An Autonomic Implicit Event Frame Generation Method for Asynchronously Perceiving Information
Regarding the generation method of event frame mode for event cameras,the event frames generated by traditional fixed time interval methods and fixed event number methods are subject to the phenomena of overly sparse image texture and blurred images,which lead to the reduced effectiveness of the visual tasks performed by the event frames gener-ated based on such methods.To address the above problems,an autonomic implicit event frame generation method with a-synchronously perceiving information is proposed,which can generate event frames according to the amount of information of accumulated event data.First,an image quality discrimination network based on residual network is designed to implicitly discriminate the semantic information amount contained in event frames during asynchronous sensory events;second,the event frames defined by the quality discriminative network are embedded into the semantic segmentation task,which re-duces the information acquisition latency of the event-based camera task while maintaining the segmentation accuracy;and third,the proposed method is verified on the public dataset DSEC.Compared with traditional methods,this event genera-tion method reduces the information acquisition delay of each event frame by 30 ms while maintaining the semantic segmen-tation accuracy.
asynchronous sensingevent-cameracomputer visionevent-framessemantic segmentationdelays in obtaining information
NETL
NSTL
万方数据
