Target classification method based on multi-modal data characteristics

Abstract

The invention provides a target classification method based on multi-modal data characteristics. The method comprises the following steps: collecting laser radar point cloud data and RGB image data; acquiring a plurality of features of the laser radar point cloud data; obtaining a plurality of laser radar feature maps according to the plurality of features; performing up-sampling and densificationon the laser radar feature map; acquiring three-channel data of the laser radar feature map, and fusing the three-channel data with the RGB image data to form six-channel data; and training the six-channel data by adopting a deep learning network model to obtain a classification result. According to the invention, three features obtained from a laser radar are used as three-channel data and are fused with three-channel data of an RGB image to form six-channel data; six-channel data are trained by using a deep learning network model, a plurality of different probabilities can be obtained for each target after training, and a numerical value with the maximum probability is selected from the plurality of probabilities to serve as a final target classification result, so that the target can be detected more accurately, and the accuracy is higher.

Description

technical field [0001] The invention relates to the technical field of multi-sensor fusion and pattern recognition, in particular to a target classification method based on multi-modal data features. Background technique [0002] Object classification is one of the necessary technologies for the development of intelligent driving vehicles. In order to enhance the environmental awareness of intelligent driving vehicles, multiple types of sensors are often installed around the vehicle. Most of the current object classification methods are image-based computer vision methods. Although visual images contain the richest semantic information, traditional optical cameras are very sensitive to environmental lighting changes, and image calculations consume a lot of computer resources. More importantly, under highly dynamic driving conditions, visual odometry is not stable, and image-based methods cannot estimate object distance and distinguish overlapping objects on the road due to ...

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the problem of not being able to better detect and classify targets in the prior art, a target classification method based on multimodal data features is provided.

Images