| With the rapid development of our country’s economy in recent years,the transaction volume of the jadeite market has also increased significantly,and people have higher demand for jadeite valuation.The traditional jadeite valuation method is mainly completed by offline testing agencies in 1-2 weeks,and at the same time,the appraisers need to complete the jadeite valuation based on their own experience in the market.In addition,although our country has promulgated jadeite grading standards,due to its poor operability and the particularity of jadeite structure,the jadeite market has not established a unified standard.Therefore,proposes a method of using image retrieval to complete the valuation of jadeite to be evaluated,and designs a jadeite valuation system based on image retrieval,users can upload jade images through the We Chat public account to get the price evaluation report.The main work content is as follows:(1)Constructing the Jadeite Jewelry Information Database.Mainly select four large-scale e-commerce platforms : including the price,listing time,and classification of each piece of jadeite.And Constructing the jadeite jewelry information database.(2)Constructing the jade images retrieval library.Preprocessing and data enhancement were performed on the crawled images of jade ornaments,and all images have been annotated.At the same time,through the perceptual hash technology,feature extraction is performed on each image to obtain the hash sequence representing the image,and the Jade Image Retrieval Library is established.By randomly selecting 1600 images of image retrieval experiments,the similar jadeite result set reaching the threshold is output.The average retrieval time per image is 1.08 seconds,and the accuracy rate reaches 89.9%.This method greatly improves the efficiency of jadeite evaluation,reduces the subjectivity of jadeite evaluation,and provides a basis of accurate evaluation of jadeite.(3)The realization of the jade evaluation method based on main detect.In view of the possible problems in the jadeite images uploaded by users,the SSD network is used to complete the main detect of the jadeite images to be estimated,and the optimal jadeite subjects in the images to be detected are extracted.And 20% of the jade image retrieval library was selected as the jadeite to be detected for the experiment.The final experimental results showed that the SSD network performed well on the emerald image retrieval library,and the m AP reached 94.1%.At the same time,the detection speed is also expected to be within the acceptable range of users.(4)The Realization of Jade Valuation System.Developed a complete jadeite price evaluation system based on Py Charm through the We Chat public account interface,providing users with the function of completing jadeite price estimation by uploading jadeite images.The user completes the upload of the jade image to be evaluated through the We Chat official account,and the system automatically completes image preprocessing,main detect and similar image retrieval,and finds a similar image set that reaches the threshold;Analyze the price range of the Top 10 pieces of jadeite in the similar jadeite result set,the price distribution range of all jadeite in the result set,and generate final price report.The report provides users with the price range and market situation of jadeite to be estimated in the market.Aiming at the problems existing in the current jade evaluation market,the jade valuation system based on image retrieval solves the pain points of low efficiency,lack of recognized standards,and low convenience,and it has certain practical value. |
PDF Full Text Request