Object detection method and apparatus for object detection
AssigneeApollo Intelligent Driving Technology (Beijing) Co., Ltd.
InventorsYuanfan Xie, Zishu Yang, Liang Wang
Life of this patent
Estimating the term…
Abstract
Embodiments of the present disclosure disclose an object detection method, an object detection apparatus, a related device, a storage medium and a vehicle. The method includes the following. Scene point cloud data is inputted into a point cloud feature extraction model pretrained to obtain a point cloud descriptive feature of the scene. Scene image data is input into an image feature extraction model pretrained to obtain an image descriptive feature of the scene. The point cloud descriptive feature and the image descriptive feature of each target region in the scene are input into an object identifying model to obtain a detection result of an object included in each target region.
Claims15
An object detection method, comprising:
- inputting scene point cloud data of a scene into a point cloud feature extraction model pretrained to obtain a point cloud description feature of the scene, and inputting scene image data of the scene into an image feature extraction model pretrained to obtain an image descriptive feature of the scene into an image feature extraction model pretrained to obtain an image descriptive feature of the scene;
- and
- inputting the point cloud descriptive feature and the image descriptive feature of each target region in the scene into an object identifying model pretrained, to obtain a detection result of an object comprised in each target region;
- wherein inputting the point cloud descriptive feature and the image descriptive feature of each target region in the scene into the object identifying model pretrained comprises:
- projecting each point cloud target region onto a scene image according to parameters of a laser radar and parameters of an imager, to determine a respective image target region associated with each point cloud target region;
- and
- inputting the point cloud descriptive feature of each point cloud target region and the image descriptive feature of the respective image target region associated with each point cloud target region into the object identifying model.
An electronic device, comprising:
- one or more processors;
- and
- a memory, configured to store one or more programs that when executed by the one or more processors cause the one or more processors to input scene point cloud data of a scene into a point cloud feature extraction model pretrained to obtain a point cloud descriptive feature of the scene, and input scene image data of the scene into an image feature extraction model pretrained to obtain an image descriptive feature of the scene;
- and
- input the point cloud descriptive feature and the image descriptive feature of each target region in the scene into an object identifying model pretrained, to obtain a detection result of an object comprised in each target region;
- wherein the one or more processors are caused to:
- project each point cloud target region onto a scene image according to parameters of a laser radar and parameters of an imager, to determine a respective image target region associated with each point cloud target region;
- and
- input the point cloud descriptive feature of each point cloud target region and the image descriptive feature of the respective image target region associated with each point cloud target region into the object identifying model.
A non-transitory computer readable storage medium, having computer programs stored thereon that when executed by a processor cause the processor to perform the object detection method, the object detection method comprises:
- inputting scene point cloud data of a scene into a point cloud feature extraction model pretrained to obtain a point cloud descriptive feature of the scene, and inputting scene image data of the scene into an image feature extraction model pretrained to obtain an image descriptive feature of the scene;
- and
- inputting the point cloud descriptive feature and the image descriptive feature of each target region in the scene into an object identifying model pretrained, to obtain a detection result of an object comprised in each target region;
- wherein inputting the point cloud descriptive feature and the image descriptive feature of each target region in the scene into the object identifying model pretrained comprises:
- projecting each point cloud target region onto a scene image according to parameters of a laser radar and parameters of an imager, to determine a respective image target region associated with each point cloud target region;
- and
- inputting the point cloud descriptive feature of each point cloud target region and the image descriptive feature of the respective image target region associated with each point cloud target region into the object identifying model.
Dependencies are inferred from the claim text.