代表的两级目标检测网络和以YOLO(Bochkovskiy et al.,2020;Redmon et al.,2016;Redmon and Farhadi,2018,2017)、SSD(W.Liu et al.,2016)为代表的一级目标检测网络。两级网络具有较高的检测精度,但是需要大量的数据集、较高的训练成本,检测速度不快。一阶段检测法直接在网络中提取特征并预测目标类别和位置,检测速度快、所需数据量小,适用小样本数据集。Existing deep learning detection algorithms are mainly based on R-CNN (Girshick, 2015; Girshick et al., 2014; He et al., 2017; Ren et al., 2017), Fast R-CNN (Girshick, 2015), Two-level target detection networks represented by Faster R-CNN (Ren et al., 2017), SPP-NET (He et al., 2014), and YOLO (Bochkovskiy et al., 2020; Redmon et al., 2016 ; Redmon and Farhadi, 2018, 2017) and SSD (W. Liu et al., 2016) are representative first-level target detection networks. The two-level network has high detection accuracy, but requires a large number of data sets, high training costs, and the detection speed is not fast. The one-stage detection method directly extracts features from the network and predicts the target category and location. It has fast detection speed, requires a small amount of data, and is suitable for small sample data sets.
目前深度学习技术在海上风电目标识别应用已有初步进展。Hoeser等人(2022)(Hoeser and Kuenzer,2022)构建了全球海上风电深度学习数据集,为海上目标检测提供了研究基础。Hoeser等人(2022)(Hoeser et al.,2022)通过该数据集结合CNN模型检测全球范围海上风电,得到2016-2021年全球海上风电的时空分布数据集。但仅以北海(欧洲)、东海(中国)的海上风电作为验证集不足以评估全球不同海况、不同下垫面海上风电的识别情况,且实验在轻量级部署等方面仍有改进的空间。即使深度神经网络在海上风电探测领域已经取得了进展,然而,基于深度学习的方法依然面临着数据集可用性、模型泛化性和探测结果鲁棒性这三点挑战。首先,深度学习通常需要大量的样本数据对模型进行训练,从而获得较高的区域精度;然而,目前开源可获取的高质量训练数据集仍缺乏。其次,海上风电在海域中分布稀疏,海表杂波背景在遥感影像中存在时相和空间差异,这些因素给深度学习模型的泛化性带来巨大挑战。再者,海上风电目标对象面积小,海上漂浮物或临时移动的物体(如船舶、云、石油钻井平台)干扰使得风电检测算法既耗时又缺乏鲁棒性。At present, deep learning technology has made preliminary progress in the application of offshore wind power target recognition. Hoeser et al. (2022) (Hoeser and Kuenzer, 2022) constructed a global offshore wind power deep learning data set, providing a research basis for offshore target detection. Hoeser et al. (2022) (Hoeser et al., 2022) used this data set combined with the CNN model to detect global offshore wind power, and obtained the spatiotemporal distribution data set of global offshore wind power from 2016 to 2021. However, only using offshore wind power in the North Sea (Europe) and the East China Sea (China) as a verification set is not enough to evaluate the identification of offshore wind power in different sea conditions and different underlying surfaces around the world, and the experiment still has room for improvement in aspects such as lightweight deployment. Even though deep neural networks have made progress in the field of offshore wind power detection, methods based on deep learning still face three challenges: data set availability, model generalization, and robustness of detection results. First, deep learning usually requires a large amount of sample data to train the model to obtain high regional accuracy; however, there is still a lack of high-quality training data sets available in open source. Secondly, offshore wind power is sparsely distributed in the sea area, and there are temporal and spatial differences in the sea surface clutter background in remote sensing images. These factors bring huge challenges to the generalization of deep learning models. Furthermore, the target area of offshore wind power is small, and interference from floating objects on the sea or temporarily moving objects (such as ships, clouds, and oil drilling platforms) makes the wind power detection algorithm time-consuming and lacks robustness.
因此,如何对国家尺度的海上风电目标进行精准高效的识别成为一个亟需解决的问题。Therefore, how to accurately and efficiently identify national-scale offshore wind power targets has become an urgent problem that needs to be solved.
发明内容Contents of the invention
为了克服上述问题或者至少部分地解决上述问题,本发明提供一种基于yolo算法的海上风电识别方法及系统,可对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。In order to overcome the above problems or at least partially solve the above problems, the present invention provides an offshore wind power identification method and system based on the yolo algorithm, which can accurately and stably detect offshore wind power, effectively improving detection efficiency and detection accuracy.
为解决上述技术问题,本发明采用的技术方案为:In order to solve the above technical problems, the technical solutions adopted by the present invention are:
第一方面,本发明提供一种基于yolo算法的海上风电识别方法,包括以下步骤:In a first aspect, the present invention provides an offshore wind power identification method based on the yolo algorithm, which includes the following steps:
采集多个地区的海上风电场遥感图像,并制作海上风电样本集;Collect remote sensing images of offshore wind farms in multiple areas and create an offshore wind power sample collection;
基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型;Build an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module;
利用海上风电样本集对初始海上风电识别模型进行训练和验证,以得到目标海上风电识别模型;Use the offshore wind power sample set to train and verify the initial offshore wind power identification model to obtain the target offshore wind power identification model;
获取实时海上风电影像,利用目标海上风电识别模型对海上风电目标进行识别。Obtain real-time offshore wind images and use the target offshore wind power identification model to identify offshore wind power targets.
本发明创建了一个适用于中国海上风电目标识别的深度学习训练数据集,并基于YOLO算法中的YOLOv5算法进行优化改进,引入CA注意力机制和RFB多分支卷积模块,构建有效的海上风电识别模型(YOLOv5s-CR模型),加入坐标注意力CA模块提高模型的定位能力,融合上下文增强模块RFB增加感受野减少在大尺度复杂海表背景下模型的错检率,可以实现中国海上风电的快速检测和精确定位。基于构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明构建的海上风电识别模型时相泛化和地理泛化性能好。This invention creates a deep learning training data set suitable for China's offshore wind power target recognition, optimizes and improves it based on the YOLOv5 algorithm in the YOLO algorithm, introduces the CA attention mechanism and RFB multi-branch convolution module, and constructs effective offshore wind power recognition Model (YOLOv5s-CR model), the coordinate attention CA module is added to improve the positioning ability of the model, and the context enhancement module RFB is integrated to increase the receptive field and reduce the error detection rate of the model in the large-scale complex sea surface background, which can realize the rapid development of China's offshore wind power detection and precise positioning. Based on the constructed offshore wind power identification model, offshore wind power is accurately and stably detected, effectively improving detection efficiency and detection accuracy. The offshore wind power identification model constructed by the present invention has good temporal generalization and geographical generalization performance.
基于第一方面,进一步地,上述制作海上风电样本集的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of producing an offshore wind power sample set includes the following steps:
对海上风电场遥感图像进行预处理;Preprocess remote sensing images of offshore wind farms;
对预处理后的海上风电场遥感图像进行均值计算,对海上风电场遥感图像进行优化;Perform mean calculation on the preprocessed offshore wind farm remote sensing images and optimize the offshore wind farm remote sensing images;
对优化后的海上风电场遥感图像进行样本标记,以得到海上风电样本测试集和海上风电样本验证集;Perform sample labeling on the optimized offshore wind farm remote sensing images to obtain an offshore wind power sample test set and an offshore wind power sample verification set;
通过ArcGIS Pro图像处理工具,制作海上风电样本测试集和海上风电样本验证集的样本切片,并将所有样本的位置信息和类别信息写入对应的xml文件中;Use the ArcGIS Pro image processing tool to create sample slices of the offshore wind power sample test set and offshore wind power sample verification set, and write the location information and category information of all samples into the corresponding xml files;
选取部分样本切片数据作为训练集和测试集,形成VOC格式的数据集。Select some sample slice data as the training set and test set to form a data set in VOC format.
基于第一方面,进一步地,上述对海上风电场遥感图像进行预处理的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method for preprocessing offshore wind farm remote sensing images includes the following steps:
对海上风电场遥感图像的轨道状态向量进行更新,针对海上风电场遥感图像中的地距多视影像GRD,使用修正后的轨道文件更新轨道元数据;Update the orbit state vector of offshore wind farm remote sensing images, and use the corrected orbit file to update orbit metadata for the ground-distance multi-view image GRD in offshore wind farm remote sensing images;
通过对低值像元掩膜抑制GRD场景边缘的低强度噪声和无效数据,消除海上风电场遥感图像的子条带中的附加噪声;Suppress low-intensity noise and invalid data at the edge of the GRD scene by masking low-value pixels to eliminate additional noise in sub-strips of offshore wind farm remote sensing images;
采用元数据中的传感器校准参数计算反向散射强度来进行辐射校准,并结合数字高程模型对海上风电场遥感图像进行正射校正。The sensor calibration parameters in the metadata are used to calculate the backscattering intensity for radiometric calibration, and the digital elevation model is used to perform orthorectification of remote sensing images of offshore wind farms.
基于第一方面,进一步地,该基于yolo算法的海上风电识别方法,还包括以下步骤:Based on the first aspect, further, the offshore wind power identification method based on the yolo algorithm further includes the following steps:
对数据集中的图像进行图像增强处理。Image enhancement is performed on the images in the dataset.
基于第一方面,进一步地,上述利用海上风电样本集对初始海上风电识别模型进行训练和验证的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of training and verifying the initial offshore wind power identification model using the offshore wind power sample set includes the following steps:
分别训练集和测试集对初始海上风电识别模型进行训练和验证。The initial offshore wind power identification model is trained and verified using the training set and the test set respectively.
基于第一方面,进一步地,上述基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of constructing an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module includes the following steps:
在YOLOv5算法的Backbone模块中引入CA注意力机制,构建CS-CA模块;Introduce the CA attention mechanism into the Backbone module of the YOLOv5 algorithm and build the CS-CA module;
在YOLOv5算法的主干网络中引入RFB多分支卷积模块;Introducing the RFB multi-branch convolution module into the backbone network of the YOLOv5 algorithm;
基于YOLOv5算法、CS-CA模块和RFB多分支卷积模块构建初始海上风电识别模型。An initial offshore wind power identification model is constructed based on the YOLOv5 algorithm, CS-CA module and RFB multi-branch convolution module.
第二方面,本发明提供一种基于yolo算法的海上风电识别系统,包括样本集制作模块、初始模型构建模块、模型训练与验证模块以及海上风电识别模块,其中:In the second aspect, the present invention provides an offshore wind power identification system based on the yolo algorithm, including a sample set production module, an initial model construction module, a model training and verification module, and an offshore wind power identification module, wherein:
样本集制作模块,用于采集多个地区的海上风电场遥感图像,并制作海上风电样本集;The sample set production module is used to collect remote sensing images of offshore wind farms in multiple regions and create offshore wind power sample sets;
初始模型构建模块,用于基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型;Initial model building module, used to build an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module;
模型训练与验证模块,用于利用海上风电样本集对初始海上风电识别模型进行训练和验证,以得到目标海上风电识别模型;The model training and verification module is used to train and verify the initial offshore wind power identification model using the offshore wind power sample set to obtain the target offshore wind power identification model;
海上风电识别模块,用于获取实时海上风电影像,利用目标海上风电识别模型对海上风电目标进行识别。The offshore wind power identification module is used to obtain real-time offshore wind power images and use the target offshore wind power identification model to identify offshore wind power targets.
本系统通过样本集制作模块、初始模型构建模块、模型训练与验证模块以及海上风电识别模块等多个模块,构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明创建了一个适用于中国海上风电目标识别的深度学习训练数据集,并基于YOLO算法中的YOLOv5算法进行优化改进,引入CA注意力机制和RFB多分支卷积模块,构建有效的海上风电识别模型(YOLOv5s-CR模型),加入坐标注意力CA模块提高模型的定位能力,融合上下文增强模块RFB增加感受野减少在大尺度复杂海表背景下模型的错检率,可以实现中国海上风电的快速检测和精确定位。基于构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明构建的海上风电识别模型时相泛化和地理泛化性能好。This system uses multiple modules such as the sample set production module, initial model construction module, model training and verification module, and offshore wind power identification module to build an offshore wind power identification model to accurately and stably detect offshore wind power, effectively improving detection efficiency and detection Accuracy. This invention creates a deep learning training data set suitable for China's offshore wind power target recognition, optimizes and improves it based on the YOLOv5 algorithm in the YOLO algorithm, introduces the CA attention mechanism and RFB multi-branch convolution module, and constructs effective offshore wind power recognition Model (YOLOv5s-CR model), adding the coordinate attention CA module to improve the positioning ability of the model, integrating the context enhancement module RFB to increase the receptive field and reduce the error detection rate of the model in the large-scale complex sea surface background, can achieve rapid development of China's offshore wind power detection and precise positioning. Based on the constructed offshore wind power identification model, offshore wind power is accurately and stably detected, effectively improving detection efficiency and detection accuracy. The offshore wind power identification model constructed by the present invention has good temporal generalization and geographical generalization performance.
第三方面,本申请提供一种电子设备,其包括存储器,用于存储一个或多个程序;处理器;当一个或多个程序被处理器执行时,实现如上述第一方面中任一项的方法。In a third aspect, the present application provides an electronic device, which includes a memory for storing one or more programs; a processor; when one or more programs are executed by the processor, any one of the above first aspects is implemented. Methods.
第四方面,本申请提供一种计算机可读存储介质,其上存储有计算机程序,该计算机程序被处理器执行时实现如上述第一方面中任一项的方法。In a fourth aspect, the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method in any one of the above-mentioned first aspects is implemented.
本发明至少具有如下优点或有益效果:The present invention has at least the following advantages or beneficial effects:
1、创建了一个适用于中国海上风电目标识别的深度学习训练数据集,为后续提供精准全面的数据集。1. Created a deep learning training data set suitable for China's offshore wind power target recognition to provide an accurate and comprehensive data set for subsequent use.
2、基于YOLO算法中的YOLOv5算法进行优化改进,引入CA注意力机制和RFB多分支卷积模块,构建有效的海上风电识别模型(YOLOv5s-CR模型),加入坐标注意力CA模块提高模型的定位能力,融合上下文增强模块RFB增加感受野减少在大尺度复杂海表背景下模型的错检率,可以实现中国海上风电的快速检测和精确定位。基于构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。2. Based on the YOLOv5 algorithm in the YOLO algorithm, optimize and improve it, introduce the CA attention mechanism and RFB multi-branch convolution module, build an effective offshore wind power identification model (YOLOv5s-CR model), and add the coordinate attention CA module to improve the positioning of the model Ability to integrate the context enhancement module RFB to increase the receptive field and reduce the error detection rate of the model in the large-scale complex sea surface background, which can achieve rapid detection and precise positioning of China's offshore wind power. Based on the constructed offshore wind power identification model, offshore wind power is accurately and stably detected, effectively improving detection efficiency and detection accuracy.
3、本发明构建的海上风电识别模型时相泛化和地理泛化性能好。3. The offshore wind power identification model constructed by the present invention has good temporal generalization and geographical generalization performance.
附图说明Description of drawings
为了更清楚地说明本发明实施例的技术方案,下面将对实施例中所需要使用的附图作简单地介绍,应当理解,以下附图仅示出了本发明的某些实施例,因此不应被看作是对范围的限定,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他相关的附图。In order to explain the technical solutions of the embodiments of the present invention more clearly, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and therefore do not It should be regarded as a limitation of the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without exerting creative efforts.
图1为本发明实施例一种基于yolo算法的海上风电识别方法的流程图;Figure 1 is a flow chart of an offshore wind power identification method based on the yolo algorithm according to an embodiment of the present invention;
图2为本发明实施例中YOLOv5s-CR模型构造及流程示意图;Figure 2 is a schematic diagram of the structure and flow of the YOLOv5s-CR model in the embodiment of the present invention;
图3为本发明实施例中某地区潮滩典型海上风电场时序实验的检测结果示意图;Figure 3 is a schematic diagram of the detection results of the timing experiment of a typical offshore wind farm on tidal flats in a certain area in the embodiment of the present invention;
图4为本发明实施例中全球其他国家典型海上风电场检测结果示意图;Figure 4 is a schematic diagram of the detection results of typical offshore wind farms in other countries around the world in the embodiment of the present invention;
图5为本发明实施例中中国海上风电研究区及其样本标注示意图;Figure 5 is a schematic diagram of China's offshore wind power research area and sample labeling in the embodiment of the present invention;
图6为本发明实施例一种基于yolo算法的海上风电识别系统的原理框图;Figure 6 is a functional block diagram of an offshore wind power identification system based on the yolo algorithm according to an embodiment of the present invention;
图7为本发明实施例提供的一种电子设备的结构框图。Figure 7 is a structural block diagram of an electronic device provided by an embodiment of the present invention.
附图标记说明:100、样本集制作模块;200、初始模型构建模块;300、模型训练与验证模块;400、海上风电识别模块;101、存储器;102、处理器;103、通信接口。Explanation of reference signs: 100. Sample set production module; 200. Initial model building module; 300. Model training and verification module; 400. Offshore wind power identification module; 101. Memory; 102. Processor; 103. Communication interface.
具体实施方式Detailed ways
为使本发明实施例的目的、技术方案和优点更加清楚,下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本发明一部分实施例,而不是全部的实施例。通常在此处附图中描述和示出的本发明实施例的组件可以以各种不同的配置来布置和设计。In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, rather than all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.
因此,以下对在附图中提供的本发明的实施例的详细描述并非旨在限制要求保护的本发明的范围,而是仅仅表示本发明的选定实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。Therefore, the following detailed description of the embodiments of the invention provided in the appended drawings is not intended to limit the scope of the claimed invention, but rather to represent selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.
应注意到:相似的标号和字母在下面的附图中表示类似项,因此,一旦某一项在一个附图中被定义,则在随后的附图中不需要对其进行进一步定义和解释。It should be noted that similar reference numerals and letters represent similar items in the following figures, therefore, once an item is defined in one figure, it does not need further definition and explanation in subsequent figures.
需要说明的是,在本文中,诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来,而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且,术语“包括”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or operations are mutually exclusive. any such actual relationship or sequence exists between them. Furthermore, the term "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus including a list of elements includes not only those elements but also other elements not expressly listed, Or it also includes elements inherent to the process, method, article or equipment. Without further limitation, an element defined by the statement "comprises a..." does not exclude the presence of additional identical elements in a process, method, article, or apparatus that includes the stated element.
在本发明实施例的描述中,“多个”代表至少2个。In the description of the embodiments of the present invention, “plurality” represents at least two.
实施例:Example:
如图1-图5所示,第一方面,本发明实施例提供一种基于yolo算法的海上风电识别方法,包括以下步骤:As shown in Figures 1-5, in the first aspect, embodiments of the present invention provide an offshore wind power identification method based on the yolo algorithm, which includes the following steps:
S1、采集多个地区的海上风电场遥感图像,并制作海上风电样本集;S1. Collect remote sensing images of offshore wind farms in multiple areas and create an offshore wind power sample set;
S2、基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型;S2. Build an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module;
S3、利用海上风电样本集对初始海上风电识别模型进行训练和验证,以得到目标海上风电识别模型;S3. Use the offshore wind power sample set to train and verify the initial offshore wind power identification model to obtain the target offshore wind power identification model;
S4、获取实时海上风电影像,利用目标海上风电识别模型对海上风电目标进行识别。S4. Obtain real-time offshore wind film images, and use the target offshore wind power identification model to identify offshore wind power targets.
本发明创建了一个适用于中国海上风电目标识别的深度学习训练数据集,并基于YOLO算法中的YOLOv5算法进行优化改进,引入CA注意力机制和RFB多分支卷积模块,构建有效的海上风电识别模型(YOLOv5s-CR模型),加入坐标注意力CA模块提高模型的定位能力,融合上下文增强模块RFB增加感受野减少在大尺度复杂海表背景下模型的错检率,可以实现中国海上风电的快速检测和精确定位。基于构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明构建的海上风电识别模型时相泛化和地理泛化性能好。This invention creates a deep learning training data set suitable for China's offshore wind power target recognition, optimizes and improves it based on the YOLOv5 algorithm in the YOLO algorithm, introduces the CA attention mechanism and RFB multi-branch convolution module, and constructs effective offshore wind power recognition Model (YOLOv5s-CR model), the coordinate attention CA module is added to improve the positioning ability of the model, and the context enhancement module RFB is integrated to increase the receptive field and reduce the error detection rate of the model in the large-scale complex sea surface background, which can realize the rapid development of China's offshore wind power detection and precise positioning. Based on the constructed offshore wind power identification model, offshore wind power is accurately and stably detected, effectively improving detection efficiency and detection accuracy. The offshore wind power identification model constructed by the present invention has good temporal generalization and geographical generalization performance.
基于第一方面,进一步地,上述制作海上风电样本集的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of producing an offshore wind power sample set includes the following steps:
对海上风电场遥感图像进行预处理;Preprocess remote sensing images of offshore wind farms;
对预处理后的海上风电场遥感图像进行均值计算,对海上风电场遥感图像进行优化;Perform mean calculation on the preprocessed offshore wind farm remote sensing images and optimize the offshore wind farm remote sensing images;
对优化后的海上风电场遥感图像进行样本标记,以得到海上风电样本测试集和海上风电样本验证集;Perform sample labeling on the optimized offshore wind farm remote sensing images to obtain an offshore wind power sample test set and an offshore wind power sample verification set;
通过ArcGIS Pro图像处理工具,制作海上风电样本测试集和海上风电样本验证集的样本切片,并将所有样本的位置信息和类别信息写入对应的xml文件中;Use the ArcGIS Pro image processing tool to create sample slices of the offshore wind power sample test set and offshore wind power sample verification set, and write the location information and category information of all samples into the corresponding xml files;
选取部分样本切片数据作为训练集和测试集,形成VOC格式的数据集。Select some sample slice data as the training set and test set to form a data set in VOC format.
基于第一方面,进一步地,上述对海上风电场遥感图像进行预处理的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method for preprocessing offshore wind farm remote sensing images includes the following steps:
对海上风电场遥感图像的轨道状态向量进行更新,针对海上风电场遥感图像中的地距多视影像GRD,使用修正后的轨道文件更新轨道元数据;Update the orbit state vector of offshore wind farm remote sensing images, and use the corrected orbit file to update orbit metadata for the ground-distance multi-view image GRD in offshore wind farm remote sensing images;
通过对低值像元掩膜抑制GRD场景边缘的低强度噪声和无效数据,消除海上风电场遥感图像的子条带中的附加噪声;Suppress low-intensity noise and invalid data at the edge of the GRD scene by masking low-value pixels to eliminate additional noise in sub-strips of offshore wind farm remote sensing images;
采用元数据中的传感器校准参数计算反向散射强度来进行辐射校准,并结合数字高程模型对海上风电场遥感图像进行正射校正。The sensor calibration parameters in the metadata are used to calculate the backscattering intensity for radiometric calibration, and the digital elevation model is used to perform orthorectification of remote sensing images of offshore wind farms.
在本发明的一些实施例中,海上风电因其独特三叶片刚性多面体形态,高分辨率SAR遥感图像上表现为高亮度纺锤状或长椭圆亮斑。与弱后向散射海表面在图像上反差明显。因此,本文使用了C波段、空间分辨率为10米的Sentinel-1SARGRD产品,选择干涉宽(IW)条带模式和垂直-垂直(VV)极化模式。产品可以通过GoogleEarthEngine(GEE)平台获得。Sentinel-1卫星由两颗极地轨道卫星A和B组成。双星协同工作模式使得Sentinel-1的重访周期从12天缩短至6天,且可以提供全天候观测。样本实际制作流程如下。Step1:Sentinel-1图像预处理。包括更新SAR影像的轨道状态向量,针对地距多视影像(GRD),使用修正后的轨道文件更新轨道元数据;通过对低值像元掩膜抑制GRD场景边缘的低强度噪声和无效数据;消除子条带中的附加噪声以减少多条带采集模式中不同子带之间的不连续性;采用元数据中的传感器校准参数计算反向散射强度来进行辐射校准,并结合数字高程模型(DEM)数据对影像进行正射校正,消除地形的影响。Step2:逐年对2015年-2022年每年12月研究区内的SAR影像进行均值计算并下载。图像切片的长宽比保持一致,且不小于2048像素。Step3:对样本区中国海上风电进行手动标记(图3(b)-(d)),分别得到2143个海上风电样本(测试集)、744个海上风电样本(验证集)。Step4:通过ArcGISPro图像处理工具,制作256×256像素的样本切片,并将样本的位置信息和类别信息写入相应的xml文件中。Step5:之后将输出的2022年2991张的样本切片数据作为训练集、2021年的771张切片数据作为测试集,形成VOC格式的数据集。In some embodiments of the present invention, offshore wind power, due to its unique three-blade rigid polyhedral shape, appears as a high-brightness spindle-shaped or oblong bright spot on high-resolution SAR remote sensing images. The contrast with the weakly backscattered sea surface is obvious in the image. Therefore, this article uses the Sentinel-1SARGRD product with a C-band and a spatial resolution of 10 meters, and selects the interference wide (IW) strip mode and the vertical-vertical (VV) polarization mode. Products are available through the Google Earth Engine (GEE) platform. The Sentinel-1 satellite consists of two polar-orbiting satellites A and B. The dual-star cooperative working mode shortens the revisit period of Sentinel-1 from 12 days to 6 days, and can provide all-weather observations. The actual sample production process is as follows. Step1: Sentinel-1 image preprocessing. Including updating the orbit state vector of SAR images, using the corrected orbit file to update orbit metadata for ground-range multi-view images (GRD); suppressing low-intensity noise and invalid data at the edge of the GRD scene by masking low-value pixels; Eliminate additional noise in sub-strips to reduce discontinuities between different sub-bands in multi-strip acquisition mode; use sensor calibration parameters in metadata to calculate backscatter intensity for radiometric calibration, combined with digital elevation model ( DEM) data is used to orthorectify the image to eliminate the influence of terrain. Step2: Calculate and download the average of the SAR images in the study area in December each year from 2015 to 2022. The aspect ratio of image slices remains consistent and is no less than 2048 pixels. Step 3: Manually label Chinese offshore wind power in the sample area (Figure 3(b)-(d)), and obtain 2143 offshore wind power samples (test set) and 744 offshore wind power samples (validation set) respectively. Step 4: Use the ArcGISPro image processing tool to create 256×256 pixel sample slices, and write the sample's location information and category information into the corresponding xml file. Step5: Then use the output 2991 sample slice data in 2022 as the training set and the 771 slice data in 2021 as the test set to form a VOC format data set.
基于第一方面,进一步地,该基于yolo算法的海上风电识别方法,还包括以下步骤:Based on the first aspect, further, the offshore wind power identification method based on the yolo algorithm further includes the following steps:
对数据集中的图像进行图像增强处理。Image enhancement is performed on the images in the dataset.
了增强模型的泛化能力,对数据集进行图像数据增强,对训练集的2991幅原始中国海上风电进行随机亮度增强、随机对比度增强和锐化、Gamma增强。In order to enhance the generalization ability of the model, image data enhancement was performed on the data set. Random brightness enhancement, random contrast enhancement and sharpening, and Gamma enhancement were performed on the 2991 original Chinese offshore wind power images in the training set.
基于第一方面,进一步地,上述利用海上风电样本集对初始海上风电识别模型进行训练和验证的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of training and verifying the initial offshore wind power identification model using the offshore wind power sample set includes the following steps:
分别训练集和测试集对初始海上风电识别模型进行训练和验证。The initial offshore wind power identification model is trained and verified using the training set and the test set respectively.
基于第一方面,进一步地,上述基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型的方法包括以下步骤:Based on the first aspect, further, the above-mentioned method of constructing an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module includes the following steps:
在YOLOv5算法的Backbone模块中引入CA注意力机制,构建CS-CA模块;Introduce the CA attention mechanism into the Backbone module of the YOLOv5 algorithm and build the CS-CA module;
在YOLOv5算法的主干网络中引入RFB多分支卷积模块;Introducing the RFB multi-branch convolution module into the backbone network of the YOLOv5 algorithm;
基于YOLOv5算法、CS-CA模块和RFB多分支卷积模块构建初始海上风电识别模型。An initial offshore wind power identification model is constructed based on the YOLOv5 algorithm, CS-CA module and RFB multi-branch convolution module.
YOLOv5(是一种高效的单阶段目标检测算法,网络结构由Input模块、Backbone模块、Neck模块和Head模块组成。步骤1:YOLOv5s-CR模型在Input模块中,采用平移、锐化、增强对比度等,提高模型泛化能力和鲁棒性。步骤2:Backbone模块包括Conv、CoT模块和SPFF模块,负责提取图像特征。Backbone模块中我们在C3模块中加入CA注意力机制,加强视觉主干,提高定位能力、实现更好的性能。步骤3:Neck网络使用路径聚合网络(PANet增强特征信息,提高了小目标的检测性能。此外引入RFB模块,通过不同尺寸的卷积核来扩大感受野、捕获特征信息,控制感受野大小和偏心度之间的比率来生成感受野的空间阵列。步骤4:Output模块主要为带有边界框位置、置信度、类别概率的预测框。YOLOv5s-CR提升了模型检测的效率和精确度(该模型的检测流程及架构如图2所示)。YOLOv5 (is an efficient single-stage target detection algorithm. The network structure consists of Input module, Backbone module, Neck module and Head module. Step 1: In the Input module, the YOLOv5s-CR model uses translation, sharpening, contrast enhancement, etc. , improve the model generalization ability and robustness. Step 2: The Backbone module includes Conv, CoT module and SPFF module, which is responsible for extracting image features. In the Backbone module, we add the CA attention mechanism to the C3 module to strengthen the visual backbone and improve positioning ability to achieve better performance. Step 3: The Neck network uses the path aggregation network (PANet) to enhance feature information and improve the detection performance of small targets. In addition, the RFB module is introduced to expand the receptive field and capture features through convolution kernels of different sizes. Information, control the ratio between receptive field size and eccentricity to generate a spatial array of receptive fields. Step 4: The Output module is mainly a prediction box with bounding box position, confidence, and category probability. YOLOv5s-CR improves model detection Efficiency and accuracy (the detection process and architecture of this model are shown in Figure 2).
遥感目标检测领域面对的主要挑战之一是处理复杂背景下的目标识别、小目标误报问题。本发明在C3模块中引入坐标注意力(CoordinateAttention,CA)模块以显著提升识别海上风电特征提取和定位能力。该模块的设计巧妙地融合了中国海上风电的位置信息和通道间关系的高度敏感性,将位置信息嵌入到通道注意力中,可以获取更广泛的上下文信息,CA模块的结构如图2(b)所示。对输入特征图X的每个通道分别使用尺寸为(H,1)和(1,W)的池化内核沿水平和垂直方向进行平均池化操作,编码空间坐标信息。这个操作分别沿两个空间方向聚合特征,产生方向感知的特征图和,分别用于建模沿空间方向的长期依赖性以及保留沿另一个空间方向的精确位置信息。。经过F1操作(1×1卷积核降维)和非线性激活形成中间特征图f,随后中间特征图f被分解成高度注意力张量和宽度注意力张量。通过2个1×1卷积进行升维度操作并结合sigmoid函数降低模型的复杂性和计算开销,获得高度和宽度的注意力向量和,最后将输入特征图X分别与高度和宽度的注意力权重和相乘,生成最终的注意力特征图Y。C3-CA模块通过整合通道和位置信息、运用注意力机制,提高中国海上风电的检测性能、降低误报率。为应对SAR复杂背景下的遥感目标检测问题提供了有效的解决方案且几乎不会增加计算开销。One of the main challenges faced in the field of remote sensing target detection is to deal with target recognition and small target false alarms in complex backgrounds. The present invention introduces a coordinate attention (CA) module into the C3 module to significantly improve the ability to identify offshore wind power feature extraction and positioning. The design of this module cleverly integrates the position information of China's offshore wind power and the high sensitivity of the relationship between channels. By embedding the position information into the channel attention, a wider range of contextual information can be obtained. The structure of the CA module is shown in Figure 2(b). ) shown. For each channel of the input feature map This operation aggregates features along two spatial directions, producing direction-aware feature maps and , which are used to model long-term dependencies along the spatial direction and preserve precise position information along the other spatial direction, respectively. . After F1 operation (1×1 convolution kernel dimensionality reduction) and nonlinear activation, an intermediate feature map f is formed, and then the intermediate feature map f is decomposed into a height attention tensor and a width attention tensor. Through two 1×1 convolutions, the dimensionality operation is performed and combined with the sigmoid function to reduce the complexity and computational cost of the model, and obtain the attention vector sum of height and width. Finally, the input feature map X is combined with the attention weights of height and width respectively. and multiplied to generate the final attention feature map Y. The C3-CA module improves the detection performance of China's offshore wind power and reduces the false alarm rate by integrating channel and location information and using an attention mechanism. It provides an effective solution to the problem of remote sensing target detection in complex SAR backgrounds with almost no increase in computational overhead.
为了在中国海上风电目标检测任务中提高速度和精确度,本发明引入RFB(ReceptiveFieldBlock)模块。多分支卷积层是RFB模块的核心组成部分,巧妙地采用瓶颈结构,通过1x1步幅为2的卷积层降低通道特征的维度,然后引入nxn的卷积层。为了有效控制参数数量、增加非线性层的深度,模块采用2个3x3卷积来替代5x5卷积,同时在直连层中引入不带激活函数的1x1卷积层。此外,RFB模块通过膨胀卷积技术控制每个分支的偏心度,模拟感受野大小和偏心度之间的比率,最终串联并进行1×1卷积,生成感受野的空间阵列。保持较少的参数量的同时获得更高分辨率的特征表示。融合不同卷积核大小和膨胀因子的多个分支进一步提升了特征提取能力,为模块的性能提升提供了强有力的支持。我们将RFB模块整合到轻量级YOLOv5s的主干网络中,可以有效地提升特征表示的质量以及精确地衡量感受野大小与偏心度之间的关系,从而更具区分性和鲁棒性。与传统的深度加深主干网络相比,RFB模块引入了更为灵活的特征提取机制,在保证精确度的同时实现了任务的加速。In order to improve the speed and accuracy in China's offshore wind power target detection tasks, the present invention introduces the RFB (ReceptiveFieldBlock) module. The multi-branch convolution layer is the core component of the RFB module. It cleverly uses a bottleneck structure to reduce the dimension of channel features through a 1x1 convolution layer with a stride of 2, and then introduces an nxn convolution layer. In order to effectively control the number of parameters and increase the depth of the nonlinear layer, the module uses two 3x3 convolutions to replace the 5x5 convolution, and introduces a 1x1 convolution layer without an activation function in the direct connection layer. In addition, the RFB module controls the eccentricity of each branch through dilated convolution technology, simulating the ratio between receptive field size and eccentricity, and finally concatenates and performs 1×1 convolution to generate a spatial array of receptive fields. Obtain higher resolution feature representation while maintaining a smaller number of parameters. Fusion of multiple branches with different convolution kernel sizes and expansion factors further improves feature extraction capabilities and provides strong support for module performance improvement. We integrate the RFB module into the lightweight YOLOv5s backbone network, which can effectively improve the quality of feature representation and accurately measure the relationship between receptive field size and eccentricity, making it more discriminative and robust. Compared with the traditional deep deepening backbone network, the RFB module introduces a more flexible feature extraction mechanism, which accelerates tasks while ensuring accuracy.
在本发明的另一些实施例中,为了深入验证所提出的YOLOv5s-CR模型的性能,以时间序列的方式,对2015年至2022年中国海上风电的SAR影像数据集进行了逐年测试。所选用的测试数据集包括不同的海域、多样下垫面条件、不同的风机类型、水深和离岸距离,以及未经过海陆分割的复杂陆地和海洋背景的SAR影像。我们特别关注了江苏潮滩典型海上风电场,以时间序列为基准展示了自2015年至2022年各个年份海上风电的识别可视化结果,具体呈现在图3中(a)-(h)之中,红色框内代表海上风电,(a)–(h)分别代表2015年-2022年的检测样本。结果表明,模型都能几乎无遗漏地识别每个年份的中国海上风电目标,且从未出现将相邻海上风电目标误判为同一目标,或将单一中国海上风电误判为两个目标的情况。此外,在未经过海陆分割的复杂背景下,模型仍然能够精准地识别出中国海上风电。实验结果全面验证了我们的模型在不同时间SAR影像的泛化能力,从而进一步证明了其具有高可信度的特点。YOLOv5s-CR模型在时序实验中持续表现出色,充分证明了其在时相泛化性方面的卓越表现,并突显了鲁棒性和可持续优化的潜能。In other embodiments of the present invention, in order to deeply verify the performance of the proposed YOLOv5s-CR model, the SAR image data set of China's offshore wind power from 2015 to 2022 was tested year by year in a time series manner. The selected test data sets include SAR images of different sea areas, diverse underlying surface conditions, different wind turbine types, water depths and offshore distances, as well as complex land and ocean backgrounds that have not been separated by sea and land. We paid special attention to the typical offshore wind farms on the tidal flats of Jiangsu, and used the time series as the benchmark to display the identification and visualization results of offshore wind power in each year from 2015 to 2022. The details are shown in (a)-(h) in Figure 3. The red box represents offshore wind power, and (a)–(h) represent the test samples from 2015 to 2022 respectively. The results show that the model can identify China's offshore wind power targets in every year with almost no omissions, and has never misjudged adjacent offshore wind power targets as the same target, or misjudged a single Chinese offshore wind power target as two targets. . In addition, the model can still accurately identify China's offshore wind power in a complex background that has not been separated from land and sea. The experimental results comprehensively verify the generalization ability of our model on SAR images at different times, further proving its high reliability. The YOLOv5s-CR model continues to perform well in time series experiments, fully proving its excellent performance in temporal generalization and highlighting the potential for robustness and sustainable optimization.
为了深入研究YOLOv5s-CR模型的地理泛化性,我们扩展了模型的应用范围,将其用于全球其他国家的典型海上风电场识别检测,包括英国、德国、丹麦、法国、比利时等多个国家。通过在不同地理环境下的海上风电场SAR影像数据上进行模型识别,验证模型在全球范围内的稳定性和广泛适应性。In order to further study the geographical generalization of the YOLOv5s-CR model, we expanded the application scope of the model and used it for identification and detection of typical offshore wind farms in other countries around the world, including the United Kingdom, Germany, Denmark, France, Belgium and other countries. . By conducting model identification on SAR image data of offshore wind farms in different geographical environments, the stability and broad adaptability of the model on a global scale are verified.
我们从每个国家的海上风电场数据中精选出代表性的SAR影像样本制作成不同尺度的测试集。为了更全面地考察模型的适用性,这些影像均未经过海陆分割。我们将YOLOv5s-CR模型应用于这些选定的影像数据,以自动化的方式进行海上风电目标检测,其中部分检测结果如图4(a)-(n)所示。经过对识别结果进行分析,我们观察到YOLOv5s-CR模型能够准确识别大部分海上风电。然而也存在一些漏检的情况(图4(d)),陆地和海洋同时存在的复杂影像中更容易出现错检、漏检的情况,陆地上的建筑物(图4(l))、海洋中的堤坝(图4(m))干扰了模型检测。这种现象可能与影像中存在的噪声干扰以及风电场与周围环境的复杂交互作用有关。综合对全球其他国家的海上风电场数据集进行模型测试和分析,我们认为YOLOv5s-CR模型在不同地理环境下都能表现出较高的识别精度和性能稳定性,具有较高的地理泛化性和广泛适用性。We selected representative SAR image samples from offshore wind farm data in each country to create test sets of different scales. In order to examine the applicability of the model more comprehensively, these images were not segmented into land and sea. We apply the YOLOv5s-CR model to these selected image data to detect offshore wind power targets in an automated manner. Some of the detection results are shown in Figure 4(a)-(n). After analyzing the identification results, we observed that the YOLOv5s-CR model can accurately identify most offshore wind power. However, there are also some missed detections (Figure 4(d)). In complex images where land and ocean exist at the same time, false detections and missed detections are more likely to occur. Buildings on land (Figure 4(l)), oceans The embankment in (Fig. 4(m)) interferes with the model detection. This phenomenon may be related to the noise interference present in the image and the complex interaction between the wind farm and the surrounding environment. Based on comprehensive model testing and analysis of offshore wind farm data sets from other countries around the world, we believe that the YOLOv5s-CR model can show high recognition accuracy and performance stability in different geographical environments, and has high geographical generalization. and broad applicability.
如图6所示,第二方面,本发明实施例提供一种基于yolo算法的海上风电识别系统,包括样本集制作模块100、初始模型构建模块200、模型训练与验证模块300以及海上风电识别模块400,其中:As shown in Figure 6, in the second aspect, the embodiment of the present invention provides an offshore wind power identification system based on the yolo algorithm, including a sample set production module 100, an initial model construction module 200, a model training and verification module 300, and an offshore wind power identification module. 400, of which:
样本集制作模块100,用于采集多个地区的海上风电场遥感图像,并制作海上风电样本集;The sample set production module 100 is used to collect remote sensing images of offshore wind farms in multiple regions and create an offshore wind power sample set;
初始模型构建模块200,用于基于YOLOv5算法、CA注意力机制和RFB多分支卷积模块构建初始海上风电识别模型;The initial model building module 200 is used to build an initial offshore wind power identification model based on the YOLOv5 algorithm, CA attention mechanism and RFB multi-branch convolution module;
模型训练与验证模块300,用于利用海上风电样本集对初始海上风电识别模型进行训练和验证,以得到目标海上风电识别模型;The model training and verification module 300 is used to train and verify the initial offshore wind power identification model using the offshore wind power sample set to obtain the target offshore wind power identification model;
海上风电识别模块400,用于获取实时海上风电影像,利用目标海上风电识别模型对海上风电目标进行识别。The offshore wind power identification module 400 is used to obtain real-time offshore wind power images and identify offshore wind power targets using a target offshore wind power identification model.
本系统通过样本集制作模块100、初始模型构建模块200、模型训练与验证模块300以及海上风电识别模块400等多个模块,构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明创建了一个适用于中国海上风电目标识别的深度学习训练数据集,并基于YOLO算法中的YOLOv5算法进行优化改进,引入CA注意力机制和RFB多分支卷积模块,构建有效的海上风电识别模型(YOLOv5s-CR模型),加入坐标注意力CA模块提高模型的定位能力,融合上下文增强模块RFB增加感受野减少在大尺度复杂海表背景下模型的错检率,可以实现中国海上风电的快速检测和精确定位。基于构建的海上风电识别模型对海上风电进行准确稳定的检测,有效提升了检测效率以及检测精准度。本发明构建的海上风电识别模型时相泛化和地理泛化性能好。This system uses multiple modules such as the sample set production module 100, the initial model construction module 200, the model training and verification module 300, and the offshore wind power identification module 400 to construct an offshore wind power identification model to accurately and stably detect offshore wind power, effectively improving the Detection efficiency and detection accuracy. This invention creates a deep learning training data set suitable for China's offshore wind power target recognition, optimizes and improves it based on the YOLOv5 algorithm in the YOLO algorithm, introduces the CA attention mechanism and RFB multi-branch convolution module, and constructs effective offshore wind power recognition Model (YOLOv5s-CR model), the coordinate attention CA module is added to improve the positioning ability of the model, and the context enhancement module RFB is integrated to increase the receptive field and reduce the error detection rate of the model in the large-scale complex sea surface background, which can realize the rapid development of China's offshore wind power detection and precise positioning. Based on the constructed offshore wind power identification model, offshore wind power is accurately and stably detected, effectively improving detection efficiency and detection accuracy. The offshore wind power identification model constructed by the present invention has good temporal generalization and geographical generalization performance.
如图7所示,第三方面,本申请实施例提供一种电子设备,其包括存储器101,用于存储一个或多个程序;处理器102。当一个或多个程序被处理器102执行时,实现如上述第一方面中任一项的方法。As shown in FIG. 7 , in a third aspect, an embodiment of the present application provides an electronic device, which includes a memory 101 for storing one or more programs; and a processor 102 . When one or more programs are executed by the processor 102, the method as in any one of the above first aspects is implemented.
还包括通信接口103,该存储器101、处理器102和通信接口103相互之间直接或间接地电性连接,以实现数据的传输或交互。例如,这些元件相互之间可通过一条或多条通讯总线或信号线实现电性连接。存储器101可用于存储软件程序及模块,处理器102通过执行存储在存储器101内的软件程序及模块,从而执行各种功能应用以及数据处理。该通信接口103可用于与其他节点设备进行信令或数据的通信。It also includes a communication interface 103. The memory 101, the processor 102 and the communication interface 103 are directly or indirectly electrically connected to each other to realize data transmission or interaction. For example, these components may be electrically connected to each other through one or more communication buses or signal lines. The memory 101 can be used to store software programs and modules. The processor 102 executes the software programs and modules stored in the memory 101 to perform various functional applications and data processing. The communication interface 103 can be used to communicate signaling or data with other node devices.
其中,存储器101可以是但不限于,随机存取存储器(Random Access Memory,RAM),只读存储器(Read Only Memory,ROM),可编程只读存储器(Programmable Read-OnlyMemory,PROM),可擦除只读存储器(Erasable Programmable Read-Only Memory,EPROM),电可擦除只读存储器(Electric Erasable Programmable Read-Only Memory,EEPROM)等。The memory 101 may be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable memory. Read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.
处理器102可以是一种集成电路芯片,具有信号处理能力。该处理器102可以是通用处理器,包括中央处理器(Central Processing Unit,CPU)、网络处理器(NetworkProcessor,NP)等;还可以是数字信号处理器(Digital Signal Processing,DSP)、专用集成电路(Application Specific Integrated Circuit,ASIC)、现场可编程门阵列(Field-Programmable Gate Array,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。The processor 102 may be an integrated circuit chip with signal processing capabilities. The processor 102 can be a general-purpose processor, including a central processing unit (CPU), a network processor (Network Processor, NP), etc.; it can also be a digital signal processor (Digital Signal Processing, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.
在本申请所提供的实施例中,应该理解到,所揭露的方法及系统,也可以通过其它的方式实现。以上所描述的方法及系统实施例仅仅是示意性的,例如,附图中的流程图和框图显示了根据本申请的多个实施例的方法及系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上,流程图或框图中的每个方框可以代表一个模块、程序段或代码的一部分,所述模块、程序段或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意,在有些作为替换的实现方式中,方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如,两个连续的方框实际上可以基本并行地执行,它们有时也可以按相反的顺序执行,这依所涉及的功能而定。也要注意的是,框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合,可以用执行规定的功能或动作的专用的基于硬件的系统来实现,或者可以用专用硬件与计算机指令的组合来实现。In the embodiments provided in this application, it should be understood that the disclosed methods and systems can also be implemented in other ways. The method and system embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show possible implementation systems of the methods and systems, methods and computer program products according to multiple embodiments of the present application. Architecture, functionality and operations. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more components for implementing the specified logical function(s). Executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two consecutive blocks may actually execute substantially in parallel, or they may sometimes execute in the reverse order, depending on the functionality involved. It will also be noted that each block of the block diagram and/or flowchart illustration, and combinations of blocks in the block diagram and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts. , or can be implemented using a combination of specialized hardware and computer instructions.
另外,在本申请各个实施例中的各功能模块可以集成在一起形成一个独立的部分,也可以是各个模块单独存在,也可以两个或两个以上模块集成形成一个独立的部分。In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, each module can exist alone, or two or more modules can be integrated to form an independent part.
第四方面,本申请实施例提供一种计算机可读存储介质,其上存储有计算机程序,该计算机程序被处理器102执行时实现如上述第一方面中任一项的方法。所述功能如果以软件功能模块的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM,Read-Only Memory)、随机存取存储器(RAM,Random AccessMemory)、磁碟或者光盘等各种可以存储程序代码的介质。In a fourth aspect, embodiments of the present application provide a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor 102, the method in any one of the above-mentioned first aspects is implemented. If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other various media that can store program code.
以上仅为本发明的优选实施例而已,并不用于限制本发明,对于本领域的技术人员来说,本发明可以有各种更改和变化。凡在本发明的精神和原则之内,所作的任何修改、等同替换、改进等,均应包含在本发明的保护范围之内。The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.
对于本领域技术人员而言,显然本申请不限于上述示范性实施例的细节,而且在不背离本申请的精神或基本特征的情况下,能够以其它的具体形式实现本申请。因此,无论从哪一点来看,均应将实施例看作是示范性的,而且是非限制性的,本申请的范围由所附权利要求而不是上述说明限定,因此旨在将落在权利要求的等同要件的含义和范围内的所有变化囊括在本申请内。不应将权利要求中的任何附图标记视为限制所涉及的权利要求。It is obvious to those skilled in the art that the present application is not limited to the details of the above-described exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be regarded as illustrative and non-restrictive from any point of view, and the scope of the application is defined by the appended claims rather than the above description, and it is therefore intended that all claims falling within the claims All changes within the meaning and scope of the equivalent elements are included in this application. Any reference signs in the claims shall not be construed as limiting the claim in question.
Claims (9)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311697617.9A CN117671504A (en) | 2023-12-11 | 2023-12-11 | An offshore wind power identification method and system based on yolo algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311697617.9A CN117671504A (en) | 2023-12-11 | 2023-12-11 | An offshore wind power identification method and system based on yolo algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117671504A true CN117671504A (en) | 2024-03-08 |
Family
ID=90069586
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311697617.9A Pending CN117671504A (en) | 2023-12-11 | 2023-12-11 | An offshore wind power identification method and system based on yolo algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117671504A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118279751A (en) * | 2024-05-31 | 2024-07-02 | 华东师范大学 | Offshore wind turbine installed capacity calculation method and system based on Sentinel-2 image and Unet model |
CN119479006A (en) * | 2024-09-24 | 2025-02-18 | 国家海洋环境监测中心 | A real-time AI identification and positioning method and system for birds around offshore wind farms |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115880642A (en) * | 2022-12-29 | 2023-03-31 | 中国华能集团清洁能源技术研究院有限公司 | Marine target proximity degree early warning method and system based on machine vision |
CN117152625A (en) * | 2023-08-07 | 2023-12-01 | 西安电子科技大学 | A method, system, equipment and medium for remote sensing small target recognition based on CoordConv and YOLOv5 |
-
2023
- 2023-12-11 CN CN202311697617.9A patent/CN117671504A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115880642A (en) * | 2022-12-29 | 2023-03-31 | 中国华能集团清洁能源技术研究院有限公司 | Marine target proximity degree early warning method and system based on machine vision |
CN117152625A (en) * | 2023-08-07 | 2023-12-01 | 西安电子科技大学 | A method, system, equipment and medium for remote sensing small target recognition based on CoordConv and YOLOv5 |
Non-Patent Citations (4)
Title |
---|
WENXIAO ZHAO 等: ""CRAS-YOLO:A Novel multi-category vessel detection and classification model based on YOLOv5s algorithm"", 《IEEE》, 19 February 2023 (2023-02-19) * |
XINGSHUAI YANG 等: ""Detection of river floating garbage based on improved YOLOV5"", 《MATHEMATICS》, 20 November 2022 (2022-11-20) * |
YINGCHAO ZHANG 等: ""converge of coordinate attention boosted yolov5 model and quantum dot labeled fluorescent biosensing for rapid detection of the poultry disease"", 《COMPUTER ANDELECTRONICS IN AGRICULTURE》, 15 February 2023 (2023-02-15) * |
朱昭云 等: ""基于YOLOv5s的海上风电设施检测与预警评估"", 《计算机技术与发展》, 1 April 2023 (2023-04-01) * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118279751A (en) * | 2024-05-31 | 2024-07-02 | 华东师范大学 | Offshore wind turbine installed capacity calculation method and system based on Sentinel-2 image and Unet model |
CN119479006A (en) * | 2024-09-24 | 2025-02-18 | 国家海洋环境监测中心 | A real-time AI identification and positioning method and system for birds around offshore wind farms |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110705457B (en) | Remote sensing image building change detection method | |
CN111898507B (en) | Deep learning method for predicting earth surface coverage category of label-free remote sensing image | |
CN112287807B (en) | A road extraction method from remote sensing images based on multi-branch pyramid neural network | |
Sellars et al. | Computational Earth science: Big data transformed into insight | |
CN117671504A (en) | An offshore wind power identification method and system based on yolo algorithm | |
CN111598174A (en) | Training method, image analysis method and system of image feature classification model | |
CN113610070A (en) | Landslide disaster identification method based on multi-source data fusion | |
CN116861347B (en) | A method for calculating magnetic anomaly data based on deep learning model | |
CN113096129B (en) | Method and device for detecting cloud cover in hyperspectral satellite image | |
CN116400426B (en) | Electromagnetic method-based data survey system | |
CN114119621A (en) | SAR remote sensing image water area segmentation method based on depth coding and decoding fusion network | |
Hu et al. | A study of automatic recognition and localization of pipeline for ground penetrating radar based on deep learning | |
CN117671480B (en) | Automatic landslide identification method, system and computer equipment based on visual large model | |
CN110751077A (en) | An optical remote sensing image ship detection method based on component matching and distance constraints | |
CN112084860A (en) | Target object detection method and device and thermal power plant detection method and device | |
CN112633213A (en) | Zhang Heng satellite lightning whistle sound wave detection method and system based on YOLO neural network | |
CN116561536A (en) | Landslide hidden danger identification method, terminal equipment and medium | |
Han et al. | Targets mask U-Net for wind turbines detection in remote sensing images | |
Saravanan et al. | Automated evaluation of degradation in stone heritage structures utilizing deep vision in synthetic and real-time environments | |
CN117409329B (en) | Method and system for reducing false alarm rate of underground cavity detection by three-dimensional ground penetrating radar | |
Li et al. | Three-dimensional dynamic monitoring of crevasses based on deep learning and surface elevation reconstruction methods | |
CN118134870A (en) | A method and system for detecting a directional marine vessel | |
Jun1a et al. | A review of ground camera-based computer vision techniques for flood management | |
Sokolowsky et al. | tobac v1. 5: Introducing fast 3D tracking, splits and mergers, and other enhancements for identifying and analysing meteorological phenomena | |
Mohammad et al. | Updating of GIS maps with Change Detection of Buildings using Deep Learning techniques |
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 |