Study On Rapid Detection Of Ammonia/Alkali Corn Stalk Composition Based On Near Infrared Technology

Corn straw is an important source of ruminant forage.Its rational exploitation and utilization play active roles in reducing animal husbandry cost.The research shows that crude protein in corn straw is one of the important indexes to measure its nutritional value.Coarse fiber is one of the important parameters to evaluate its palatability.By applying proper ammonia,alkali and ultrasonic treatment can effectively accelerate the degradation of straw,increase the crude protein content and reduce the crude fiber content of corn stalk.How to quickly and effectively determine the crude protein and fiber content of straw is very important for determining the optimum straw ammonification,alkalization and ultrasonic conditions and improving the comprehensive utilization ratio of straw.The chemical laboratory method is routinely adopted to measure the crude protein and fiber content in straws.The high price,long time and low efficiency of this complex method can not meet the basic needs of the on-line detection of large quantities of samples.Based on that,this paper established model of corn stalk composition by using near infrared spectroscopy analysis technology to realize fast measurement of straw crude protein and hemicellulose.The aim is to find a new fast,accurate and efficient method for chemical detection of straw components.This article mainly dealt with the following work:(1)Preparation of ammoniated and alkalized corn straw sample.54 kinds of maize stalks mainly planted in Heilongjiang province in the experimental base of Northeast Agricultural University were selected as experimental samples 5%urea and 4%sodium hydroxide solution were used for ammoniation and alkali treatment respectively,and the treated straw was treated by adopting the method of ultrasonic treatment.The crude protein content of 54 corn stalks was measured according to the national standard method.The content of crude fiber was determined by the normal form method,and the hemicellulose was extracted.Statistics show that the crude protein content was 2.536-6.854%,the average value was 3.715%,the hemicellulose content was20.739-32.254%,and the average value was 25.914%.(2)The collection of corn straw spectral image information and preparation and classification of samples.Firstly the crude protein and hemicellulose in the corn straw were analyzed by adopting abnormal samples,and 4 crude proteins and 2.5 hemicellulose abnormal samples were removed respectively.The model interaction validation decision coefficient R~2C increased from0.6790 and 0.6533 to 0.8402 and 0.8010 respectively.The results of sample culling were obvious.The remaining 50 crude protein and 52.5 hemicellulose samples were normally analyzed and distribution characteristics were counting.The results showed that the samples were evenly distributed and were representative.According to the K-S method,the sample sets were divided.40 calibration sets and 10 verification sets were selected for the rough protein model,40 the selection correction set and 12 verification set of hemicellulose were selected.Finally,the near infrared spectrometer Antaris II was selected to determine the corn stalk spectrum in the4000-12000 cm~-11 band.(3)Noise treatment of corn stalk spectrum.The effects of spectral noise removal by adopting smoothing noise,derivative denoising and orthogonal signal decomposition methods were compared respectively.The results showed that the best removal effect of crude protein and hemicellulose was found by using the orthogonal signal decomposition method and the two order derivative+smooth 21 points.After treatment,the interaction validation decision coefficient R~2C of the samples was relatively improved,respectively.It reached 0.8642 and 0.8282,which indicates that denoising for spectral noise removal is effective.(4)The selection of corn stalk spectrum's characteristics wave lengths.MWPLS and the correlation coefficient method were adopted to select the characteristic wavelengths.The results showed that 4000-12000 cm~-11 and 8500-9000 cm~-11 were the best for the model analysis of crude protein and hemicellulose,and the determination coefficient R~2C of validation set were 0.8642 and0.8489 respectively.(5)The construction of rapid measurement model of corn straw components.Quantitative analysis model of SVR,PLS and PCR were constructed and verified.The results showed that by using SVR,the rapid detection mode of crude protein was the most accurate.The root mean square error root RMSEP of the validation set was 0.2962,and the determination coefficient R~2P was0.9148.By using PLSR,the rapid detection model of hemicellulose was the most accurate.The root mean square error root RMSEP of the validation set was 0.8606 and the determination coefficient R~2P was 0.9007.In conclusion,the rapid detection model of maize straw composition based on near infrared spectroscopy is more accurate and feasible.It can provide new ideas and new references for the rapid detection method of straw composition.