| Ancient villages are precious cultural heritage.From this perspective,the construction of digital archives is an important way to protect ancient villages.However,the wide distributions of many ancient villages make it difficult to collect ancient village digital resources.To solve this issue,crowdsourcing provides a more efficient solution to get the ancient village digital resources,which distributes the collection of ancient village digital resources especially ancient village images to the volunteers in the Internet.However,there are still some problems that need to be investigated underlying crowdsourcing.Due to the different skills and devices,some ancient village images taken by participants in crowdsourcing are distorted.In addition,some participants shoot multiple images for the same object with the same angle.As a result,some ancient village images are near-duplicate.In order to ensure the data quality of ancient village digital archives,ancient village images collected by crowdsourcing need to be screened efficiently and objectively.To deal with the above issues,this thesis proposes a logistic regression based screening model for ancient village images to filter out ancient village images with poor quality and nearduplicate.By this way,the proposed approach accelerates the construction of the ancient village digital archives and provides digital protection for ancient villages.First of all,the proposed approach predicts the quality of ancient village images by the edge grey gradient and the quality score calculated by NIQE.Secondly,the contents of ancient village images are described by their ancient village name,three class classification and color feature vector.Meanwhile,the similarity threshold used to judge whether the two images are near-duplicate is determined through experiments.Furthermore,the proposed approach calculates the number of near-duplicate images and relative quality score associated with near-duplicate images for each ancient village image.Moreover,a logistic regression based screening model for ancient village images is designed and realized in which maximum likelihood estimation algorithm is used to estimate the parameters of the proposed model.What’s more,to validate the efficiency,HosmerLemeshow goodness of fit statistic,classification table and model Chi-square statistic are used to measure the proposed model.Finally,this thesis applies the proposed model to the cloud service platform of ancient village.In order to verify the applicability and accuracy,the proposed model is compared with a screening model that only considers the quality of ancient village images and a SVM based screening model.The comparison results show that the proposed model outperforms the baseline models in accuracy. |
