| Research of on-line detection technology and equipment for fruit internal quality detection are of great significance to increase the economic income of fruit farmers and increase the economic benefits of the fruit industry.However,due to the difficulty of obtaining internal quality information and low signal-to-noise ratio,the detection technology and equipment especially for detecting deep internal disorders are difficult to develop.Apple is one of the main fruit varieties grown in China.And the planting area and output of apple are in the forefront of garden fruits.Due to climate,nutrient elements,and other factors,watercore disorder,as an internal physiological disorder that appears around the vascular bundles and core of apples,occurs in major apple producing areas such as Shaanxi and Xinjiang.It has a greater impact on the storage and commercialization of apples.Watercore disorder mostly occurs in more mature apples.Apples with watercore are sought after by consumers because of their unique taste,and they are commercially called ‘Honeyed apples'.This subject carried out experimental research aiming at the problems with the endogenousness of apple watercore disorder,no obvious characteristic signal,and serious difficulty of on-line detection.The feasibility and development of Visible/Near Infrared(Vis/NIR)spectroscopy technology for on-line detection apple watercore severity index(WSI)and soluble solids content(SSC)were explored.The purposes of this article were to propose an on-line detection method for WSI and SSC of apples with watercore disorder,design and develop a new conveying and detecting system for internal quality detection,and provide theoretical foundation and equipment support for promoting the commercialization of fruit after production.The main research contents and results of this paper were as follows:(1)In response to the demand for watercore disorder and SSC detection of apples,the optical property of apples with watercore disorder were studied,and the research on the on-line dynamic detection of watercore disorder and SSC in apples was carried out.Using the free tray type fruit sorting equipment developed by our team,the feasibility of apple watercore on-line detection was carried out under the condition of the layout of two light source.The results were as follows.The peak intensity of the intensity spectra of watercore apples were higher than that of the sound apple.With the increase of WSI,the peak intensity of the spectra gradually increased.The randomly distributed watercore tissues of different sizes and shapes changed the light transmittance of apples,causing apples' spectra had an obvious trend of changing with the change of WSI.This might be the reason for the differences in the spectra of apples at different detection positions,and it also led to the deterioration of the SSC prediction results.In the classification of watercore apples and sound apples,the recognition accuracy of knearest neighbor classification(k NN),back propagation neural network(BPNN),support vector machine(SVM)and one-dimensional convolutional neural network(1D-CNN)was above 95%.These results showed that it was feasible to use Vis/NIR spectroscopy technology for on-line classifying watercore apples and sound apples.In the k NN binary classification of different severity watercore apples and sound apples,the accuracy of discriminating apples with moderate watercore and sound apples was poor(68%),while the accuracy of discriminating apples with moderate or severe watercore and sound apples was higher(91%,100%).The prediction results of WSI and SSC also reflected the influence of watercore tissues on the dynamic detection of watercore disorder apples.(2)Aiming at the problem that the watercore tissues of different sizes and distributions in apples in the above study had a great impact on the WSI and SSC detection of apples.In this study,based on optical simulation research and experimental research,a dynamic detection method for apple watercore disorder and SSC with a fourlight source layout was established.And the impact of apple size on detection was analyzed.The results were as follows.The optical simulation research carried out on Light Tools software had concluded that the fruit spectra obtained under the four-light source layout could carry more internal information.The model established by the average spectra of apples of the same size had excellent performance.In the partial least squares regression(PLSR)prediction model of the SSC of the same size of apples,the better root mean square error of calibration set(RMSEC)and root mean square error of prediction set(RMSEP)were 0.34?Brix and 0.37?Brix,respectively,and the residual predictive deviation(RPD)reached to 3.78.The better RMSEC,RMSEP and RPD of the model using PLSR algorithm for WSI prediction were 2.00%,1.82% and 1.69,respectively.In the SSC and WSI predictions of apples of different sizes under the duallight source layout and the four-light source layout,the detection effect of the four-light source layout was better than that of the dual-light source layout.Using PLSR algorithm to predict the SSC under the four-light source layout could also obtain better RMSEC and RMSEP,which were 0.35?Brix and 0.43?Brix,respectively,and the RPD value is 3.58.The experimental results verified the conclusions of the optical simulation,proposed a better four-light source layout,and evaluated the impact of apples with different sizes on the detection.(3)Combined the optimized light source layout in the above research and the estimated impact of apple size on the online detection of apple watercore disorder,a new internal quality on-line detection equipment based on fruit size adaptive light source adjustment mechanism was developed.Aiming at the problems of poor positioning accuracy of free-tray fruit cups,the unsuitability of roller-type fruit cups for full-transmittance or half-transmittance spectral detection mode,a tray-type fruit and vegetable sorting cup with chain transportation was developed to meet the fulltransmittance or half-transmittance mode needs and it had the functions of conveying,grading and automatic resetting for quasi-spherical fruits.On the basis of light source layout optimization,an adaptive light source adjustment mechanism based on fruit size was developed to realize the spectral detection of fruits of different sizes.Based on the new tray-type fruit cup and adaptive light source adjustment mechanism,the whole machine was designed and developed,and the mechanical structure was designed to prevent damage.Using programmable logic controller(PLC)as the control center,a control system for on-line detection equipment was developed with table instructions as the core to realize the detecting and grading functions of new equipment.Based on the design of the control system,a strong electric power circuit and a weak electric control circuit were designed in the circuit system.Established a stable circuit system assembly of strong and weak electrical isolation,PLC load isolation,and spectrometer trigger signal isolation hardware system.The performance of the production line under static and dynamic conditions was tested using apples and reference(PTFE),and the spectral characteristics of apples and PTFE at different speeds were tested.The feasibility of the production line for spectral detection was determined.(4)Aiming at the optimization requirements of the four-light source layout developed in the above research on the apple posture,the influence of the apple posture on the apple watercore disorder and SSC on-line detection was studied,and the better posture with the fruit stem facing upward in the four-light source layout was optimized.On the prototype,three possible detection attitudes of apples were comprehensively considered and experimental studies were carried out(position 1: stem up,attitude 2:stem-calyx axis parallel to the conveying direction,attitude 3: stem-calyx axis perpendicular to the conveying direction).The research results showed that compared with posture 2 and posture 3,PLSR modeling algorithm could realize the SSC prediction performance with RMSEC 0.45?Brix,RMEP 0.49?Brix,RPD 2.91.The model effect could meet SSC on-line detection needs.In the discrimination of apple watercore disorder,the optimal posture and algorithm were posture 1 and SVM.Its' discrimination accuracy was consistent with partial least square discriminant analysis(PLS-DA)of posture 2,both were 96%.However,SVM showed better sensitivity and specificity reflecting the better recognition effect on watercore apples and sound apples.Research conclusions showed that the delivery method with the fruit stem facing upwards(posture 1)had certain advantages in detecting apple watercore disorder and SSC on the newly developed equipment.(5)For different size apples under a fixed light source,the relative position of the light-receiving area was inconsistent,which affect the correction of the spectra of different size apples.The adaptive light source adjustment mechanism developed in the above research could adjust the height of the light source illuminating the fruit surface.On the basis of the detection posture optimized by the above research,this research proposed a spectral correction method for different size apples based on the adaptive light source adjustment mechanism.The model effects before and after the correction were compared and analyzed.The heights of the light source were grouped according to the 4-size group of apples and then adjusted according to the illumination heights of the group to simplify the experiment process.The spectra with the same relative optical path length were obtained to study the method of spectra correction.The research results showed that in the modified spectral model of the adaptive light source detection system,the PLSR algorithm could be used to obtain a better detection model.The RMSEC,RMSEP,and RPD were 0.44?Brix,0.47?Brix,and 2.19,respectively.Compared with the spectra before correction,this model could obtain a lower RMSEP and a relatively close RPD value.The size-corrected spectra for apple watercore disorder identification under the PLS-DA algorithm had an accuracy of 81%.Although the accuracy was lower than the results of watercore disorder discrimination of apples with the same size.However,compared with the models without light source height adjustment and spectral correction,the more balanced sensitivity and specificity and higher accuracy showed the effectiveness and performance of the spectral correction method. |
PDF Full Text Request