APPLICATION OF ARTIFICIAL NEURAL NETWORKS TO MACHINE VISION FLAME DETECTION

CURRENT TECHNIQUE OF OPTICAL FLAME DETECTION RELY ON THE USE OF NARROW BANDWIDTH SENSORS WHICH HAVE BEEN DEMONSTRATED TO BE PRONE TO FALSE ALARMS. TO ADDRESS THIS NEED, AMERICAN RESEARCH CORPORATION OF VIRGINIA (ARCOVA) PROPOSES TO DEVELOP AN INTELLIGENT MACHINE VISION INTERFACE FOR THE DETECTION AND CLASSIFICATION OF SPECTRAL FLAME SIGNATURES. THE APPROACH BEING PROPOSED WOULD USE MACHINE VISION TECHNIQUES TO GENERATE HSI FORMATTED DIGITIZED VIDEO DATA THAT WOULD BE PRESENTED TO AN ARTIFICIAL NEURAL NETWORK FOR ANALYSIS. DEVELOPMENT OF A SILICON-BASED NEURAL NETWORK VIDEO CLASSIFICATION SYSTEM WOULD ALLOW ON-LINE MONITORING, DETECTION, AND CLASSIFICATION OF FLAME PATTERNS AS DERIVED FROM MACHINE VISION DATA. THE IMPROVEMENTS THE HSI VIDEO/NEURAL NETWORK SYSTEM OVER STANDARD FLAME DETECTION TECHNIQUES WOULD BE APPARENT IN AN IMPLEMENTATION OF AN ARTIFICIAL NEURAL NETWORK SIMULATION. HSI TECHNIQUES ARE CURRENTLY BEING IMPLEMENTED TO REDUCE IMAGE PROCESSING EFFORTS AND ARTIFICIAL NEURAL NETWORKS HAVE BEEN REPEATEDLY USED WITH SUCCESS IN THE CLASSIFICATION AND REDUCTION OF COMPLEX DATA INPUTS AND IN THE RECOGNITION OF PATTERNS BETWEEN DIFFERENT INPUT DATA. THEREFORE, ARCOVA ANTICIPATES POSITIVE RESULTS IN THE APPLICATION OF NEURAL NETWORKS IN AN INTELLIGENT VIDEO DATA CLASSIFICATION AND PATTERN RECOGNITION SYSTEM FOR MONITORING AND FLAME DETECTION PURPOSES.