Pavement water body detection method

Abstract

The invention discloses a pavement water body detection method which comprises the following steps: acquiring a road scene image by using a visible light camera, zooming the road scene image to a specified size, and acquiring a mask containing water body position information in the acquired image by using a labeling method; constructing a conditional generative adversarial network combined with areflection attention unit, and training the conditional generative adversarial network combined with the reflection attention unit by utilizing the acquired road scene image and the marked mask; and zooming an image to be detected to a specified size, inputting the scaled image into the trained conditional generative adversarial network combined with the attention reflection unit, and obtaining abinary image which is output by a generator of the conditional generative adversarial network and is used for representing a water body detection result. According to the pavement water body detectionmethod, the surface water body area can be accurately and efficiently detected by utilizing the pavement image acquired by the camera, and the detection result has relatively high accuracy and recallrate, and the pavement water body detection method can be applied to tasks related to surface classification in the field of unmanned driving.

Description

technical field [0001] The invention relates to the technical fields of image segmentation and deep learning, and in particular to a road surface water body detection method. Background technique [0002] Surface classification is a key and important task in the field of unmanned driving. The water body on the road often means puddles, and it is difficult to estimate its depth information. pose unpredictable dangers. Due to the confusion caused by the reflection characteristics of road water bodies, traditional water body detection algorithms based on edge detection and texture detection are difficult to identify road water bodies well, and are prone to false detection or missed detection. At the same time, because the number and shape of road water bodies are random, the problem of road water body detection should be attributed to the problem of image segmentation. With the development of deep learning and artificial intelligence technology, there are countless examples o...

AI Technical Summary

Problems solved by technology

[0004] However, the missed detection rate of the above method is still high, and the effect is still affected by the reflection of the water surface. The training of the network is greatly affected by the uneven distribution of samples, and there is still a lot of room for improvement.

Images