Saliency map generation system, method and device based on dynamic visual sensor

Abstract

The invention belongs to the field of data processing, specifically relates to a dynamic visual sensor-based saliency map generation system, method and device, and aims to solve the problem. The present invention includes: acquiring event data through a dynamic visual sensor, accumulating and caching the DVS event data and information obtained from interrelated neurons through each neuron in the target detection network, and passing the activation function at preset times Obtain the neuron membrane potential, and output a pulse signal according to the relationship between the neuron membrane potential and the preset saturation threshold, the membrane potential u(t0) at the previous moment is attenuated and accumulated with the current received excitation, and the result is used for the pulse signal and membrane potential update judgment. The invention eliminates redundant background information from the saliency map generated by the target detection network array, improves the reliability and accuracy of target detection, and avoids misjudgment.

Description

technical field [0001] The invention belongs to the field of data processing, and in particular relates to a dynamic visual sensor-based saliency map generation system, method and device. Background technique [0002] Dynamic vision sensor (Dynamic vision sensor, DVS) has gradually gained attention. DVS is an event-driven photoelectric sensor. The information output by this type of sensor is different from the concept of "frame" of traditional image sensors, but is composed of a large number of independent events. In the data stream, each event contains information about an activated pixel point, so that the pixel information not only has brightness information, but also implies the time sequence information of event occurrence. At the same time, the dynamic vision sensor only generates output for areas where the brightness contrast changes, so it has the characteristics of filtering the background and only capturing moving targets, which can greatly reduce redundant backgro...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problems in the prior art that the dynamic visual sensor data cannot eliminate redundant background information, the amount of data is huge, too many resources are occupied, and the captured moving targets are not clear, the present invention provides a saliency map generation for dynamic visual sensors A system, the system comprising an event data acquisition module and an event data processing module;

Images