Within the domain of ferrite ring production, the vast majority of available data consists of repetitive, defect-free samples, while data containing defects are considerably rare. To tackle this issue, a deep learning-based defect detection strategy tailored for small sample scenarios has been introduced. The approach begins with the precise extraction of defective areas using image masking technology, followed by the processing of these defect images through regional enhancement transformation techniques to create new instances of defects. These newly generated defect images are employed to build an offline defect sample library. During the model training phase, defective samples are randomly selected from this library and merged with the foreground of normal samples to generate new defect images for training. The deep learning network trained using this methodology demonstrates enhanced capability in distinguishing between defective and defect-free data, with experimental results showing an increase of 3.87% in the Area Under the Curve (AUC) metric. This approach not only addresses the challenge of scarce defect data but also significantly improves the performance of defect detection in small sample environments.
With the continuous development of industry, the discharge of pollutants in marine areas has attracted widespread attention. As a rapidly developing city, Hong Kong faces significant challenges regarding the large-scale discharge of pollutants into its marine areas. Therefore, studying the inversion model of chlorophyll-a concentration in the coastal waters of Hong Kong is an important topic. Traditional BP neural networks are widely used for their ability. However, it is challenging to scientifically and effectively determine the parameters, and the stability of the model is insufficient. In this study, based on the measured data of chlorophyll-a concentration in the coastal waters of Hong Kong and Landsat 8 OLI data of the Hong Kong offshore area, the traditional BP neural network is optimized and improved. Three neural network models are established: BP neural network model, BP neural network optimized by genetic algorithm (GA-BP), and BP neural network model improved by whale optimization algorithm (WOA-BP). By comparing the prediction accuracy , the results show that the WOA-BP neural network model achieves high accuracy and good stability in the inversion of chlorophyll-a concentration, with an average relative error of 12.91%, which is lower than that of the traditional BP neural network model.
This study aims to investigate the wetland vegetation phenology in the Yellow River Delta using remote sensing data from the Sentinel-2 satellite and four classification algorithms: Support Vector Machine (SVM), Classification And Regression Tree (CART), Gradient Tree Boost (GTB), and Random Forest (RF). By constructing feature sets and analyzing time series curves, the phenological information of wetland vegetation, including Spartina alterniflora (SA), Suaeda salsa (SS), Phragmites australis (PA), and Willow (WW), was explored and the classification accuracy of different algorithms was compared. The results demonstrate distinct phenological characteristics of wetland vegetation in the Yellow River Delta, with RF algorithm showing excellent performance in accurately extracting large areas of SA and achieving good results in mixed vegetation areas. The invasion of SA poses a significant threat to native vegetation, gradually occupying their growth space. This study provides scientific decision support for the ecological restoration and conservation of wetland vegetation in the Yellow River Delta.
Hyperspectral remote sensing has been widely used in mineral identification using the particularly useful short-wave infrared (SWIR) wavelengths (1.0 to 2.5 μm). Current mineral mapping methods are easily limited by the sensor’s radiometric sensitivity and atmospheric effects. Therefore, a simple mineral mapping algorithm (SMMA) based on the combined application with multitype diagnostic SWIR absorption features for hyperspectral data is proposed. A total of nine absorption features are calculated, respectively, from the airborne visible/infrared imaging spectrometer data, the Hyperion hyperspectral data, and the ground reference spectra data collected from the United States Geological Survey (USGS) spectral library. Based on spectral analysis and statistics, a mineral mapping decision-tree model for the Cuprite mining district in Nevada, USA, is constructed. Then, the SMMA algorithm is used to perform mineral mapping experiments. The mineral map from the USGS (USGS map) in the Cuprite area is selected for validation purposes. Results showed that the SMMA algorithm is able to identify most minerals with high coincidence with USGS map results. Compared with Hyperion data (overall accuracy=74.54%), AVIRIS data showed overall better mineral mapping results (overall accuracy=94.82%) due to low signal-to-noise ratio and high spatial resolution.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.