Fire Detection Algorithms Using Multimodal ... - Bilkent University

CHAPTER 6. WILDFIRE DETECTION 84are determined from actual forest fire videos and test fires. They are updated usingthe least-mean-square (LMS) algorithm during initial installation. The LMSalgorithm was initially developed by Widrow in 1960’s together with neural networks[91]. The error function in the LMS adaptation is defined as the differencebetween the overall decision of the compound algorithm and the decision of anoracle. In our case, the oracle is the security guard. The compound decision algorithmwill obviously produce false alarms. The system asks the guard to verifyits decision whenever an alarm occurs. In this way, the user actively participatein the learning process.The chapter is organized as follows: Section 6.2 describes briefly each oneof the four sub-algorithms which make up the compound (main) wildfire detectionalgorithm. Adaptive active fusion method is described in Section 6.3. InSection 6.4, experimental results are presented and the proposed online activefusion method is compared with the universal linear predictor and the weightedmajority algorithms. Finally, conclusions are drawn in Section 6.5.6.2 Building Blocks of Wildfire Detection AlgorithmWildfire detection algorithm is developed to recognize the existence of wildfiresmoke within the viewing range of the camera monitoring forestal areas. The proposedwildfire smoke detection algorithm consists of four main sub-algorithms:(i) slow moving object detection in video, (ii) smoke-colored region detection,(iii) rising video object detection, (iv) shadow detection and elimination of shadowregions, with decision functions, D 1 (x, n), D 2 (x, n), D 3 (x, n) and D 4 (x, n), respectively,for each pixel at location x of every incoming image frame at timestep n. Computationally efficient sub-algorithms are selected in order to realizea real-time wildfire detection system working in a standard PC.Decision functions D i , i = 1, ..., M of sub-algorithms do not produce binaryvalues 1 (correct) or −1 (false), but they produce zero-mean real numbers for