Object-level edge detection method based on deep residual network

Abstract

The invention relates to an object-level edge detection method based on a deep residual network. The method comprises the following four parts: (1) establishing a neural network: taking a deep residual error network as a basic network structure, replacing a convolution residual error structure in the basic network structure with a mixed cavity convolution residual error block, and adding a multi-scale feature enhancement module and a pyramid multi-scale feature fusion module; (2) performing training optimization on the neural network through data set enhancement and hyper-parameter setting; (3) completing the training of the neural network; and (4) detecting the general image by using the trained neural network, and outputting an object-level edge detection image. Compared with a traditional edge detection operator and an existing neural network edge detection method, the method has a better edge detection effect, a detection result is closer to a real value, and noise is lower.

Description

technical field [0001] The invention belongs to the field of intelligent vision processing, and in particular relates to an object-level edge detection method based on a deep residual network. Background technique [0002] Object-oriented edge detection aims to propose visually salient edges and object boundaries from natural images, and is a fundamental problem in the field of intelligent vision processing. At present, most detection methods still stay in the stage of traditional edge detection operators. These operators mainly use low-level visual features such as color gradient, brightness and texture of images to realize edge detection. Due to the high real-time performance, traditional detection methods are still widely used in many fields, but their limitations are also obvious. It is difficult to detect edges at the semantic level through low-level features, and it is difficult to be directly applied to automatic driving, 3D reconstruction, and intelligent image compr...

AI Technical Summary

Problems solved by technology

[0007] In order to solve the problems of low edge resolution and high noise in existing object-level edge detection methods, the present invention provides an object-level edge detection method based on a multi-scale residual network, which specifically includes the following steps

Images