Study On The Acoustics Methods For Detecting Insect Damaged Wheat Kernels

The quality of wheat is related to people’s health. In the process of storage, the wheat is highly vulnerable to erosion of pests, became the insect damage wheat. Insect damaged wheat seriously affect the appearance and reduce commodity value and use value of the wheat, influence the wheat flour processing quality and eating quality (such as taste, smell, etc.) and processing production rate. Currently, the detection of wheat insect damaged was mainly used the different appearance, proportion and suspended velocity between the perfect and insect grain by artificial, that not only working intensity, slowly, and affected easily by the subjective results in the decrease of accuracy.This paper analysis the progress and applications on wheat quality of the nondestructive testing technology at home and abroad, to analyze the acoustic signals of wheat by signal processing technology, to study a better expression of the sound signal for wheat insect damage grain, to explore a rapid detection for insect damage wheat kernels by acoustic methods. First, select and prepare the samples of wheat, adjust the parameters of the test apparatus to make the grain fall with the same initial state and gather the sound signal of wheat grain which hit a metal target; then detect the endpoint of the gathering wheat signal based on the MATLAB software, analysis and extract the characteristic of grain in the time frequency domain, and last, extracted the characteristics better correlated to insect damage wheat quality, which is FFT and DCT transform characteristics TZPOWER, TZFRE1, DFT1, DFT2;Finally, built the recognition model of BP neural network and support vector machine (SVM) according to the requirements of identification, then extract the test sample characteristic of perfect and insect damage wheat kernels, then classified test samples of wheat grain quality by established model and analyze the recognition rate. To test eight set of samples, the accurate recognition rate is as high as91%, by using BP neural network, another accurate recognition rate is as high as89%by support vector machine (SVM), At the same time, the recognition correct rate can reach more than70%of the other two samples of wheat varieties. It proves that the two recognition models is good and. the method of acoustic recognition of wheat insect damage is feasible.