Electrostrictive Strain Detection Of Dielectric Elastomer Based On Image Analysis

Dielectric Elastomer (DE) is a new type of intelligent polymer materials, which is developing rapidly in recent years. It can produce deformation quickly to response the external voltage. Once the external voltage was removed, it will restore again. This feature provides DE a huge potential application in the field of aerospace and intelligent bionic. But the electrostrictive strain of DE is a micro deformation under high voltage, which makes it very difficult to take a non-contact man-made measurement.In order to solve such problem, this paper proposes a method of measurement based on image analysis for the electrostrictive strain of DE. In this method, we use a digital camera to record the photographs of measured DE materials on various experiment conditions, and then a series of measures will be adopted for the further photo processing. Firstly, an appropriate image enhancement algorithm would be used to highlight the detail information of these photographs. Secondly, we improve the image segmentation algorithm to segment the target deformation area from the total photographs accurately. The most important business is designing reasonable algorithms to locate and exact the scale which was photographed into the picture as an object of reference, and then we calibrate the actual physical size that an individual pixel represents in the current image. Finally, we calculate the measured material's accurate electrostrictive strain data according to the above steps.The main works done in this paper are as follows:(1) Researching on image enhancement algorithm to solve the problem that noises and non-uniform illumination in the photos will affect further processing. This paper chose HSV color model which conforms well to human subjective vision to decompose the image in order to get the luminance information. We introduce the morphological top-hat algorithm which has a significant inhibiting effect on non-uniform illumination, combining with Nonsubsampled Contourlet Transform (NSCT). The experimental results show that this method can enrich the linear detailed features along with restraining the influence of noises and non-uniform illumination.(2) Improving the image segmentation algorithm. The traditional FCM algorithm used in image segmentation just considers the gray information of pixels ignoring the spatial relationship between them. What's more, it also has many deficiencies such as poor de-noising and complex computation. In order to solve those problems, a new FCM image segmentation method based on Markov random is presented. The proposed method improves the clustering objective function based on FCM algorithm by using Markov random field to describe the neighborhood relationship between image pixels. Such improvements contribute to the new segmentation algorithm a fast operational speed and high efficiency in de-noising.(3) Some algorithms are designed and optimized to calibrate the actual physical size that an individual pixel represents by using the scale which was photographed into the picture as an object of reference. On the aspect of locating and extracting the scale, we use edge detection and linear filtering extract it accurately according to the characteristic that the target scale area is rich in small graduation lines. In image tilt correction, an improved image tilt correction algorithm based on coordinate axis rotation and vertical projection is proposed to make this work easy and low-calculation. Through scan the binary graduation lines we just processed by using progressive scanning line by line, this paper gives the algorithm and formulae calibrate the actual physical size that an individual pixel represents according to the condition of these scanning line's jump points.