Study On Extraction And Recognition Of Fading Words Based On Spectral Imaging

China has a large number of cultural relics,including murals,calligraphy and painting.They are Chinese traditional cultural relics,which can record the spiritual life and important historical events.Among them,the handwriting can most intuitively understand the information on cultural relics.Due to the long history of Chinese civilization,handwriting was found as early as in the Paleolithic age.From ancient bone inscriptions to modern simplified characters,handwriting have always been important symbols for human beings to record events and express emotions.The ancients left us many precious cultural relics.In addition,most of these cultural relics are described or decorated by handwriting.Exploring the information of handwriting can better restore the real history and help us understand the culture of different periods.However,due to the influence of natural and human factors,the surface of some cultural relics has faded,which affected the appearance and made the handwriting illegible.Therefore,in the protection of cultural relics,it is particularly important to explore the information of handwriting.Most of the traditional ways are visual identification,relying on the experience of cultural relics workers,but it is difficult to distinguish faded words by eyes.Moreover,in the process of extracting text,once there is a mistake,it may cause irreparable loss.Therefore,we need to use modern science and technology to better serve the work of cultural relics protection.Hyperspectral technology has many characteristics,such as noncontact,spectrum integration,wide spectrum range,etc.It can be used for deep preservation and analysis of cultural relics.Using the unique advantages of hyperspectral,we can capture the information that the human eye can’t observe,which is of great significance for the interpretation and exploration of handwriting.There are three hyperspectral data studied in this paper,which are hyperspectral images of a Chinese painting,a stone carving and the bottom of a tomb.These three cultural relics have been damaged to varying degrees,which makes the information on the surface of cultural relics difficult to identify.After data preprocessing,the fading information is enhanced and extracted in this paper.Then the convolution neural network is used to recognize the handwriting,which provides reference for the handwriting recognition.The specific work is as follows:(1)The enhancement index of black handwriting is constructed for extraction.The spectrum of handwriting and background are selected in different wavelength positions to establish the mean spectrum curve and analyze its spectral characteristics.The blue and near infrared bands are selected as sensitive bands used to construct the three black handwriting enhancement indexes,respectively.They are compared with the common Hyperspectral Feature extraction methods.Then,the best threshold value of handwriting and background area is determined by density slice method,and the handwriting extraction image is obtained.Finally,three faded images are taken as samples for experimental verification.The results show that the extraction method based on the handwriting enhancement index can effectively enhance the faded text information.In addition,this method has achieved good results in three data,which provides a scientific means for the extraction of faded text information of cultural relics.(2)Convolution neural network(CNN)is used to recognize the handwriting of cultural relics.Handwriting extraction can be looked as a problem of multi-Classification by CNN.In this paper,we introduce the structure and characteristics of LeNet-5 CNN.Based on the experimental data set and experimental environment,we adjust the structure and parameters of the network,realize the construction of the network,and train the model.Finally,the convolution neural network is applied to the real cultural relic data to get the handwriting recognition results,and the experimental results are discussed and analyzed.The results show that it is feasible to use CNN to recognize the faded handwriting of cultural relics,which provides a possibility for researchers.