In a word, one needs to study the light source optimization choice and design method combination for different detection tasks and work environments.Moreover, the small proportions of the target defects relative to the entire picture in a micrograph, uneven surface illumination for high curvature surfaces and natural metal textures all contribute to make contrast of defect regions and background regions small, and image segmentation comparatively difficult. PCNN has been widely used in every field of image processing, such as denoising [17], segmentation [18], fusion [19] and feature extraction [20], but the elementary PCNN model framework is complex, and there are multiple undetermined parameters such as attenuation constants, amplification coefficients and connect coefficients.

Most parameters are configured by artificial tests, which affect PCNN image processing speed and make it difficult to implement automatic image processing. Richard [21] used a genetic algorithm for setting optimum PCNN parameters, yet the key of genetic algorithms is the accurate setting of parameters such as variation and cross operator, which, if not set properly, will destroy the developmental stability. Particle Swarm Optimization (PSO) is an efficient search strategy [22], which features quick convergence and requires less parameter settings. Chao [23] used a PSO to search for the best parameter value of a generalized diffusion coefficient function that was used for anisotropic diffusion defect detection in low-contrast surface images.

The PSO algorithm is used to automatically set PCNN optimization key parameters by fitness function of maxima between cluster variances, which carries out automatic PCNN image processing.The paper is organized as follows: Section 2 introduces the structure of the Brefeldin_A gyroscope pivot bearing dimension measurement and surface detection system; Section 3 presents task-oriented illumination system design methods; Section 4 presents self-adaptive parameter settings obtained Cilengitide by integrating the PSO algorithm and PCNN; Section 5 describes experimental results and comparisons. Finally, some conclusions and future development are illustrated in Section 6.2.?Detection System Design2.1. System FrameworkThe shape of a gyroscope pivot bearing is shown in Figure 1.