مجلة الاربعين المحكمة
This paper presents a fire detection and localization system designed to address the increasing incidence of fires in farms and outdoor areas. The proposed system utilizes a video camera capable of withstanding environmental changes, such as illuminance and color correlation to fire smoke and flames, offering a suitable alternative to traditional smoke sensor-based detection methods, which does not operate in open areas. By integrating color and motion detection approaches, the proposed fire detection system achieves accurate results. The input video decomposes using lifting wavelet transform to reduce the data size and as a result reduce the data processing time while preserving fire features for detection. The decomposed video frames are processed for color and motion detection to identify fire characteristics, with morphological post-processing eliminating unwanted pixels. The system calculates the detected fire area and applies threshold conditions for bounding. Fire localization is achieved through inverse camera parameter calibration and fire position mapping in pixels. The approach utilizes a projective transformation matrix to establish the relationship between frame pixels and real-world positions. Experimental results demonstrate a high accuracy rate for both offline and online (real-time) tests. The system achieves an average accuracy of 96% for real-time smoke and flame detection and 98% for offline flame and smoke detection. Fire localization is achieved with an error of less than 0.4 meters.