US 2023/0146260 A1 · VIDEO PROCESSING SYSTEM AND VIDEO PROCESSING METHOD USING SPLIT LEARNING

Abstract

Disclosed is a video processing system including multiple unmanned aerial vehicles (UAVs) configured to capture a video of a fire site, wherein each UAV has a control unit including an input layer and a first hidden layer (hidden layer 1) and a central server connected to the multiple UAVs by wireless communication, wherein the central server includes multiple hidden layers and an output layer. The video processing system performs a learning process to determine whether a fire has occurred using a feature map in which an original video is difficult to recognize and personal information is protected. Thus, it is possible to fundamentally prevent the exposure and infringement of personal information.

Claims8

Dependent claims

A video processing system comprising:

multiple unmanned aerial vehicles (UAVs) configured to capture a video of a fire site, wherein each UAV has a control unit including an input layer and a first hidden layer (hidden layer 1);
and
a central server connected to the multiple UAVs by wireless communication, wherein the central server comprises multiple hidden layers and an output layer.

A video processing method comprising operations of:

(A) allowing a central control unit of a central server to transmit a signal for requesting video information of a fire site to each of multiple unmanned aerial vehicles (UAVs) through a central transceiving unit;
(B) allowing a control unit of each of the UAVs to extract a feature map obtained by distorting acquired video information of the fire site and transmit the feature map to the central server;
(C) allowing the central control unit of the central server to sum and regenerate the feature map;
(D) allowing the central control unit to perform computation by substituting the regenerated feature map into an activation function;
(E) allowing the central control unit to execute a pooling operation using a value output using the activation function;
(F) allowing the central control unit to compute a loss value using a pooling operation value and a preset target output value;
(G) allowing the central control unit to acquire a modified value for a parameter using the computed loss value and update the parameter using the acquired modified value for the parameter;
and
(H) allowing the central control unit to re-execute operations (C) to (G) using the updated parameter.

6 more elements

Dependencies are inferred from the claim text.

Video processing system and video processing method using split learning

Life of this patent

Video processing system and video processing method using split learning

Life of this patent

Abstract

Claims8