Intelligent workshop scene target lightweight semantic segmentation method

Abstract

The invention provides an intelligent workshop scene target lightweight semantic segmentation method. According to the method, a lightweight semantic segmentation network fusing two-way average pooling and a three-branch attention mechanism is provided for the problem of balance between the semantic recognition precision and the real-time performance of a workshop target. An encoder of the network takes a lightweight MobileNet v2 network as a reference so as to realize high real-time performance of segmentation; a two-way average pooling module is constructed in a decoder, lost feature information in an encoder is accurately repaired through a contextual information fusion method, richer semantic information and more accurate target position information are extracted, and high-precision semantic segmentation is achieved. A three-branch attention mechanism module is constructed in a decoder, pixel-level multiplication is performed on original features by using information of different scales, the features of different scales are represented in a vector form, and the multi-scale problem of semantic segmentation is efficiently solved.

Description

Technical field [0001] The present invention belongs to the field of computer vision, deep learning, specifically relates to a smart workshop scene target lightweight semantic segmentation method. Background [0002] The intelligent workshop is the most core execution unit of the intelligent factory, and promoting the construction of the intelligent workshop is a key step to achieve intelligent manufacturing. Semantic level perception recognition of workshop scene targets is the basis for realizing workshop intelligence, such as workshop intelligent security, mobile robot intelligent navigation tasks first need to semantically perceive and identify the workshop scene targets, that is, to identify the type, shape, and posture of the target, and then make reasoning decisions based on the recognition results. Due to the complexity of smart workshop scenarios and goals, the semantic level recognition of smart workshop goals faces the following two challenges: [0003] 1. Balance bet...

AI Technical Summary

Problems solved by technology

Some early semantic segmentation networks based on fully convolutional networks (FCN) mainly ensure segmentation accuracy through complex network structures and numerous parameters, and lack real-time segmentation.

The emergence of lightweight networks such as BiseNet has improved the real-time performance of semantic segmentation, but the reduction of network structure and parameter volume has reduced the segmentation accuracy to a certain extent.

Images