Abstract
Hyperspectral data consists of hundreds of bands. The high dimension of hyperspectral data is a challenge for the researcher to design efficiency and accuracy image segmentation algorithm. In this paper, a new approach to extraction features and independence of hyperspectral image proposes using Discriminant independent component analysis (DICA) and multilevel thresholding techniques based on Otsu problems for image segmentation are introduced. Image segmentation is initial process of image analysis and recognition. Otsu’s problem with multilevel thresholding is solved in each band from hyperspectral data that has been reduced dimensionality using DICA. The main purpose using multilevel thresholding is to get an optimal threshold that maximizes variance between classes. The swarm optimization approach used in this study is Darwinian particle swarm optimization (DPSO). The result of experiment showed that DPSO was better compared to other swarm optimization approaches. DPSO shows a statistically significant increase, both from CPU time processing and fitness value. DPSO technique is able to find optimal threshold with greater variance between classes and smaller search times compared to particle swarm optimization (PSO).
Highlights
Multispectral and hyperspectral are two groups in spectral imaging
The data dimension reduction method used in this paper is discriminant independent component analysis (DICA)
From this it can be seen that Darwinian particle swarm optimization (DPSO) can find the optimal threshold value with a faster time compared to particle swarm optimization (PSO). From this it can be concluded that the DPSO method is more recommended to be used as an image segmentation method, especially in high-dimensional images such as hyperspectral images
Summary
Multispectral and hyperspectral are two groups in spectral imaging. Multispectral and Hyperspectral are two groups in spectral imaging. Most of the methods used are designed for gray-level analysis, color image or multispectral image not for hyperspectral images Applying this method to HSI will be a big challenge because of the increasing dimensions of data. Feature selection or feature extraction techniques are often displayed as pre-processing on hyperspectral data analysis [1] Such processing is often carried out in multispectral imagery to improve class to separate or to eliminate certain types of noise. A number of multispectral and hyperspectral image segmentation methods have been introduced in some literature. The PSO consists of a number of particles that collectively move in the search space (for example the pixels of the image) to find global optimality (e.g. maximizing inter-class variance from the distribution of intensity levels in a given image). While the segmentation method used is swarm optimization approach namely particle swarm optimization (PSO) and Darwinian particle swarm optimization (DPSO)
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: IOP Conference Series: Materials Science and Engineering
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
