Scene segmentation method and system

Abstract

The invention belongs to the technical field of scene segmentation, and particularly relates to a scene segmentation method and system. The scene segmentation method comprises the step of carrying outscene segmentation on each frame of image in a scene video by utilizing a lightweight network, wherein the lightweight network comprises a plurality of convolutional networks, the network architecture is preset, and the training process comprises the following steps: inputting an image in a training set with normalized pixel points into a lightweight network with a known architecture; adopting convolution for feature extraction in the encoding stage, and combining convolution and bilinear interpolation in the decoding stage to restore information of an input image, and obtianing a feature mapwith semantic information; learning different types of feature maps by adopting a plurality of groups of convolution kernels, and fusing the feature maps in different periods; and optimizing each convolutional network in the lightweight network through a cross entropy loss function.

Description

technical field [0001] The invention belongs to the technical field of scene segmentation, and in particular relates to a scene segmentation method and system. Background technique [0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art. [0003] Scene segmentation, also known as semantic segmentation, is to mark each category in the picture with a color, so as to obtain a segmented picture with semantic category information. The picture is composed of many pixels. Different categories have different RGB values ​​of the pixels. Combined with the surrounding pixels, we can determine which category a certain area belongs to. By predicting the category to which each pixel belongs, the computer can Get segmented images with semantic information. At present, scene segmentation is applied to the segmentation of remote sensing satellite images, farmland segmentation, medical image p...

AI Technical Summary

Problems solved by technology

[0004] The inventors found that the current scene segmentation has the problems of overlapping objects, dim light, and overly complex road scenes, which affect the accuracy of scene segmentation. In addition, the neural network for scene segmentation requires a large amount of memory and memory while achieving high precision. longer time to train the network

Images