Road image multi-scale edge detection model and method based on residual network

Abstract

The invention discloses a road image multi-scale edge detection model and method based on a residual network, and the model comprises a data set construction module which is used for constructing a data set for model training by using the images of various types of driving roads; a training sample loading module which is used for loading training samples in batches during training, cutting the images into the same size during loading, and inputting the images into a multi-scale edge detection network; and the multi-scale edge detection network which is used for helping different layers of thenetwork to learn and detect image edge features of different scales through mutual constraint of output results between different layers during training, and finally enabling the network to detect image edges from small scales to large scales from a shallow layer to a deep layer.

Description

technical field [0001] The invention relates to the technical field of computer image processing, in particular to a residual network-based multi-scale edge detection model and method for road images. Background technique [0002] At present, there are already many edge detection algorithms for images. For example, the traditional edge detection algorithms proposed earlier pay more attention to the intensity, color gradient and texture in the image, such as the Canny operator, but because some significant edges in the image are in the The change in the color gradient is not obvious, which leads to poor detection of edges using these methods in some cases. [0003] Learning-based edge detection algorithms use supervised models and manually labeled features, such as structured random forests to introduce structural information of images to detect edges of images, and methods for edge detection based on deep learning have also been proposed, using convolutional neural networks ...

AI Technical Summary

Problems solved by technology

However, these methods are not good at multi-scale detection of image edges, such as large-scale edges of roads and small-scale edges of plant branches and leaves.

Images