Miner detection method based on conditional random field

For low detection ratio, accuracy and efficiency of existing object detection methods of video image, a miner detection method based on conditional random field was proposed. The method mainly includes two sections, namely model foundation and recognition of miner detection. In the model foundation stage, characteristics of histogram of oriented gradient of sample images are extracted, whose dimensionalities are reduced by principal component analysis. Then interested regions are signed by conditional random field to calibrate training sample, and parameters of conditional random field model are trained. In the recognition stage, characteristics of histogram of oriented gradient of detected image are extracted, whose dimensionalities are reduced. The trained conditional random field model is used to calibrate each sub window of the detected image by local binary pattern method, so as to get region where miner is. The experimental results show that the method can detect miner in image correctly.