Real-time face detection in digital video-based on Viola-Jones supported by convolutional neural networks

Face detection is a critical function of security (secure witness face in the video) who appear in a scene and are frequently captured by the camera. Recognition of people from their faces in images has recently piqued the scientific community, partly due to application concerns, but also for the difficulty this characterizes for the algorithms of artificial vision. The idea for this research stems from a broad interest in courtroom witness face detection. The goal of this work is to detect and track the face of a witness in court. In this work, a Viola-Jones method is used to extract human faces and then a particular transformation is applied to crop the image. Witness and non-witness images are classified using convolutional neural networks (CNN). The Kanade-Lucas-Tomasi (KLT) algorithm was utilized to track the witness face using trained features. In this model, the two methods were combined in one model to take the advantage of each method in terms of speed and reduce the amount of space required to implement CNN and detection accuracy. After the test, the results of the proposed model showed that it was 99.5% percent accurate when executed in real-time and with adequate lighting.