Author Email: firstname.lastname@example.org
Tampering detection techniques have been continuously studied with advancements in technologies. However, many existing alternatives cannot generate object tampering results, because of the hashes attached to the image shortfall the outline information. Object tampering detection is the process of locating the modifications in the structure of an object. In this paper, we presented a method of object tampering detection which will generate the results showing the tampered images of the object. Firstly, different images of the same object are taken from different angles. Furthermore, each image of the object will then be processed to localize the region of tampering in the object. The model estimates the parameters associated with an image such as shape hash, color hash, and hash vector. Finally, a classification system, Support Vector Machine (SVM) will classify the images as tampered or not tampered based on the input fed to the classification system. This method will be useful for the product manufactures to meet the growing demands of the product as tampered products lead to non-acceptability to the end user. This technique will reduce the manpower for quality check control of the products and will identify the tampered objects from the collection.
With the advancement in technologies and continuous diversified market needs, it becomes mandatory for the product manufacturers to meet the growing demands of the product with high performance in less delivery time. To meet this growing demand, a lot of control efforts are required. For a product to be effective, its quality is a major concern that can be evaluated in various parameters. Quality of the product also called as features can be measured in terms of internal and external factors. Internal quality factors include the raw material required for the manufacturing of the product whereas external quality factors concern with the finished goods. Hence to identify object tampering, we proposed a novel technique to identify the tampered regions in the product. This method will analyse different images of the object and will generate the results showing the
tampering in images of the object and its location where the object has tampered. This technique will reduce the manpower for quality check control of the products and will identify the tampered objects from the collection. The widespread use of technologies and the growing market needs, it becomes necessary for the manufacturers to deliver the products with good performance and with short delivery time. Some recent work and studies based on tampering detection and hashing is discussed.  Chi-Man Pun, Caiping Yan, and Xiao-Chen Yuan proposed a design of Image Alignment based Multi Region Matching (IAMRM) algorithm for object level tampering detection. In this scheme, an adaptive segmentation model is used to divide an image into regions on the basis of strong edges. Color based multi region matching along with color location based multi region matching was used to locate the tampered part in the image.
The scheme evaluates three aspects of object tampering detection: (1) database creation where analysis of shape hash, color hash and hash vector of the image is stored to decide whether the input image has the same content as the original one or not; (2) training of SVM is performed to record the number of inputs along with the bias value. The bias value less than 0.5 indicates more number of tampered images are recorded in the database and the bias value greater than 0.5 indicates more number of non-tampered images are recorded in a database. (3) Evaluation of SVM- classification of images are done which are stored in the database using SVM to label them as original or the tampered one which helps in the analysis of an input image. This localizes the region of tampering in the image of the object using the tamper map. The proposed approach can be used to detect any modification in the object content. Results show that the color hash and the shape hash is used to indicate or find the changes occurred in the image of the object. Also, here accuracy is evaluated in terms of number of images tested versus number of correct outputs of the tested images. In our future work, we will investigate how geometrical transformation affects the tempering detection of objects. In this paper, we presented an approach of using color hash and shape hash to detect any modification in the structure of an object. The modified values of the color hash and shape hash indicated a change in color texture and shape of the object respectively. The alteration in the structure of an object can, however, can be shown with histograms but using color hash and shape hash gave better accuracy than histograms. In Figure 9, the Binary Probability Tamper Map shows the regions of abrupt change in the structure of object after tampering. The region coloured in white shows the more concentrated area of tampering found in the image. The region coloured in black shows less concentrated regions of tampering. This method helps us to localize the region of tampering in object. Figure 10 shows the detected regions of image.
279 total views, 1 views today