Pattern Recognition Of Gas Data Detected By Gas Monitoring System Based On Colorimetric Array Sensor

With technology advances, the accelerated process of industrialization and theimprovement of the decorative, the toxic and harmful gases in the atmosphere are greatincreased，and they are seriously harm to human health, so it’s very important to detecttoxic and hazardous gases, and design an efficient and effective gas detection device isurgently needed. It has very important significance to identify and then targeted to clearthese toxic and harmful gases for improving the atmospheric environment.Based on that the combination of certain chemicals and gas react to produce thecolor change mechanism, devised with the specific recognition Colorimetric Arraysensor, this paper reported that a gas detection system based on visual sensors wasdeveloped. The system can to make fast and accurate detection and identification ofgases. The article introduces the detection system, the establishment of the database, thepattern recognition of gases detection. Research work is mainly focused in the followingareas:①With specific gas molecules selectivity of the Colorimetric Array sensor, theprinciple of the feature map produced by reaction of sensors and gases correspondenceto gases, gas detection system was designed combination of upper and lowercomputer. The system architecture and feature extraction process was described.②According to the detection principle of the sensor, software system was designedfor the detection system. The software system includes the sensor image acquisition,image processing, feature extraction, and so on.③In order to reduce interference and error in the gas detection process, improvedmethods are designed for impurity gases, reaction temperature, gas flows uniformity,light and other factors impact on detection.④Establish a database system for the gas detection system by using access database.The database system is to manage the gas data and to provide a basis for patternrecognition of the unknown gas.⑤The gases data pattern recognition method was designed for these gases data, Thegases data pattern recognition divided into qualitative identification (identify gases’species) and quantitative identification (identify gas concentrations), the qualitativeidentification using cluster analysis to analyze. Due to the small difference between thegas concentrations and characteristics may contain cross-cutting between the values, cluster analysis is difficult to achieve quantitative identification, So the BP neuralnetwork and RBF neural network with learning and memory function were use toanalysis concentrations, and the improved BP neural network and improved RBF neuralnetwork methods could make good recognitions.