Study On Detection Methods Of The Physical Characteristics And Quanlity Of Beef

The traditional detection methods of beef quality have disadvantages, such as detection of long time. The relationship between characteristics of electricity, ultrasound, optical and image-related and beef quality had been studied.Detection method of beef quality based on physical characteristics of beef was proposed. Methods can provide a theoretical basis for the internal composition and structural changes of beef and have great significance for the development of beef quality non-destructive testing methods and testing equipment. 1. Frequency characteristic of beef complex impedance was studied at frequency range from 100Hz to 10MHz. It not only can provide the frequency spectrum of complex impedance, but also provide suitable fixed value resistance for electrical characteristics detection circuit.The relationship between the electrical characteristics of beef and beef quality was studied. Electrical characteristics of beef, such as the apparent resistivity, conductivity, slope of voltage-related line, slope of voltage attenuation line, and natural logarithm of impedance were selected to predict the content of water, protein and fat in beef. The results showed that the prediction accuracy for water content of beef by low-frequency was 90.08±9.17%, which was smaller than the prediction accuracy by high-frequency conductivity (90.64±1.26%). High-frequency electrical characteristics can better reflect the beef quality. Fat content of beef can be better predicted by conductivity, and the prediction accuracy was 94.79±5.02%. The relationship between electrical characteristics of beef and protein content had been studied. The result showed that the detection accuracy was relative small (85.9±4.22%) and detection of protein by electrical properties had poor reproducibility.2. An ultrasonic velocity measuring system based on microprocessor control was proposed for the first time, and the unit of emission, receiving, counting and communication was designed.The test result showed that the detection accuracy for time was 99.75±0.58%. The relationship between the ultrasonic velocity and water content in beef was studied. Correlation coefficient reached 0.8163. Ultrasonic velocity was used to predict water content of beef, and the prediction accuracy was 96.11±2.73%. The prediction result proved relatively high prediction accuracy. The study also found that decrease of ultrasonic attenuation coefficient was accompanied by increase of water content. The factors which affect the measurement accuracy were discussed, such as temperature, holddown extent of probe, relative overlap area of probe and the direction of fiber in beef. The results showed that temperature played an important role in measurement. The increase of ultrasonic velocity was accompanied by increase of temprature with significant difference (R2=0.869). The average temperature coefficient was 4.6 m/s/℃. Attenuation coefficient did not change significantly with the increase of temperature. Relative overlap area between two probes was the most influential factor of ultrasonic attenuation coefficient. The overlap area was larger, the smaller the degree of attenuation when other factors were in the same conditions.3. Near-infrared analyzer with 11 narrow-band filtering film was selected. The wavelengths were 2336nm, 2310nm, 2230nm, 2180nm, 2100nm, 1940nm, 1680nm, 1759nm, 1818nm, 1982nm and 2139nm. The relationship between beef quality and response value from near-infrared analyzer was studied. High correlation was found in water content of beef with near-infrared response value (R2=0.652~0.866). The protein content of beef was moderately correlated with near-infrared response value (R2=0.558~0.695). The fat content of beef has a bad correlation with near-infrared response value (R2=0.387~0.572). To raise the accuracy of prediction, characteristic parameters were extracted by wavelet analysis. Beef qualitis have high correlation with low frequency wavelet coefficient, and bad correlation with high frequency wavelet coefficient. A characteristic parameter based on Daubechies wavelet decomposition was proposed, which was combination vector composed of low frequency wavelet coefficient with the first scale and second scale. The test showed that the prediction error(94.34±5.33%) for water content in beef from combination vector composed of first scale and second scale was smaller than the error from orginal near-infrared response value(88.62±3.22%). Method with wavelet analysis proved accurate and reliable. The color of beef was captured by solid color sensor, and was used for fitting the beef muscle and fat color grade model. The results showed that the color grade model can better predicted color grade of beef muscle and fat, the assessment accuracy were 94.56 % and 100%, respectively.4. This paper presented an improved watershed method used for beef marbling and rib-eye segmentation. In the HSV space, S and V were computered to a new parameter 2×S-V, and threshold segmentation was operated for pre-treatment. And then marked watershed algorithm was used for the segmentation of rib-eye region and beef marbling. The ultimate measurement accuracy for area and perimeter of rib-eye region were 96.98% and 92.67%, respectively. Segmentation error rate of rib-eye region was 9.44%. The ultimate measurement accuracy for area of beef marbling was 93.41%. Measurement accuracy of beef marbling particle number was 91.27%. Grading parameters reprented the grading of beef marbling were selected by correlation analysis. And marbling grade was evaluated by grading model, and compared to the manual evaluation method, the accuracy rate of grading was 95.0%. Physical maturity was also estimated by computer image processing technology for the first time. The calcification extent of cartilage in thoracic vertebra was estimated by computer vision in place of manual method. Retrieval method based on regional moment invariants and Hopfield neural network was proposed. The recognition accuracy for cartilage region was 92.31%. Classified parameter was optimized by the correlation with actual calcification extent of cartilage. The classifying accuracy was 86.0%. The beef carcass grade based on the beef marbling grade and physical maturity grade was estimated, and assessment accuracy was 95.83%.In this paper, detection methods of the physical characteristics and quality of beef were proposed with less time and higher measurement accuracy than the traditional method. It can provide detection method and theoretical basis for detection of beef quality based on multi-information fusion.