US 12,249,158 B2
Object detection method
Yun-Ling Chang, Lugong (TW); Yi-Feng Su, Lugong (TW); and Ying-Ren Chen, Lugong (TW)
Assigned to AUTOMOTIVE RESEARCH & TESTING CENTER, Lugong (TW)
Filed by Automotive Research & Testing Center, Lugong (TW)
Filed on Dec. 15, 2021, as Appl. No. 17/644,537.
Prior Publication US 2023/0186642 A1, Jun. 15, 2023
Int. Cl. G06V 20/58 (2022.01); G01S 17/89 (2020.01); G06T 7/246 (2017.01); G06T 7/277 (2017.01); G06T 7/33 (2017.01)
CPC G06V 20/58 (2022.01) [G01S 17/89 (2013.01); G06T 7/248 (2017.01); G06T 7/277 (2017.01); G06T 7/337 (2017.01); G06T 2207/20212 (2013.01)] 9 Claims
OG exemplary drawing
 
1. An object detection method that is to be performed by a computing device which is in communication with a camera device and a lidar module that are positioned on a carrier device, the camera device being configured to continuously capture a series of images showing a scene which is in a vicinity of the carrier device and which includes at least one object, the lidar module being configured to continuously scan the scene in the vicinity of the carrier device to generate in series multiple pieces of point cloud data representing the scene, the computing device being configured to receive the images from the camera device and to receive the pieces of point cloud data from the lidar module, the object detection method comprising following steps that are to be performed with respect to each of the pieces of point cloud data after the piece of point cloud data is received from the lidar module:
selecting a first to-be-combined image from among the images that have been received from the camera device, wherein the first to-be-combined image corresponds in time to the piece of point cloud data;
selecting a second to-be-combined image from among the images that have been received from the camera device, wherein the second to-be-combined image is an Nth image before the first to-be-combined image in a time order of image capture, and N is an integer that is no less than three;
combining the first to-be-combined image and the second to-be-combined image to generate a combined image;
generating a result image by incorporating the piece of point cloud data into the combined image; and
inputting the result image into a trained machine learning model in order to determine a class to which each object shown in the result image belongs.