Automatic online porn detection and tracking.
暂无分享,去创建一个
In this paper Hue Saturation, Value (HSV) modeling is used for segmenting human skin colours and for detection of pornographic images. The clustering of human skin colours in the HSV space is demonstrated. Using a mixed set of images of Caucasian, Asian and Africans, we show that the model can be used for pornographic image detection and recognition as an aid for online pornographic activity tracking. Introduction Within the past few years pornographic image exchange and reported cases of online pornographic activities leading to arrests of pedophiles has increased. It is estimated that close to 20% of office workers in the US download pornographic materials for more than 80% of all downloads. This trend is not unique to the US as the Internet has provided its users the stealth and anonymity that so many pornographic outlets have desired for decades. Law enforcement agencies in many countries are however grappling with the huge tasks of tracking child pornography often deployed on the Internet by pedophiles. Considering the seriousness and magnitude of the problem, technologies that aid law enforcement agencies to automatically detect and track online human exposure are essential. It is therefore necessary to provide technologies that can detect human skin colours the predominant content of all porn and pedophile images. Distinguishing between and accurate description of reasonable human body exposure and excessive exposure are contentious issues. Therefore to be reliable, the technology should be able to isolate skin-like pixels from other background colours and provide a reliable measure of how much skin content is in the photo and how much of it should be considered as pornography. Human skin detection technologies have largely been used in image processing and compression and one of the most promising approach involves the use of colour histograms and hue-saturation modeling of the human body colours. Recently we demonstrated the suitability of skin modeling with HSV technique for human face recognition using mobile phones [1, 2]. HSV Model The basic RGB colour model triplet (R,G,B) represents not only colour but also luminance. Luminance varies across the person’s face due to the ambient lighting. Therefore direct representation of the human skin colour with its RGB components is inefficient. However, the HSV colour model a nonlinear transformation of the RGB colour space is user-oriented and is based on artist’s notions of tint, shade, and tone. It has independent values for Hue, Saturation, and Value, corresponding, respectively to wavelength, excitation, and brightness. Figure 1 shows the HSV coordinate system as a hexacone. At the base of the hexacone is black with HSV = (0, 0, 0). Thus the facial HSV model of Africans is clustered around the origin of this coordinate system. Despite this across the African continent, modest variations occur which do not depart considerably from this value. Most colour pictures are recorded as (R, G, B) triplets. Given a colour defined by (R, G, B) where R, G, and B are normalized to 0.0 to 1.0, an equivalent (H, S, V) colour is determined by the following set of formulas. Figure 1: The HSV Colour Model Considering MAX to be the maximum of the (R, G, B) values, and MIN the minimum of those values the model is:
[1] Johnson I. Agbinya,et al. Face Recognition Programming on Mobile Handsets , 2005 .
[2] Richard Bowden,et al. Learning Non-linear Models of Shape and Motion , 1999 .
[3] Osamu Nakamura,et al. Human-face extraction using modified HSV color system and personal identification through facial image based on isodensity maps , 1995, Proceedings 1995 Canadian Conference on Electrical and Computer Engineering.