Quantitative detection of moisture content in rice seeds based on hyperspectral technique

To explore the best method for quantitative detection of moisture content in rice seeds, the total of 120 samples of rice seeds with different moisture content were studied by hyperspectral technique in the experiment. Sensitive wavelengths of moisture were firstly selected by calculating the migration rate, after that successive projections algorithm (SPA) was used to select characteristic wavelengths. The clustering method was proposed to increase the ability of prediction model by increasing the discrimination of hyperspectral eigenvalues of each sample group. Firstly, fuzzy C‐mean clustering (FCM) algorithm was applied to cluster the characteristic wavelengths selected by SPA. Then the prediction model was established by support vector regression (SVR). Due to the unsatisfied clustering effect, simulated annealing genetic algorithm (SAGA) was introduced for clustering. By comparing the results based on original eigenvalues, FCM and SAGA clustering, respectively, it was found that the best method was SAGA. The SAGA‐SVR mode achieved the value with Rp2 of .8892 and RMSEP of 0.0296. The relaxation variable was introduced to reduce interval threshold because the Rp2 was not ideal, and the final value achieved with Rp2 of .9318 and RMSEP of 0.0264. It was proved that the SAGA‐SVR model can be used for moisture detection of rice seeds. PRACTICAL APPLICATIONS: Rice is widely cultivated around the world, the moisture content of rice seed is an important index for judging the quality of rice seeds. The moisture detection of single rice seed is absolutely feasible, but the moisture content of single rice seed is not representative. Traditional methods cannot make batch detection of moisture content in rice seeds, which are also easy to cause secondary pollution in the process of detection, so they cannot meet the requirements of fine management in modern agricultural. Hyperspectral technology, a safe and effective technique, has been widely used in moisture detection of foods. This study showed that hyperspectral technology can accurately predict moisture content in rice seeds after proper data processing.