Application Research Of Computer Vision Method In Construction And Acceptance Assessment Of Meteorological Observation Field

Meteorological observation service is the foundation of meteorological operation.Meteorological observing station is the site which possesses representation,accuracy and comparability for local meteorological sounding in the surface area.Therefore,site selection and construction of meteorological observing station is a very scientific and rigorous work.The work should not only meet the requirements of site selection,but also need real-time monitoring and long-term maintenance in the construction of meteorological observing station.Focus on monitoring the layout of meteorological observation devices is whether accord with the standard?And is the height of the barrier around the observing station whether to meet the requirements?In recent years,with the development and application of computer vision technology,it provides a new way for the acceptance evaluation of meteorological observing station's construction.According to the related requirements of meteorological observing station,we develop a software that is used to detect the height of of the barriers(such as houses and trees)around the observing station is whether in conformity with the specification.The main idea is as follows:Firstly,matching point pairs are extracted from the surrounding images of the segmented images,and the panoramic images are merged by the optimized image fusion algorithm.After obtaining the panoramic image,an improved interactive image segmentation algorithm based on image segmentation is used to segment the panoramic image to obtain the coordinate position of the highest barrier in the image.Finally,the height of the barrier in the image is estimated by the method of image measurement.The main research works are carried out as follows:(1)for image matching,an improved image matching algorithm based on depth convolution network is proposed.The operational principle and algorithm procedure of the proposed matching algorithm are discussed in detail,andthe algorithm is compared with the classical image matching algorithm,obtaining the relatively ideal comparison of experimental results,and the image matching of the collected meteorological observation images is implemented by the algorithm.(2)For image mosaic based on matching point,the algorithm procedure is optimized.This way could reduce the possibility of error transmission,the algorithm is applied to mosaic the meteorological station images which have been extracted matching points and obtain the panoramic image which meets the experimental requirement.(3)For image segmentation,the energy function of the graph cut algorithm is optimized.The image color and texture information are merged.The unit potential energy item of the energy function and the energy function are redefined according to the local and global information.The multi-scale iterative segmentation of the image is realized by using the two-dimensional discrete wavelet transform.By compared with the classical image segmentation algorithmthe ideal experimental results are obtained and the segmentation of the object in the panoramic image is implemented.(4)For barriers height measurement,two image measurement methods based on invisibility and crossover invariance are applied.Obtaining barrier target height using multiple reference objects and estimating the error range is compared with the field measurement data.The detection method proposed in the thesis has high accuracy,and can provide good decision making forthe acceptance evaluation of meteorological observing station's construction.