Spectral Data Detection System Based On Machine Vision Technology Raindrops

With regard to our country present stage, raindrop spectrum data examination will be mainly finished by the artificial way, and working procedure is numerous and diverse,and the worker have heavy intensity, and the examination cycle is long and need a lot of time, and the efficiency is low, the omission or wrong examines of probability is very difficult to be controlled in the process of examination,at the same time, the method exist in subjectivity. Along with computer technology and kinds of visual algorithm fast development, so the machine vision of application already more and more widespread. The machine vision technology is applied in the middle of the raindrop spectrum data examination process, because this method has non-contact, the speed quick, the better stability and many superiority may enhance the examination of efficiency.The topic come from the practical application, we has developed to the raindrop spectrum data examination research work, specially, the raindrop spectrum data imagery processing is the topic research primary coverage. This method mainly include how to carry on the effective sample area in the raindrop spectrum data image, how to identificate the raindrop pellet from the background, how to divide the adhesion raindrop pellet and the particle size of statistics content, and the paper has conducted the relative research to the correlation technology. In the image pre-processing process, the paper proposes that union-level transform complex algorithms and interpolation combined each other in order to get a valid sampling of data area in the raindrop images, at the same time, this method can effectively solve the distortion that the problem generated by the image transformation and improve identification accuracy; using nonlinear filtering method, which can filter small raindrops by sputtering particles in order to improve accuracy and processing speed of raindrop feature extraction; the use of Otsu method and threshold evaluation function,which can improve the accuracy of identification from the background of raindrop; the integrated use of distance transform, morphological reconstruction and watershed segmentation, which separates the adhesion of raindrop Particles; the paper propose that curve fitting, regression analysis and morphological together work to gain real raindrop particles and stain-pixel correspondence relative. we start to classificate raindrop of size and statistical on the basis of this method.Experimental results show that,using machine vision processing raindrop images, it be able to quickly and accurately achieve the raindrop particles segmentation, statistical, and this method has high efficient, easy automated features, simultaneously, there are good practical. so this method provides a new solution for the raindrop of data detection.