| Postpartum grading of navel orange is an effective way to increase the added value.Using the modern technology to detect the appearance quality and internal quality of navel orange can realize the classified sale,high quality and high price.Foreign developed countries have launched intelligent citrus sorting equipment,but due to the high price and maintenance cost,it is difficult for domestic enterprises to widely use.The research of related fields in China started late.In particular,the intelligence and automation technology level of online quality inspection equipment for navel orange is not mature.Soluble solids content is an important factor affecting the taste of navel orange,and it is also an important index for internal quality detection and evaluation of navel orange.Scholars at home and abroad have studied online detection of soluble solids content in navel orange,but most of them get spectra at the equator or average spectral of several positions of navel orange for modeling,and use the same way to obtain spectra to detect soluble solids content online.However,in the application of industrial automatic detection,navel orange enters the detection line through the automatic fruit loading device,and its position state is random.For navel oranges with spherical shape,it is difficult to obtain the spectra of each navel orange at the equator.In addition,due to the large number of navel oranges detected in industrial production,navel oranges are all passing through the detection line at a very fast speed,and it is impossible to repeatedly obtain the spectra of multiple position of navel oranges and use the average spectra for detection.Therefore,the existing spectral model has some limitations in the application of fast online detection of soluble solids content in navel orange,which is in random position.In order to explore the online detection method of soluble solids content in navel orange,in this paper,Cara Cara navel orange in Fengjie,Chongqing was taken as the research object.The computer image recognition technology,visible near infrared spectrum technology and the knowledge of image processing,spectrum analysis,chemometrics and system control were applied to the online detection of soluble solids in navel orange.The main research work are as follows:First,a diffuse reflectance system and a diffuse transmittance spectroscopy system were set up to detect the soluble solids content in navel orange under static condition.Diffuse reflectance and diffuse transmittance spectra of three representative position of navel orange were obtained,including the equator,the navel and the pedicel.Four spectral preprocessing methods were used,including multiple scattering correction,standard normal variable transformation,first derivative combined with savitzky-Golay smoothing and second derivative combined with savitzky-Golay smoothing.Seven feature wavelength selection methods were used,including continuous projection algorithm,competitive adaptive reweighting sampling,Monte Carlo uninformative variable elimination,backward interval partial least squares,competitive adaptive reweighting sampling combined with continuous projection algorithm,Monte Carlo uninformative variable elimination combined with continuous projection algorithm and backward interval partial least squares combined with continuous projection algorithm.The spectral detection model of soluble solids content was established and compared.The results showed that the position of spectrum acquisition had great influence on the detection of soluble solids content in navel orange.When diffuse reflectance measurement was used,the diffuse reflectance spectrum obtained from the pedicle could not be used for modeling.When diffuse transmittance measurement was used,the signal-to-noise ratio of diffuse transmittance spectrum obtained from the pedicle was low,and the accuracy of modeling was poor,with a prediction root mean square error of 0.824 Brix.In addition,the diffuse transmittance spectral model based on the equator of navel orange was the best.The partial least square regression model established had a high detection accuracy,with a prediction root mean square error of 0.398 Brix.Second,an automatic image and spectrum online acquisition system of navel orange was designed,which included sample delivery system,image acquisition system,spectral acquisition system and control system.The system was tested with navel orange samples.The results showed that the designed image and spectrum online acquisition system of navel orange was able to acquire clear navel orange image and stable diffuse transmittance spectral signal under the set speed,which meets the basic requirements of online detection of navel orange.Third,the recognition method of online position of navel orange was studied.The top view images of 594 navel orange samples in random position were collected online.A method of online position classification of navel orange was proposed,which used 45° spatial latitude projection to segment the inspection area of top view image of navel orange.The inspection area was used to recognize the pedicle and navel of navel orange.Histograms of oriented gradients(HOG)and scale invariant feature transform(SIFT)were used to extract the features of the inspection area.Support vector machine(SVM)models were established.The modeling results of linear kernel function,polynomial kernel function,radial basis function(RBF)kernel function and Sigmoid kernel function were compared.The results showed that the SIFT-SVM model based on RBF kernel function was the best,the recognition accuracy of training set was 98.3%,and the recognition accuracy of verification set was 94.8%,which verified the correctness of position classification and recognition method.Fourth,two spectral models for online detection of soluble solids content in navel orange were studied and compared.A total of 3104 diffuse transmittance spectra of 194 navel orange samples at 16 specific positions and 85 navel orange samples at 3 random positions were collected online.The online detection model of soluble solids content based on the average diffuse transmittance spectrum at the equator of navel orange and the online detection model of soluble solids content based on the average diffuse transmittance spectrum of 16 positions of navel orange were established respectively.The two models were tested with navel orange samples in random position for three times.The results showed that the two models were greatly affected by the online position of navel orange.The accuracy of detection was poor,with a maximum prediction root mean square error of 0.848 Brix and 1.081 Brix,minimum prediction root mean square error of 0.732 Brix and 0.923 Brix,and average prediction root mean square error of 0.788 brix and 1.01 Brix.Fifth,in order to improve the accuracy of soluble solids content detection in navel orange when the online position of navel orange changes randomly,the pattern recognition technology was proposed to identify the position of navel orange.The recognition of the online position was combined with the corresponded spectral detection model of position.The online detection software system of soluble solids content in navel orange was developed based on the computer(MacBook Pro,Dual Intel Core i7 CPU,16 GB RAM),Mac operation system,Qt framework and Python language.The online detection of soluble solids content in navel orange at random position was realized combined with the designed online image and spectrum acquisition system of navel orange.The online detection system was tested with navel orange samples in random position for three times.The accuracy of detection was acceptable,with a maximum prediction root mean square error of 0.69 Brix,minimum prediction root mean square error of 0.672 Brix,and average prediction root mean square error of 0.68 Brix.The experimental results showed that the combination of online position recognition and spectral detection model of corresponded position for online detection of soluble solids content in navel orange was good.It could overcome the shortcoming of single spectral detection model and improve the accuracy of online detection of soluble solids content in navel orange effectively.This study has a reference value for improving the detection accuracy of soluble solids content of the quasi spherical fruit in online random position. |
PDF Full Text Request