Research On The Method Of Automatic Extraction Of Toponymic Annotation In Multilingual Maps

Extracting toponymic annotation from existing overseas scanned map and the electronic map published in the network is an effective means of collecting overseas data in the construction of overseas geographic information database.However,many of the current collection work of geographical name relies on artificial progress,so the realization of extracting the toponymic annotation from maps is of great significance.At present,the text extraction methods often works for specific languages,which are difficult to be applied to the other language systems.In addition,the geographical names often cross the graphic symbols of other geographical elements,especially in complex background areas,such as stains and shadows.In view of the above difficulties,this thesis use the support vector machine classification method in the CIEL*a*b* color space to achieve image segmentation,and the automatic extraction of toponymic annotation in multilingual maps is realized by extraction,combination and selection of the text line.The main results are as follows:(1)Aiming at the problem that color is difficult to be segmented in complex background,an image segmentation method based on support vector machine in CIEL*a*b* color space is proposed.The color segmentation method can be adapted to the color segmentation of toponymic annotation.In the sample collection,a method based on prior knowledge and artificial acquisition is proposed to improve the efficiency of sample collection.(2)On the basis of analyzing the common features of toponymic annotation in multilingual maps,this thesis proposes a method to extract multilingual names based on the combination of text lines.The text lines extracted from the expanded image are combined into the text area,and then the text area are screened and combined to achieve the extraction of the toponymic annotation.According to the different characteristics of different names in different languages,the expansion coefficient,the threshold length of text lines and the minimum number of text line combinations are further refined.As a result,this method can realize the extraction of multilingual toponymic annotations.(3)According to the specific characteristics of the scanned map and Google Maps,the extraction strategy of toponymic annotation is designed.And the extraction of toponymic annotation in multilingual map is realized.(4)The design and implementation of the system for the extraction of place names are given,which verifies the validity and feasibility of the method.