Landform classification method based on trinocular visual system

Abstract

The invention discloses a landform classification method based on a trinocular visual system, which comprises the following steps firstly sampling the landform and outputting abundant 3D data by a trinocular three-dimensional visual system; then transmitting 3D data to a geometrical classifier, and in consideration of color characteristics of the landform, finally differentiating the landform through a color classifier. The landform classification method has functions of realizing simple and quick vehicle automatic guiding based on the trinocular visual system, preventing large-data-volume training, and realizing no requirement for manual landform marking. The trinocular visual system can be used as a supplement for a complicated vehicle automatic guiding system, and furthermore can be prepared to a simple object transport trolley automatic guiding system. The landform classification method has good market prospect.

Description

technical field [0001] The invention relates to a vehicle automatic guidance device, in particular to a terrain classification method based on a trinocular vision system. Background technique [0002] Vehicle automatic guidance is the core research part of vehicle automatic driving. The development of modern artificial intelligence technology makes vehicle automatic guidance technology more mature. Since Electronic launched the first automatic guided vehicle system and put it into industrial production, the automatic guidance of the vehicle has gradually entered the industrial and agricultural production and terrain detection, the transportation of goods in the warehouse, the automatic driving of the Mars rover and other applications have been familiar to everyone . At the same time, the technology of terrain detection using image recognition has gradually attracted attention, but there are still deficiencies in the design of classifiers. Traditional classifiers cannot aut...

AI Technical Summary

Problems solved by technology

In this way, the function of automatic guidance is not realized at all. The amount of terrain data that needs to be marked manually is large, and the universality of the terrain cannot be taken into account. The performance of terrain recognition depends on the terrain sample library. Due to the variety of terrain, the terrain sample library is difficult to cover all types. Terrain, the correct rate of terrain recognition is greatly affected by the training process of the classifier, basically all conventional terrain classifiers have such problems

Images