3D target detection method based on single-vehicle multi-sensor information fusion

• Main Authors: FENG SHAOFEI, ZHENG HAIFENG, ZHENG QIYU, FENG XINXIN
• Format: Patent
• Language: Chinese; English
• Published: 04.06.2024
• Subjects: ANALOGOUS ARRANGEMENTS USING OTHER WAVES; CALCULATING; COMPUTING; COUNTING; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION ORRERADIATION OF RADIO WAVES; MEASURING; PHYSICS; RADIO DIRECTION-FINDING; RADIO NAVIGATION; TESTING

The invention provides a 3D target detection method based on single-vehicle multi-sensor information fusion. The method comprises the following steps: S1, extracting an ROI in a target scene by using a traditional point cloud 3D target detection algorithm; step S2, utilizing the covariance between the central position of each ROI and the position of the laser radar to guide the cutting of the false point cloud by utilizing the ROI generated in the step S1, and obtaining a false point cloud ROI area; s3, performing multi-level and multi-scale feature extraction on the cut false point cloud ROI region by using a fractional order Gabor convolutional network; s4, fusing the features of the original point cloud ROI region and the features of the false point cloud ROI region by using a fractional Fourier transform intersection attention mechanism to obtain a final target ROI feature; and S5, generating a class confidence coefficient and a bounding box by using the final target ROI feature. For the false point cloud