Target object automatic segmentation method based on spectral imaging

• Main Authors: ZHANG JIN'EN, LI FEI, FANG ZHUOQUN, LEE SO KYU, LIU CUIWEI, SUN DANNING, WANG KE, WANG YAN, LI HONGLI
• Format: Patent
• Language: Chinese; English
• Published: 30.01.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention belongs to the technical field of hyperspectral image target segmentation, and particularly relates to an automatic target object segmentation method based on spectral imaging. The method comprises the steps that S1, a target object data set needing to be segmented in a Pascal VOC format is manufactured, and the data set is divided and marked as a training set, a verification set and a test set; s2, inputting the training set pictures into the trained YOLOv5 network model for training; s3, testing the trained YOLOv5 network by using the test set; s4, modifying the YOLOv5 network, adjusting parameters and the proportion of data in a training set, and performing model training again; s5, a detection model with the most accurate detection result is obtained through continuous testing and training adjustment; inputting all target pictures into the model for target type detection and judgment; according to the method, the segmentation speed is far higher than that of manual segmentation, inversion wo