Oil is one of the most important energy supplies for economic development. In recent years, the fire safety problems of petrochemical enterprises have become prominent, with serious casualties and property losses. The continuously monitoring of key areas through the low-cost and intelligent infrared thermal imaging video monitoring system has important engineering application significance for the improvement of petrochemical site safety problems. According to the characteristics of infrared thermal imaging fire target, this paper proposes a method of deep neural network combined with time-domain feature analysis to realize fire detection. Firstly, high thermal pixels are extracted from the infrared image, and the gray-scale image is converted into a binary gray-scale image. Based on the YOLOv4 tiny framework, multi-level channel prediction and attention mechanism are added to detect the fire candidate target of the binary image, Finally, the candidate target is finally determined by analyzing the time-domain characteristics. Compared with the traditional temperature threshold judgment infrared temperature measurement fire alarm system, it can achieve high detection rate and effectively reduce the false alarm rate of the system. The intelligent security monitoring system in Petrochemical area designed in this paper has been applied in practical engineering, and the fire detection effect is good, which realizes the requirements of low power consumption, low cost and high reliability of the security monitoring system in Petrochemical area based on infrared thermal imaging.
In infrared imaging system, nonuniformity is the key factor limiting the improvement of imaging quality. In this paper, an adaptive neural network nonuniformity correction algorithm based on motion detection is proposed. The motion estimation algorithm based on gray projection is used to select the learning reference frame of neural network algorithm. Combined with edge detection algorithm and neighborhood variance information, the learning step size of neural network is adjusted adaptively, it effectively solves the “ghost effect” caused by insufficient scene motion in the traditional neural network algorithm. The algorithm automatically extracts effective image frames from the actual scene and updates the correction parameters, without planning the correction area and scene motion mode, and realizes automatic real-time correction.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.