Spectral Image Analysis Of Mixed Organic Acids Based On A Machine Learning Approach

Complex ecological environment and biological community constitute a stable dynamic system of interaction and restriction.Generally,the materials in the system are inextricably linked,which forms the basis of harmonious and organic nature.In the natural system,most substances are complex and exist in the environment in the form of mixture.The interactions between these substances often have very complex relationships,and the material interactions are usually linear or nonlinear coupling processes.In the environmental system of human existence,a large number of unknown substances come from man-made emissions or natural formation,and the interactions are complex and uncertain.The function,exchange and behavior of material systems in the environment are influenced by the effects of material interactions.When there are many elements or compounds in the environment,the combined effects on soil,atmosphere,water and biology will be complex and difficult to estimate and analyze.Therefore,it is of great significance to study the material action in the unknown complex system to maintain the balance of ecosystem.Organic acids widely exist in the environment and have important applications in food,agriculture,medicine,biochemistry and other fields.They are important components of some agricultural products,and also affect the metabolism of soil microorganisms and plants.Organic acids test can help to know the health status of human body and screen for genetic diseases.As one of the components of industrial wastewater and waste gas,organic acids cause a certain degree of pollution.The detection and analysis of organic acids is very important to scientific research.At present,the commonly used detection methods of organic acids mainly include high performance liquid chromatography(HPLC),gas chromatography-mass spectrometry(GC-MS),liquid chromatography-mass spectrometry(LC-MS),ion chromatography(IC),etc.Generally,organic acids need to be extracted from the complex system for detection.In this process,it is difficult to realize the non-destructive detection of organic acids.When faced with a large number of mixed organic acids,the current methods take a long time to detect and the processes are tedious.Also,the consumption of chemicals makes the detection cost higher,and the number of samples for one-time detection is limited.Therefore,a more convenient and economical method is needed for the detection of mixed organic acids.In this study,a new method for identifying and analyzing mixed organic acids was proposed.In the method,the organic acids existing in complex chemical system can be directly analyzed and determined without being extracted.During the determination process,they are completely independent of other substances in the system.A large number of mixed organic acids information can be collected in a short time,and multi-dimensional measurement data can be output.We aim to find a new mode in the analysis of complex chemical systems and build a spectral device that can quickly obtain the information of the mixed organic acids.The principle is based on the different absorption characteristics of the substances to the spectrum.The light emitted by the combined light source passes through the organic acid solution to be tested,and the chromogenic reactions in the solution cause the change of absorption characteristics.The camera at the end of the device captures the residual light passing through the solution and then converts it into digital image information.These images are actually digital data that contains the spectral absorption characteristics of the substances.The large number of acquired images constitute big data.However,thousands of data are difficult to analyze by traditional methods.The machine learning technology developed in recent years has outstanding performance in processing big data because of its powerful computing power.We build a neural network for data analysis,input image data and tag values into the network for supervised learning,so as to establish a machine learning model,and finally achieve the prediction of mixed organic acid concentrations.The research provides a certain reference value for the future implementation of digital environmental management.The main conclusions are as follows:(1)Screening of color system.The color reaction of mixed organic acid solution is the basis of obtaining data.The colored material solution can selectively absorb part of visible light energy and present different colors.Therefore,we have carried out the screening of composite color agents.Finally,two color systems S1 and S2 were determined.The color system S1 is composed of methyl violet,bromothymol blue,bromophenol blue,Chromazurol S,bromocresol green,methyl orange and neutral red.In order to explore the disturbance effect of heavy metal ions on the reaction system,heavy metal ion cadmium was added to the color system S2.The color system S2 is composed of methyl violet,bromothymol blue,bromophenol blue,Chromazurol S,bromocresol green,methyl orange,neutral red and cadmium acetate.(2)The construction of spectrum equipment and the acquisition of spectrum image.The composition of the spectrum equipment mainly includes the combination surface light source,the sample cell and the camera device.The surface light source is composed of a light source and an extremely-combined filter(ECF)that contains special patterns.The light beam from the light source is transformed into an order of magnitude light beam combination modes through the combination filter.When the light passes through the sample cell containing the solution to be tested,the photographing device records the residual light in the form of pictures.The output end of the spectral device is connected with a mobile storage device,and the digital image information containing the spectral absorption characteristics of the solution is saved by it.(3)Digital processing of image data.In this process,we extracted the gray value of spectral images and drew contour maps,and analyzed the differences between spectral images under different conditions.In order to observe the different effects of the color system,the spectral images of the experimental group were analyzed by subtracting the blank background images.The characteristic spectral information visualization of mixed organic acids was realized by spectral image.(4)The establishment of machine learning model.The acquired image data set and the corresponding tag values were input into the network,and the data were divided into 80%training data set and 20%test data set,and then deep learning is carried out,and finally the model is established to realize the prediction of organic acid concentrations.The results showed that the accuracy and precision of machine learning were high in different complex chemical systems and the convolutional neural network had sufficient deep learning ability.(5)Performance analysis of spectral image.Through image analysis,we verified the stability of the operation of the spectral instrument,the accuracy and repeatability of the data,and also proved the extensive applicability of the research route in this experiment.