2results about How to "Less concern" patented technology

The invention provides a boat-borne infrared vision sea surface weak target identification and tracking method, and solves the technical problem that sea surface weak targets are missing and cannot be accurately identified and tracked in the existing night vision environment. The method comprises the following steps: identifying an infrared video frame sequence by using a target identification model to obtain coordinate information and confidence of a target, and dividing to obtain a high-confidence target, a low-confidence target and an invalid target; performing judgment, generating a new trajectory if the trajectory is null, or generating a target trajectory prediction result if the trajectory is null; matching the high-confidence target with the target trajectory prediction result; matching the low-confidence target with the track which fails in the first matching; updating the tracks which are successfully matched and failed in matching, and generating a new track for the high-confidence target which is failed in matching; and the updated coordinate information and the new trajectory are written into the trajectory sequence of the target, iteration is carried out until the last frame of image, and tracking is completed. The method can be widely applied to the technical field of target identification and tracking.

This invention belongs to the field of deep learning security technology, and particularly relates to a method for poisoning defense of modulation classification models based on Transformer semi-supervised feature remapping. The method includes: constructing a signal modulation classification model; obtaining signal samples at risk of poisoning and performing poisoning training on the signal modulation classification model to obtain a poisoned signal modulation classification model; connecting the backbone network of the poisoned signal modulation classification model to a Transformer module; inputting an unlabeled signal sample set into a joint feature remapping channel to obtain a remapped classification model; inputting a clean signal sample set into the remapped classification model and performing supervised fine-tuning; iteratively executing the feature remapping and supervised fine-tuning processes based on semi-supervised training until the model accuracy stabilizes, thus obtaining a signal modulation classification model with poisoning defense capabilities. This invention can effectively remove potential backdoors in the model, improve defense performance, and ensure the high availability of the repaired classification model.