The Design Of Intelligent Controllable Camera And Research Of Relative Key Algorithm

Due to the existence of global instability as well as people's sense of security, surveillance technology gets unprecedented development.Video surveillance system is the most important security applications.As the expansion of the monitoring area, we not only need to layout a lot of cameras,but also need a lot of people work around. This way is inefficient,costly,and fallible due to mistakes made by people.In this context,intelligent video processing technology becomes one of the most important topics in video processing,artificial intelligence,pattern recognition.This paper discusses the design of controllable cameras,and research of relative intelligent algorithms,which have practical value.This paper focuses on the camera hardware design and the implementation of intelligent algorithms in PC.Taking the image quality into account,we selected the CCD camera sensor.Then we compare a variety of CCD camera system solutions, eventually using SONY's SS-11X system solution,which you can use high solutions CCD as well as low solution CCD,so it can be used widely.The solution is relatively mature,and finally we prove it works well.We implement three algorithms:auto-focus,motion detection and depth from defocus.Auto-focus technology is one of the most important technologies in camera, manual focus is too difficult for the normal people,and it also takes much more time. We can get different pictures by changing the camera set using step motor,then the auto-focus algorithm compares the pictures and decides the moving direction of step motor,then we can get the focusing position at last.Motion detection is kind of the basis of intelligent analysis.If we can detect moving targets accurately,we can implement intrusion detection,trip wire detection,motion tracking and so on.In this paper,we combine the Gaussian mixture modeling(GMM) and Mean Shift algorithm, and get a better result under complex environment.We know that camera discards the depth information,but the depth information,sometimes,has useful applications.DFF (Depth From Focus) and DFD(Depth From Defocus) are two ways to obtain the depth information.The former needs a lot of pictures while the later requires only two pictures.So it is not surprise we choose the later one.It works well when the scene has rich texture.