Translator Disclaimer
12 March 2019 Improved N-FINDR based on object retrieval initialization for finding endmembers
Author Affiliations +
In remote sensing images, the common existence of mixed pixels affects the classification of remote sensing images and the targets detection. The N-FINDR algorithm is a widely used endmember extraction algorithm. The algorithm is full- automatic. it has no parameters and has a good selection effect. However, in the volume calculation, the random selection of the initial value of the spectral endmember limits the efficiency of the method. The dimensionality reduction of the data loses some details; The automatic target retrieval algorithm uses the method of orthogonal subspace projection to determine the feature pixels. The algorithm runs faster and is more efficient, but it does not consider the angle between the pixels. Therefore, an improved N-FINDR algorithm based on object retrieval initialization is proposed in this paper. The improved algorithm can select the initial endmember, and then use the new volume calculation formula to solve the maximum volume of the simplex, so as to obtain the endmember, which makes up for the shortcomings of the two methods. Finally, simulation data and real airport diagram are used to verify the algorithm. The improved N-FINDR algorithm is superior to the traditional internal maximum volume algorithm in terms of running time and extraction accuracy.
© (2019) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Hao Wang, Chuntong Liu, Hongcai Li, Zhenxin He, and Shixin Ma "Improved N-FINDR based on object retrieval initialization for finding endmembers", Proc. SPIE 11023, Fifth Symposium on Novel Optoelectronic Detection Technology and Application, 1102330 (12 March 2019);

Back to Top