| In recent years,non-destructive detection methods such as spectroscopy,image technology and bionic sensing technology have been widely used in the detection and analysis of agricultural and livestock product quality.These methods have the advantages of lossless,real-time,and fast detection.However,these methods generally rely on integrated hardware systems such as spectrometers,images,or sensor arrays,which are expensive,complex,and not easy to move and these devices are mostly used in laboratories or fixed occasions.It is not easy for consumers and other individual users to operate,nor is hard to promote in the market.In order to overcome the above shortcomings,it is urgently needed to develop a small structure,low price,easy to use and handheld equipment for the inspection of agricultural and livestock product quality.In this paper,the main agricultural and livestock products such as meat and fruit were taken as the research object and the detection mechanism,key technology research,software and hardware system design,and equipment performance verification,experimental analysis and evaluation of testing results were studied.The main research contents and results are as follows:1.A multi-spectrum diffuse reflectance method basing on parameters characteristic wavelengths for detecting quality parameters of agricultural and livestock products was proposed.For meat quality,66 pork meat samples and 61 beef samples were detected by using a 400-1100 nm visible/near-infrared diffuse reflectance hyperspectral technology respectively.Then original reflectivity spectrum and pre-processed spectrum were obtained and to establish partial least-squares regression(PLSR)prediction models.As a result,wavelengths releated to TVB-N parameters were selected and they are mainly distributed in wavelength range of 430,465,545,620,710,778,820,and 870 nearby.And these wavelengths are also related to the parameters of color L*,a*,b*and pH value in pork sample.For meat tenderness,wavelength range of 472,525,544,579,610,778,810,and 860 nearby were related to Warner-Braztler shear force(WBSF),and these wavelengths are also related to the parameters of color L*,a*,b*and pH value in beef samples in beef.For fruit quality,Vis/NIR detection system with wavelengths range of 400-1100nm and two external ring light source(25 W)were used to detect soluble solids content(SSC)in Fushui pears.And basing on original spectrum and pre-processed spectrum,wavelengths releted to SSC parameter were selected and distributed in 430,465,545,620,710,778,820 and 870 nm nearby.2.According to the principle of uniform illumination in the sample range,a method for designing portable integrated optical structure detector for agricultural and livestock products is proposed.For detecting meat quality,a multi ring and angle light sources structure was designed and it was composed of LED light sourceswhich were related to the quality parameters of meat.For detecting fruit quality,a structure of point symmetry multi angle polycyclic orthogonal direction light source is designed,which is arranged by two vertical directions in horizontal direction and vertical direction respectively.Finally,basing on the geometric position of light source module,size of the sample detection range,spatial position of the light source and detection lens and installation angle of each light source,the mathematical relations among them were calculated and analyzed.At last,Combined with optical simulation software Tracepro,best matching relation between the radiation intensity and the radiant area of the light source were determinated,which proved there analytical relations and caclautions were validated.3.A method for designing integrated optical detectors and handheld devices based on characteristic wavelength light source modules is proposed.The integrated detector is composed of a light source module,a detector and other driving circuits.Detection module was located on the top of the light sources.The detection module consists of a fixed focal length detection lens and a detection chip.The focal length of the lens is determined by the range of the detection sample and the size of the detection chip.And it was layout in the center area of the light source module,can be used to receive diffuse reflection light from the sample.In the driving circuit,STM32F103VET6 was used as the main control chip to control whole sytem operation.LED and supply constant current source circuits and they can achieve stable light intensity of light sources in each channel of the probe.In addition,a hierarchical amplifier circuit is designed to avoid oversaturation of the detector data acquisition.In addition,an amplifier circuit and RC filter circuit based on the chip are designed to amplify the collected weak light intensity signal.In the data transmission module,the transmission circuit of the integrated wireless module is designed to realize data real-time wireless communication with the terminal host computer or phone software.Finally,equipment shell is designed to construct a device which can install light source module,detection module,constant current source driving module,data acquisition and processing module,wireless transmission module,and display terminal.They were integrated into a small-sized,full-function small handheld device.The device can realize one-key operation to collect samples,upload and save test data.This method provides can provide a new theory for the design and construction of handheld agricultural quality detection equipments.4.In order to match the hardware devices,software using JAVA language in the Eclipse platform was to develop Android-based client-side APP software and Window system PC-side software.The software can achieve external device data and data exchange by Wireless transmission.In addition,three interfaces of pork freshness,beef tenderness and fruit quality parameters can switch simultaneously.The collected data was transmitted to the terminal interface in real time.In addition,the design result display on the device hardware can be synchronized with the software interface displaying and updating results.In addition,software and external handheld devices can communicate with each other by wireless communication.The detection results can be saved automatically to the computer-side path and the handheld device in the form of Word or excel documents type.Both softwares are capable of analysising the hardware-acquired data and build the prediction model.In addition,the external device can realize One-button operation and complete the data exchange,display detection results on the terminal.The APP client software can be used by consumer’s mobile phone,and the result can be shared and recorded in real time;the PC software can be operated in difficult wiring and complex environment.5.In order to verify the performance of devices,54 samples of fresh pork,61 samples of fresh beef and 60 samples of Fushui pear were tested and establish MLR,PLSR,and PL-SVR prediction models respectively.In the pork meat freshness detection,MLR models had the best results in parameters L*,a*,b*and TVB-N,and the prediction set correlation coefficients Rp were 0.9145,0.8933,0.9122 and 0.9040 respectively,and the prediction set error SEP was 2.0392,1.0605,0.6372 and 3.8114 mg/100g respectively.The parameter pH value had the best modeling result by using PL-SVR method,the prediction set correlation coefficient Rp and prediction set error SEP are 0.9319 and 0.0746,respectively.Then two groups of 24 samples at different time periods were used to verify stability and repeatability of the freshness prediction models.The results showed that the correlation between the predicted value and the actual value of the different time models were about 0.80,and the coefficient of variation was less than 1.4%.For beef quality parameters,the MLR models of color L*,a*and b*had the best results,and Rp are 0.9832,0.9072 and 0.9359,and the prediction set error SEP is 1.00,1.54 and 0.62,respectively.For the parameter pH value and WBSF value,PL-SVR method had the besr results.Rp are 0.9420 and 0.8120 respectively,and the prediction set error SEP are 0.19 and 6.11N.For the SSR parameters of Fushui pear,the LS-SVR model had the best result.The prediction set correlation coefficient Rp is 0.7779,and the prediction set error SEP is 0.511.The study and experiments demonstrate that handheld device designed based on characteristic wavelengths can detect the quality of agricultural and livestock products.This method can provide technical support and reference for the design of other miniaturized equipments in nondestructive detection field. |
PDF Full Text Request