Skin injury picture segmentation method based on a deep network

Abstract

The invention relates to the field of artificial intelligence, in particular to a skin injury picture segmentation method based on a deep network, and the method comprises the steps: carrying out a segmentation task by using training data instead of manually extracting skin picture features, and learning deep convolutional features suitable for the segmentation task by using the training data; according to the method, preprocessing is very simple, and only normalization of picture pixel values is carried out; besides, compared with a preprocessing mode that TDLS and Jafari use a guide filter,the problem that illumination and contrast change greatly is solved, training data are enriched in a data enhancement mode, and a model learns optimal feature representation by himself / herself to carry out segmentation; according to the method, the index of the true positive rate exceeds that of an existing method, the operation time of the method on a GPU and a CPU is far shorter than that of anexisting model, and real-time skin image segmentation can be achieved; in addition, a fully-connected conditional random field is used as a post-processing method, low-level texture color features canbe effectively utilized, and segmentation of an edge area is sharpened.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, and more specifically, to a method for segmenting skin damage pictures based on a deep network. Background technique [0002] The current skin image segmentation can be divided into two categories according to the type of skin image used: methods based on dermoscopic images and methods based on images captured by ordinary cameras. Aiming at the segmentation problem of dermoscopic images, there have been many research works that can achieve good results. However, the relatively complicated and expensive acquisition of dermoscopic images will become the bottleneck of related technologies. Therefore, current segmentation techniques are more inclined to use skin pictures taken by ordinary cameras. With the improvement of the camera function of mobile devices such as mobile phones, it is easy to obtain high-definition skin pictures. Because these ordinary skin pictures are greatly dif...

AI Technical Summary

Problems solved by technology

[0004] The present invention does not manually extract skin picture features to perform segmentation tasks, but uses training data to learn depth convolution features suitable for segmentation tasks; the preprocessing of the present invention is very simple, just normalizing the pixel values ​​of the pictures; In addition, compared with TDLS and Jafari, which use the guided filter preprocessing method to solve the problem of large changes in illumination and contrast, the present invention enriches the training data through data enhancement, allowing the model to learn the optimal feature representation for segmentation; The invention surpasses the existing method on the index of true positive rate, and the running time on GPU and CPU is much lower than the existing model, and can achieve real-time skin image segmentation; the present invention also uses a fully connected As a post-processing method, conditional random fields can effectively utilize low-level texture color features and sharpen the segmentation of edge regions

Images