Cross-country road surface extraction method based on a three-dimensional laser radar

Abstract

The invention provides a cross-country road surface extraction method based on a three-dimensional laser radar, and belongs to the technical field of image recognition. The method comprises the stepsof superposing continuous multi-frame three-dimensional laser point cloud data acquired by a laser radar, and projecting the superposed three-dimensional laser point cloud data to a top view to obtaina pavement elevation map; performing feature extraction on the pavement elevation map by using a deep convolutional neural network to obtain a traffic cost map of the pavement elevation map; discretizing the passing cost map to obtain a passable area label, an obstacle area label and a fuzzy area label; and according to the passable area label, the obstacle area label and the fuzzy area label, carrying out image visualization on the discretized passable cost map to obtain a passable cross-country road surface area map. The method is not influenced by illumination and weather, and a passable cross-country road surface can be extracted; the feature extraction method based on deep learning can adaptively learn environmental features, and is higher in adaptability in different scenes; According to the provided automatic data labeling scheme, the requirement for manually labeling samples is effectively reduced.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a three-dimensional laser radar-based off-road road surface extraction method capable of accurately identifying passable areas on the off-road road surface. Background technique [0002] In recent years, unmanned driving technology has developed rapidly, and road area extraction is one of the key technologies, which is the prerequisite for realizing safe and reliable autonomous driving of unmanned vehicles. At present, the relatively mature algorithms and application technologies are mainly aimed at the urban structured road environment, and there are few studies on off-road roads with complex structures and environments. In off-road scenarios, there are often a large number of ambiguous areas at the road boundary: for example, muddy soil or areas covered by vegetation, these areas are generally physically passable, but the cost of passage is very high, which is not str...

AI Technical Summary

Problems solved by technology

The disadvantage they all have in common is that training a deep neural network requires a large number of manually labeled samples

In the absence of public data sets for off-road road surface extraction, the human cost of labeling a large number of training samples is very high

Images