彩色可视传感阵列基元匹配快速定量算法

A fast quantitative identification algorithm of colorimetric visual-sensor-array based on basic units matching

在彩色可视传感阵列传统处理方法中,存在着数据量大,人工分析困难、种类浓度识别难以同一次实现等问题,针对这些问题,考虑到同类气体饱和响应阵列点位置一致性的特点,提出一种彩色可视阵列基元匹配快速定量识别算法。该算法首先采用设置经验阈值消除冗余量,进行去噪和特征提取,减少人工分析量;然后进行基于二值化基元图模板匹配的定量分析种类识别,减少计算量,增加气体识别效率和精度;最后,综合模糊逻辑和神经网络2种人工智能方法的优点,建立彩色传感阵列气体浓度识别的自适应模糊推理系统。算法优势在于将不同气体的种类和浓度检测分开进行,解决了种类、浓度同时识别时可能出现特征数据交叉感染导致错误识别的问题。基元模板匹配分析结果显示,氨气、氯气和二氧化硫3种气体分类识别结果准确率达100%,利用模糊神经网络方法对氨气浓度识别率准确度较高,误差在5%以内。

There are many difficulties to process a colorimetric visual-sensor-array by traditional processing methods, such as complicated manual analysis due to many data and hard to finish varieties and concentrations identification simultaneously, etc. In view of colorimetric-sensor-array＇s same location response to the same gas, a fast quantitative identification algorithm of colorimetric visual-sensor-array based on basic units matching which can solve these problems is proposed. First, denoising and feature extraction are processed by setting experienced threshold to reduce redundancies and lessen manual analysis. Second, a creative qualitative analysis method based on basic units is put forward, which not only reduces computation, but also increases efficiency and precision. Finally, a ANFIS of NH3 concentration recognition utilizing advantages of fuzzy logic and neural network is used to distinguish low concentration NH3. The advantage of this algorithm is that varieties and concentrations of different gases could be detected successively, solving the problem of recognition errors caused by characteristic data infection when varieties and concentrations based on basic units show of different gases are detected simultaneously. The results of template matching that the classification accuracy of NH：~, C12 and S（）2 are 100%. The low concentration NH：~ classification accuracy is also very high after species identification with measurement errors below 5%.