Automatic white balance method for FCN human skin segmentation correction camera

The invention provides an automatic white balance method for an FCN human skin segmentation correction camera. The method comprises the following steps: S1, collecting human skin segmentation data; s2, designing a convolutional neural network which meets the condition that the last layer of the convolutional neural network is a convolutional non-full connection layer; s3, designing a training strategy; s4, training a human skin model; s5, calibrating a white balance color temperature curve of the camera; and S6, white balance parameter correction. Before automatic white balance, image global information is considered, a full convolutional neural network is utilized to perform human body skin segmentation and image pixel-by-pixel classification, a human body skin area and a non-human body skin area are distinguished, image information is counted, skin area information is not considered when white balance parameters are weighted, the white balance parameters are more accurate, and the image quality is improved.