Flame target detection method, electronic equipment and storage medium

Abstract

The invention discloses a flame target detection method, electronic equipment and a storage medium. The method comprises the following steps: S1, obtaining an image; s2, training a flame detection model by using a deep learning convolutional neural network to obtain a detection model yo-v5; s3, each frame of the image is detected through the detection model yo-v5, and a preliminary detection result is obtained; s4, filtering the preliminary detection result through a filtering module based on flame image features; and S5, judging that the detection target meeting the judgment condition is flame, and otherwise, judging that the detection target is misjudged. According to the invention, the false detection rate during flame target detection can be reduced.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, and in particular relates to a flame target detection method, an electronic device and a storage medium. Background technique [0002] In real life, fire brings a huge threat to our personal safety and property safety. If the fire is not controlled in time, it will even cause social unrest. Therefore, fire detection is an important direction in the field of security monitoring. [0003] Existing flame detection methods can be mainly divided into two categories: [0004] (1) When machine vision technology has not been applied on a large scale, fire detection is usually achieved by installing corresponding sensor devices and using sensors to monitor the smoke, high temperature, light and other characteristics generated by the fire, so as to achieve fire detection, but due to the height of the equipment Factors such as the limitation of the detection range, the coverage range, etc...

AI Technical Summary

Problems solved by technology

[0007] (1) The traditional machine vision method uses artificial prior knowledge, and uses the channel characteristics and edge characteristics of the flame to judge the flame. However, the image-based vision method relies heavily on the artificially set threshold, and the anti-interference is not strong. Poor adaptability, prone to false negatives and false positives

[0008] (2) In recent years, with the in-depth development of deep learning in the field of machine vision, there are many solutions for using deep learning methods to detect flames. However, the training of convolutional neural networks requires a large amount of training data to support, training Data samples have a great impact on the robustness of the model. At the same time, the nature of deep learning makes the model theoretically unable to achieve 100% accuracy in the detection of flames. Some pictures similar to flames are also easily detected by the model identify as flame

Flame detection has the property of alarming in monitoring scenarios such as security, and requires a high accuracy rate or even a 100% accuracy rate. Otherwise, false detection will also cause corresponding waste of resources. For example, personnel need to go to the scene to confirm in person after the alarm

Images