Effective key parameter determination for an automatic approach to land cover classification based on multispectral...

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Environmental Science

Background:

• Land cover classification from satellite data is crucial for research but traditional methods are labor-intensive and subjective.
• Jiang et al. developed a semi-automatic method using prior maps and current imagery, but key parameters required manual selection.
• This limitation hinders efficiency and introduces potential bias in land cover classification.

Purpose of the Study:

• To develop an automated approach for selecting key parameters in land cover classification.
• To enhance the automation and reduce subjectivity of existing semi-automatic methods.
• To improve the accuracy and efficiency of land cover classification using remote sensing data.

Main Methods:

• Proposed a three-part interdependent approach: pure-pixel training-sample selection, automated key parameter determination, and an optimal combination model.
• Utilized overall accuracy, Kappa Coefficients (KC), and Time-Consuming (TC) metrics for automatic parameter selection, avoiding subjective test-decisions.
• Applied a portfolio optimization model for determining key parameters, enhancing automation.

Main Results:

• Successfully automated the selection of two key parameters in the land cover classification method.
• The automated parameter selection process avoided subjective bias, leading to more objective results.
• Demonstrated increased automation of Jiang et al.'s classification method through a case study in Weichang District, China.

Conclusions:

• The proposed method significantly increases the degree of automation for land cover classification.
• Automated parameter selection using accuracy metrics and optimization models offers a robust alternative to manual selection.
• This enhanced methodology holds potential for wide-ranging scientific applications in land cover analysis.