| Research and development of embedded intelligent video-based fire detection equipment has important practical significance, which can protect the natural and ecological resources, promote forestry development and maintaining social stability. To achieve real-time forest fire monitoring and intelligent alarm, this paper design implements a fire monitoring system by DSP/IP module as the core and using solar power and wireless network transmission, on condition that electricity and network infrastructure poor in remote forest areas, by the support of the projects of major science and technology in Zhejiang -- solar energy-based forest fire monitoring technology and system development. The main work and contributions in this paper includes the following:(1) Design implementation of system hardware modules integration and multi-tasking scheduling communication of software framework.To implementation the requirements of the video coding and video flame detection, in this paper, using hardware modules Integrated approach, DSP module as the scheduling control core of the flame detection system, the design implements real-time online detection of fire algorithm and remote response of clouds-terrace and other terminals based on DSP/BIOS multi-task scheduling communication mechanism. And IP module as the data transfer hinge of the flame detection system, the design implements H.264 video coding, wireless networks and information exchange with the DSP module based on LINUX multi-thread scheduling communication mechanism.(2) Implementation of IP module I/O redirection and wireless network transmission.In this paper, using I/O redirections technique based on file-sharing, IP module's hardware limitations has been overcome and communications between the IP and DSP module are achieved by changing the original RS232 console terminal to a normal serial port. In addition, in order to overcome the restrictions of monitoring center can not get the wireless-router's public network IP address, we changed the traditional C/S network communication mode (ie, by means of monitoring center active connect to remote monitoring equipment to establish connection), through remote monitoring equipment unblock connect to monitoring center, achieved the rapid establishment of wireless network of flame detection system.(3) Design of intelligent fire algorithm, and to achieve it in DSP/BIOS environment.Accurate and timely detection for the early flame is an effective way to improve the warning ability of fire monitoring system. Full account of the impact of outdoor illumination, wind speed and flame-colored object on the flame detection in an open environment, this paper proposed an on-line flame detection algorithm based on mixed color model and parallel Hidden Markov Model (HMM). This algorithm has the following characteristics: 1) Mixed color model can adapt to the open environment and the impact of anti-light. 2) The proposed motion detection of modified Kalman Fliter (KF) model and mixed color model combination to achieve the online detection of suspected fire area. 3) Describes the phenomenon of the flame edge of the random flicker based on parallel-HMM, according to this model and to further eliminate the interference of false fire objects, improve the detection accuracy.The flame detection system is reliable and has a higher effectiveness through site commissioning and test verification. The proposed fire detection algorithm, which can accurately and rapidly achieve fire detection and alarm of the open environment under certain conditions, basically achieved the expected results. |
PDF Full Text Request