Infrared blind pixel detection method and system based on image segmentation
By combining the improved Deeplabv3+ network and MobileNetV3 backbone network with the CA attention mechanism, the accuracy problem of infrared blind pixel detection in complex scenes is solved, achieving efficient and lightweight infrared blind pixel detection, and improving the model's adaptability and detection accuracy.
CN122391823APending Publication Date: 2026-07-14JIANGSU NORTH LAKE OPTOELECTRONICS CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGSU NORTH LAKE OPTOELECTRONICS CO LTD
- Filing Date
- 2026-05-09
- Publication Date
- 2026-07-14
Smart Images
Figure CN122391823A_ABST
Abstract
The application relates to an infrared blind element detection method and system based on image segmentation, and relates to the field of infrared detection. The method is based on a Deeplabv3+ network, and the method comprises the following steps: acquiring an infrared blind element data set, wherein the infrared blind element data set comprises real infrared blind element data and simulated infrared blind element data; constructing an infrared blind element detection model based on an improved Deeplabv3+ network; training the infrared blind element detection model according to the infrared blind element data set; and performing infrared blind element detection by using the trained infrared blind element detection model. The technical effect is that the detection accuracy of the infrared blind element is improved by using a neural network image segmentation technology.
Need to check novelty before this filing date? Find Prior Art