RGB-D image object detection and semantic segmentation method based on deep convolution network

Abstract

The invention discloses an RGB-D image object detection and semantic segmentation method based on a deep convolution network, which belongs to the field of depth learning and machine vision. According to the method provided by the technical scheme of the invention, Faster-RCNN is used to replace the original slow RCNN; Faster-RCNN uses GPU, which is fast in the aspect of feature extracting, and at the same time generates a regional scheme in the network; the whole training process is training from end to end; FCN is used to carry out RGB-D image semantic segmentation; FCN uses a GPU and the deep convolution network to rapidly extract the deep features of an image; deconvolution is used to fuse deep features and shallow features of the image convolution; and the local semantic information of the image is integrated into the global semantic information.

Description

technical field [0001] The invention belongs to the field of deep learning and machine vision, and in particular relates to an object detection and semantic segmentation method including RGB-D images, which has a very wide range of applications in real scenes, such as detecting and tracking pedestrians in surveillance videos, UAV navigation, automatic driving, etc. Background technique [0002] Object detection and semantic segmentation are two important research fields of computer vision. Object detection is mainly used to detect the position and category of objects in the image. There are two main tasks in object detection. One is to find out the region scheme of the object (Region Proposals), the regional scheme is a pre-selected frame, which represents the approximate position of an object in the image; the second is to classify the objects in the pre-selected frame. The problem solved by semantic segmentation is to assign the correct label to each pixel of the image. S...

AI Technical Summary

Problems solved by technology

In the process of semantic segmentation, the superpixel-based depth features (globe-centered pose) and geometric features (size, shape) use support vector machines to predict the category labels of superpixels, but this method is very slow and uses multi-scale combination The method of grouping generates a regional plan is a very slow process, using slow and redundant RCNN, training is divided into multiple pipeline stages, and calculating the characteristics of superpixels is a complex and slow process

Images