Research Grade Corn-based Collision Detection And Identification Of Acoustic Signals

Corn is widely planted in our country, but because of the affect of weather and storage method, corn kernels can be damaged by insects and moisture content, bringing down the quality of corn kernels. Along with the improved living standard, the demand for agricultural product quality has been raised.This research designs an agricultural products quality evaluation automatic detection device, which is used for the collection and processing the corn kernel acoustic signal, the objects of this research are three types of corn, which are undamaged kernels, insect damaged kernels and moldy damaged kernels. The methods of feature extraction are principal component analysis; decision tree and approximate entropy, respectively utilized the three methods to analyze the time domain and frequency domain characteristics of corn impact acoustic signal. In order to evaluate the quality of corn kernels, this study extracts many effective features. At last, BP neural network, fuzzy inference system and Elman neural network are respectively used to classify and evaluate the quality of corn kernels, and obtain a good recognition result. This study can provide effective basis for the quality evaluation of agricultural products.This paper mainly includes the following aspects:(1) The research background and significance of this field and the quality evaluation technologies for agricultural products are summarized. In particular, current achievement and development of quality evaluation technologies based impact acoustic signal are mainly summarized.(2) Describes the schematic of proposed experimental apparatus, experimental material and the collecting of corn kernel impact acoustic signal.(3) Analyzes impact acoustic signals from the time and frequency domain, extracts amplitudes, power spectral density, frequency and phase angle value as features, uses the principal component analysis method to reduce the dimensions of the feature data. BP neural network is used to classify the corn kernels. The classification accuracy of three kinds corn kernels are94.0%,91.7%and95.2%respectively.(4) The method combining the decision tree and fuzzy inference system is applied in feature selection and classification of corn kernels impact acoustic signal to detect corn quality. First nine kinds of statistical features including standard deviation, skewness and kurtosis are extracted as the input of C4.5algorithm to construct the decision tree; then according to the decision tree, membership function and IF-THEN regulation of fuzzy inference system are devised; finally the classification of impact acoustic signals is conducted and the recognition rate of the three types of corn in this experiment are97.6%,92.9%and96.4%respectively. The experimental result proves this approach is feasible for corn quality classification, and possesses of high accuracy and high application value.(5) The method extracts impact acoustic signal amplitude of undamaged kernels, insect damaged kernels and moldy damaged kernels, selects optimum parameters of approximate entropy, calculates the approximate entropy value of amplitude which is the feature vector of Elman neural network, then Elman neural network is introduced, the classification accuracy rate are95.0%,93.3%,96.7%for undamaged kernels, insect damaged kernels and moldy damaged kernels respectively. The experimental results show that it is feasible and effective in corn kernels quality evaluation using the approximate entropy value of amplitude as impact acoustic signal feature, presenting a new thinking for agricultural products quality evaluation.