Abstract:Aimed at the questions caused by unsuitable sampling frequency, such as data redundancy, high cost in hardware and software produced by high sampling frequency and the accuracy and stability which are hardly ensured under the condition of low sampling frequency, the grain impact frequency and sampling theorem were considered, consequently, the maximum sampling frequency was determined at 50 Hz. In data preprocessing, for high frequency sampling signal, two methods were put forward, one was an arithmetic average method, and the other was double threshold filtering mean method. The analysis result showed that the effect of double threshold value filtering and average treatment was better than that of the former. In order to investigate the influence of sampling frequency on yield estimation, the yield estimation tests of different sampling frequencies were carried out, and a frequency extraction method was proposed to generate different sampling signals. Under four kinds of sampling frequency as 1, 10, 25 and 50 Hz, four types of yield estimation models were established, and the prediction effects were also compared. The test results showed that high accuracy of the estimated yield can be obtained with high sampling frequency; using 50 Hz sampling frequency, the lowest average relative error was 3.04%; using sampling frequency higher than 10 Hz, it could ensure the average relative error was not higher than 5%. As a result, it is necessary to adopt at least 10 Hz as the sampling frequency for estimation of grain yield.