Abstarct: Analysis, motions and behavior of human have always been interesting subjects in dif-ferent areas of science. The aim of designing such systems is to get better understanding on the behavior of human using video images sequence. Human behavior analysis can be assessed from various points of views such as initialization, tracking as well as pose re-covery and recognition. Tracking is by far one of the most significant steps in behavior analysis conducted using network covering a wide range of areas. To reduce the amount of cost in such cases, cameras with disjoint views are often used. However, features like appearance of person, environmental illumination, location of person resect to planted cameras as well as camera’s angle of view involved in such systems make the results un-reliable along with considerable errors. In this thesis, an approach was introduced to be used in a network of cameras with dis-joint views. The tracking process was performed in identification and tracking steps in the view of one cameras and matching with other cameras. After doing necessary pre-processing steps and elimination of useless parts, three global features including motion, geometrical model and appearance were evaluated aiming to present their weakness and strength points. Since methods working baxsed on motions and geometrical models were not able to provide reliable results, color histogram was used as the best feature extract-ed from persons appearances. Color histogram is a suitable method for such systems as it is a very simple but effective approach. However, it cannot be properly used once brightness and size of the people going through the cameras are changed. Thus, cumula-tive brightness transform function was used to reduce the variation of brightness of those persons going through the camera’s view. To do tracking process, according to the relative distance of various parts of human body, the body was partitioned into three portions comprising head, torso and bottom part. Since this method cannot be affected by people size, the body can be very well portioned once people get close to the cameras or vice versa. Another method used in the present study to extract feature was nonlinear fuzzy robust principle component analysis (nfrPCA). Those features extracted by this method cannot be always used in entrance and exit of people as they present distinct results in various environmental conditions. Thus, fuzzy system was used to discriminate the area under consideration to three parts of close, middle and far distance to the cam-eras. To evaluate the efficiency of proposed system, experiments were conducted on a set of data extracted from Italian experiments. The results indicate a better performance for proposed method in YCbCr color space compared to other color space used in previ-ous studies.