Feature Extraction of Microscopic Image and Construction of Target Optimization Model

Rui Liu,Xingbao Gao

2019-01-01

Abstract:With the popularity of image processing technology in the world, many companies, universities and research institutions in the world are currently working on microscopic technology. However, the cost of other analytical methods is high and the system structure is complex. These shortcomings limit the popularity and promotion. In order to study the feature extraction and target optimization of microscopic images, this paper first extracts the texture feature quantity based on gray level co-occurrence matrix and the statistical feature quantity based on gray level distribution to form the initial feature set. According to the texture characteristics and brightness difference of the microscopic image, five texture features such as energy, entropy, moment of inertia, local stationarity and maximum probability are extracted based on the gray level co-occurrence matrix, supplemented by the brightness ratio and mean value based on the gray distribution statistics. Six characteristics, such as mean square error, skewness, consistency, and kurtosis, form the 11-dimensional initial feature quantity set, and draw the distribution map of the characteristic parameters. Multi-thread parallel acceleration is performed by using CUDA for the part with large calculation amount such as solving control point coefficient and resampling. The paper proposes a high-dimensional multi-target membrane algorithm based on surface film guidance. The SMG-MOMA target optimization algorithm proves that the superficial membrane guiding strategy plays a large role in the evolution of the population. The simulation results on the scalable multi-objective optimization test show that the proposed algorithm can effectively deal with high-dimensional multi-objective optimization problems.