Flame detection method and device based on long short-term memory model

Abstract

The invention discloses a flame detection method and device based on a long short-term memory model. The method comprises the steps of sample collection, feature extraction, normalization, long short-term memory model design and testing. The device comprises a three-channel infrared sensor, an ultraviolet sensor, an amplifying circuit, an A / D sampling chip and a controller, wherein the controller realizes digital filtering, feature extraction, normalization processing and long and short term memory model output results. The invention solves the problems that the traditional flame detector algorithms mostly adopt a fusion strategy of combining a threshold value method with an AND-OR relationship, but the simple linear classification and fusion strategy does not fully exert the advantages of combination of red and ultraviolet sensors and the like. Compared with a traditional BP neural network which is only related to current input, the recurrent neural network (RNN) of the invention can memorize previous input information, influence the current output result, make an optimal judgment on a weak part in a flame signal in combination with historical information, and can relieve the gradient disappearance problem.

Description

Technical field [0001] The present invention belongs to the field of detection and analysis, particularly relates to flame detection, and more particularly, to a method and a flame detection apparatus based on short and long term memory model. Background technique [0002] With the improvement of people's quality of life, work safety has become particularly important, flame detectors in their daily lives, the production of more and more applications. A flame detector means detecting a combustion chamber or flame intensity of the burner, the probe consists of two parts and signal processors, analog signals output represents the flame intensity, indicating whether the switch signal indicates a flame and (or) flame intensity video signal. UV flame detector into flame detectors, infrared detectors and flame detectors visible flame. The optical characteristics of the flame, the flame detectors used are three: one is a shorter wavelength ultraviolet flame ultraviolet radiation sensitiv...

AI Technical Summary

Problems solved by technology

Most of the algorithms use the fusion strategy of the threshold method combined with the OR relationship, and the performance is stronger than that of a single sensor. However, the simple linear classification and fusion strategy has not fully utilized the advantages of the combination of red and ultraviolet sensors, and there is still a huge room for improvement in the flame detection algorithm.

[0004] The Long-Short Term Memory model (Long-Short Term Memory) is a special recurrent neural network (RNN) model, which is proposed to solve the problem of gradient dispersion of the RNN model; in the traditional RNN, the training algorithm uses BPTT ( Back-Propagation Through Time) algorithm, when the time is relatively long, the residuals that need to be returned will decrease exponentially, resulting in slow update of network weights, which cannot reflect the effect of RNN's long-term memory, so a storage unit is needed to store memory, so LSTM model is proposed

Images