Forest fire early warning system based on unmanned aerial vehicle multi-target identification

The invention discloses a forest fire early warning system based on unmanned aerial vehicle multi-target identification, and the system is characterized in that an unmanned aerial vehicle flight module completes the fire task inspection of a forest designated area, and employs a GRO algorithm to plan the inspection track of an unmanned aerial vehicle; the method comprises the following steps: acquiring time series data of three characteristics of temperature, smoke concentration and carbon monoxide concentration, preprocessing forest multi-target time series data through a Kalman filtering technology on the acquired time series data, and inputting the preprocessed forest multi-target time series data into an improved N-HiTS model to identify and judge a forest fire phenomenon; the method for improving the N-HiTS model comprises the steps of adopting an EDO algorithm to carry out optimal solution solving on a hidden vector # imgabs0 # and an expression rate # imgabs1 # in the N-HiTS model, finally judging whet