Strain Gauge Image Feature Extraction And The Edge Detection

Automatic identification and installation of strain gauges can effectively improving productivity and saving labor. But the key technology is how to exact the feature, model and recognition. Mathematical morphology is similar to the human perception and the nervous system, and it also is a nonlinear theory for signal and image processing analysis, mouphology can estimate many geometrical structure information. And it has been widely applied in lots of field and been paid more and more attentions. In this paper, we use mathematical morphology as the basic theory and research tools to study and design the de-noising algorithm, edge detection and conner detection in the system of automatic installation of strain gauge technology.For strain gauge image processing, we study and analysis the advantages and disadvantages between the traditional and popular filtering method, then introduct the fuzzy concept and design a new de-desing algorithm based on fuzzy morphology, which would avoid damaging the image details, and in this algorithm wo use the small structural to protect the image details and reduce the algorithm complexity. We also study the de-noise algorithm in the frequence, so we proposed a new de-noiseing algorithm by combiating the the morphology and wavelet.For image edge detection and conner extraction, we study and analysis the traditional edge detector and morphological edge detector, and apply these algorithms to the strain gauge image, then we can get the advantages and disadvantages of the algorithm. Considering the actual situation, we need to suppress the noise and background to improve the precision, we introduce a new edge detector by using the probe characteristics of structural elements and the size is 5*5, and it only perform the operate once, which can enormously ruduce the algorithm complexity compared with the multi-scale structure elements, and suppress the noise and background. Besade studying edge algorithm, studying the conner extraction also is one part. Comparing and analysis the result of SUSAN etl.classic and morphology conner detectors to the strain gauge image, at last, a new algorithm was proposed based on multi-scale SE, but because there is less change between background and foreground, all the algorithm ware not well. Finally, for the actual type of the strain gauge image and regular place, in this paper, using the edge template which has been deteced ahead as the extracted feature, and using the template matching to recognize the strain gauge, which can reduce the search area, reduce the complexity, save match time and improve execution speed effectively, realize the real-time as far as possible.