Studies On Throughput Nondestructive Detection Technology For Rice Kernel Physicochemical Properties By Near-infrared Spectroscopy

Improving rice quality and yield is the most important goal of rice breeding,and rice grain traits（composition,grain weight,grain shape,etc.）are closely related to rice yield and quality.Accurate and efficient identification and screening of the above traits in rice varieties will help promote the process of rice breeding with good quality.Due to the large number of grain traits,the trait characterization and detection methods are complex and time-consuming,which seriously restricts the efficiency of grain trait phenotype selection.The near-infrared（NIR）analysis method has numerous advantages such as non-destructive,accurate,and rapid detection of population grain traits.However,the breakthrough point of rice breeding often lies in single rice grain with specific physicochemical traits,while currently NIR technique still has the disadvantages of being unable to detect the chemical composition of rice grains in the form of single grains,and cannot effectively analyze their physical properties,which makes it difficult to be practically applied in the field of rice breeding.Based on the acousto optic tunable filters-near infrared（AOTF-NIR）single-grain dynamic detection platform,this thesis develops microchemical detection and model analysis methods for rice grain compositions such as amylose,fat,cellulose,moisture;and conducts research on non-destructive detection on its physicochemical phenotypes such as grain weight,grain shape,and chalkiness.This thesis provides technical support for efficient identification and screening of rice grain traits.The main findings of this thesis are as follows:1.To solve the problem regarding difficult validation of single-grain spectral analysis,a trace-chemical wet detection method related to rice grain compositions was established:（1）The sample amount required for testing the amylose content of a single grain has been reduced from 0.05 to 0.01 g（the weight of a single grain of rice is 0.02 to 0.04 g）.The correlation analysis showed that the correlation coefficient between the results of this method and the traditional method was 0.9950,indicating that this method can be used to determine the amylose content of single grain rice.（2）Improve microdetection method for fat content,the sample amount required for detection was reduced from 2 g at least to 0.3 g,and the correlation coefficient between the results of this method and the traditional method was 0.9329,indicating that this method can be used to determine the fat content of rice in small batches（8-15 grains）.（3）The trace detection method of cellulose in the bran layer overcame the shortcomings of the original method in determining samples with low cellulose content,the RSD was as low as 1.06%with good repeatability,and the correlation coefficient between the results of this method and the traditional method was 0.9601,indicating that this method can solve the problem of accurate determination of cellulose content in rice bran layer.2.The amylose and fat content of single grain rice,and the cellulose content in the rice bran layer were measured using a dynamic spectum platform（25～30 grains/minute）,and the results were compared and verified with those of the static spectrum platform.（1）For the detection of amylose content of single grain rice in the moving state,the R²_cv of the model was 0.666,compared with the R²_cv（0.724）of the diffuse reflection model established under the static platform,the models established on the two platforms had some common characteristic absorption peaks,indicating that the model selection was reasonable.（2）Under the dynamic collection platform,the R²_cv of the rice fat content model was 0.765,compared with the R²_cv（0.646）of the diffuse reflection model established under the static platform,indicating that the multi-grain rice averaging method can help correct the physical shape interference of single grain rice under the dynamic platform,thereby improving the prediction performance of the model.（3）Under the dynamic collection platform,the R²_cv of the cellulose content model in the rice bran layer under the dynamic collection platform was 0.801,compared with the R²_cv of 0.796 for the diffuse reflection model and 0.856 for the diffuse transmission model established under the static platform,indicating that the powdered samples are also suitable for throughput analysis.3.Establishing a non-destructive detection and sorting method for the physical properties of single grain rice based on NIR technique.（1）Under the two collection platforms,the R²_cv,of single grain rice moisture content model were both greater than 0.9.Since rice samples with different moisture contents can be easily prepared,grain moisture standard samples are ideal samples for instrument calibration.（2）Under the dynamic collection platform,a grain type determination model of single-grain rice was established.Based on the dynamic spectrum collection platform,73.12%of rice with different grain sizes were correctly sorted into their corresponding categories,and the classification accuracy rate for single traits（volume or aspect ratio per unit volume）was greater than 80%.which can meet the needs of modeling and screening of dynamic platform.（3）Under the dynamic collection platform,the R²_cv of the single-grain rice weight model was 0.709,compared with the R²_cv of 0.872 for the model established under the static platform.The model established under dynamic platform can pre-screen the weight of single-grain rice,which can reduce the impact of rice weight factor on the modeling and sorting of its chemical composition.（4）Under the dynamic collection platform,a qualitative discrimination model for single-grain rice chalkiness was established.The accuracy rate of the model established using the dynamic platform was greater than 85%,while that of the model established using the static platform was less than 70%,laying the foundation for achieving flux detection of chalkiness rate.At the same time,the rationality of selecting the near-infrared band were explained.