Three-dimensional object detection method based on view cone point cloud

Abstract

The invention discloses a three-dimensional object detection method based on a view cone point cloud. The RGB image and the point cloud image are collected and annotated, the annotated RGB image dataset and the point cloud image data set are obtained, and the data set is used as a training sample and a test sample. It mainly consists of three networks: cone point cloud extraction network based ontwo-dimensional object detection, three-dimensional object segmentation network based on Point net and three-dimensional bounding box regression network based on offset residuals. The three networksare trained in turn, and the training output of the previous network is used as the input of the next network. The super parameters of each neural network model are set, and the three-dimensional object detection network model is trained by Tensorflow. When the cost loss function is reduced to the ideal degree and the training reaches the required maximum iteration number, the training model is generated. The point cloud image data set is inputted into the trained model, and the 3D object detection result is output. The method of the invention is more perfect and more difficult than the existing three-dimensional object detection method.

Description

technical field [0001] The invention relates to the technical fields of pattern recognition, machine learning and computer vision, in particular to a three-dimensional object detection method based on cone point cloud. Background technique [0002] 3D object detection is an important research field of pattern recognition and computer vision, and it is also the core problem to solve many high-level vision tasks. It has a wide range of applications in people's daily life and industrial production, such as: intelligent video surveillance, unmanned driving, intelligent robots, intelligent transportation, and augmented reality. [0003] In the past few years, with the improvement of 2D object detection methods and the popularization of depth cameras, 3D object detection technology has also achieved rapid development. 3D object detection obtains the 3D information of the object by identifying and locating the 3D object, and outputs the 3D bounding box of the object to represent t...

AI Technical Summary

Problems solved by technology

In the existing research work, unmanned driving 3D object detection and indoor environment scene understanding 3D object detection, the degree of freedom of the 3D bounding box is only 7 (l, w, h, c x 、c y 、c z , yaw), because scholars only consider the vehicles and objects in the indoor environment only have the heading angle yaw, but no pitch angle and roll angle roll

Images