Study Of Vision Inspection Technology Based On Uniform Illumination And Polarization Imaging

Computer vision measurement is widely applied in various fields that require precisionmeasurement such as modern industry manufacturing. In order to improve the contrast and thedefinition of detection images, simplify the task of image understanding and the difficulty ofvision algorithms, an indirect illumination pattern with diffuse reflection surface wasresearched in this paper, a design of diffuse reflection freeform surface for uniformillumination was proposed; Image fusion technology with polarization information and colorinformation to enhance the contrast of detection target was studied. The strategy of fusingoriginal image with polarization feature in color model directly was proposed.For the illumination requirement in image acquisition process: well multiple-wavelengthcolor mixing characteristic and illumination uniformity of near-field distance, an indirectillumination pattern with three-ring array of LEDs and reflection freeform surface is proposed.Based on the Lambertian characteristic of LEDs and ideal diffuse surface, a mathematicalmodel is developed. A set of differential equations are established according to the expectedenergy distribution on the target plane. Solving these equations numerically, the outline of thediffuse surface is obtained. By using the optical software to perform non-sequential raytracing, the simulation result shows that the irradiance uniformity of the detection plane isimproved significantly with high accepted tolerance and demonstrates that the proposedmethod is correct, effective and reliable.Polarization vision is usually a beneficial supplement to normal vision measurement, andis used to discriminate material properties. Image fusion technology is researched based onpolarization and color information. The paper takes printed circuit board as experimentalobject and the vision detection platform used for polarization imaging is built. Fusing originalimage with polarization degree and polarization phases in RGB and HIS color models basedon wavelet is attempted. According to the evaluation criterion of images, the contrast ofdetection target has been improved effectively and the true color fused image is more close tothe human being perception.