This paper presents experimental studies of coverage control for networked camera sensors. Regarding the issue, one of our previous works considered the scenario where the camera sensors with controllable orientations are distributed over the 3-D space to monitor 2-D environment. Then, we presented a visual coverage control scheme based on the gradient descent techniques on SO(3) in combination with an estimation process of the image density function describing the importance of each point over the image. In this paper, we build a testbed with four Pan-Tilt (PT) cameras and demonstrate the total process including the image acquisition, density estimation, gradient computation and camera physical motion. In particular, we show applicability of the presented scheme to a scenario of the human behavior monitoring. In addition, the density function is given in the form of a polynomial function in order to overcome the drawback of the above density estimation process that the magnitude of the motion at each pixel cannot be directly reflected to the resulting density. It is shown through experiments that the new density estimation method together with the new gradient outperforms the conventional method when the environment is rapidly changing.