On-line Sorting System Research Articles

An online automatic sorting system for defective Ginseng Radix et Rhizoma Rubra using deep learning

ObjectiveTo establish a deep-learning architecture based on faster region-based convolutional neural networks (Faster R-CNN) algorithm for detection and sorting of red ginseng (Ginseng Radix et Rhizoma Rubra) with internal defects automatically on an online X-ray machine vision system. MethodsA Faster R-CNN based classifier was trained with around 20 000 samples with mean average precision value (mAP) of 0.95. A traditional image processing method based on feedforward neural network (FNN) obtained a bad performance with the accuracy, recall and specificity of 69.0%, 68.0%, and 70.0%, respectively. Therefore, the Faster R-CNN model was saved to evaluate the model performance on the defective red ginseng online sorting system. ResultsAn independent set of 2 000 red ginsengs were used to validate the performance of the Faster R-CNN based online sorting system in three parallel tests, achieving accuracy of 95.8%, 95.2% and 96.2%, respectively. ConclusionThe overall results indicated that the proposed Faster R-CNN based classification model has great potential for non-destructive detection of red ginseng with internal defects.

Online sorting of drilled lotus seeds using deep learning

For defective drilled lotus seeds, the inner bitter lotus plumule cannot be removed normally, leading to difficulties in subsequent nutrient extraction and food processing. There is an obvious difference in visibility of drilled hole between normal and defective drilled lotus seeds in the top view; thus, an online sorting method for drilled lotus seeds based on drilled hole detection is proposed in this study. First, a drilled hole detection model based on You Only Look Once (YOLOv3) is developed to detect the drilled hole features on the lotus seed surface. The model was tested and compared with the Faster Region-based Convolutional Neural Network (Faster R-CNN) and Single Shot MultiBox Detector (SSD) models, and it showed a better comprehensive performance in terms of accuracy and speed. A sorting control algorithm is also proposed to perform online sorting based on real-time drilled hole detection results. In addition, an auxiliary algorithm is proposed to prevent the boundary misjudgement of detected hole ownership between adjacent lotus seeds during continuous sorting. An online sorting system was designed, and sorting tests were performed. A sorting accuracy of 95.8% was achieved for the mixed defective and normal drilled lotus seed samples. The proposed method is expected to not only fulfil the practical requirements for the online sorting of drilled lotus seeds but also provide references for other agricultural products that require continuous online sorting based on the detection of local features.

An automatic sorting system for electronic components detached from waste printed circuit boards.

Recycling e-waste makes for eliminating the pollution to environment and recovering critical materials as one of resource. Printed circuit boards (PCBs) serve as the important part in all e-waste, containing valuable but hazardous elements to be recycled when they reach the end of life. For the recycling of waste PCB, the electronic components (ECs) are liberated from base board and to be treated separately for element recovery. Due to the diverse element composition, ECs deserve to be further classified and sorted to improve the efficiency of recycling, achieving the concept of accurate recovery. Currently, the recycling industry only roughly screen the ECs manually by labors, which increases the risk of health for exposure to the hazardous environment. Automatic solutions are necessary for replacing labors to classify and sort the waste ECs, thus safeguarding them against the hazards of factory environment. In this work, the YOLO-V3, an emerging image detection algorithm, is utilized to train the self-made dataset and classify the ECs into specific categories. To avoid surface damage that weakens the accuracy of object detection, the technology process of detaching the ECs is improved by building a nitrogen atmosphere during the desoldering process, which delivers great protection effects on ECs. Results of YOLO-V3 detection model present satisfactory classification capability for all the classes of ECs and a smart on-line sorting system is proposed to automatically separate the ECs detached from WPCB.

Online accurate detection of soluble solids content in navel orange assisted by automatic orientation correction device

Non-destructive online sorting system for fruit quality are gaining importance in the global agro-industry as they allow for optimising harvesting, storage and marketing management decisions. The most commonly used sensing technology is visible-near infrared (Vis/NIR) spectroscopy. However, in online detection, the reproducibility of spectra is easily affected by fruit placement attitude. Up to now, little or no work considered the problem of attitude correction before spectral acquisition. This work aims at filling this gap. In this study, Vis/NIR diffuse transmission technology was used to detect the soluble solids content (SSC) in navel oranges, the analysis models of SSC in navel orange with consistent-orientation (T1: navel upward, T2: navel forward, and T3: navel leaning with an angle about 30°-60°), mixed-orientation (mix of T1, T2 and T3) and corrected-orientation (T2 correction) were established. Partial least square regression (PLSR) combined with different spectral pretreatment and effective wavelengths (EWs) selection methods were used to establish the optimal SSC analytical models. Root mean square error of cross-validation (RMSECV), correlation coefficient of calibration (RC), root mean square error of prediction (RMSEP), correlation coefficient of prediction (RP), residual predictive deviation (RPD) and the ratio of error range (RER) are used as the evaluation indicators of the model. The results show that the inconsistency of the placement attitude affects the prediction accuracy. The mixed-orientation model contains inconsistent placement information of the fruit orientation, and the robustness of the model is improved, but the prediction accuracy decreased. In the consistent-orientation model, the best analytical results can be obtained with the T2 placement orientation. The automatic orientation correction device based on image processing technology can correct the attitude of randomly placed navel orange samples according to the T2 orientation. The analytical accuracy of the model is further improved. In the corrected-orientation model, the best SSC analytical model can be obtained using NWS-MSC (Norris Williams smoothing - multiplicative scattering correction) pretreatment and VCPA (variable combination population analysis) variable selection methods, and RP, RMSEP, RPD and RER of the optimal prediction model were 0.981, 0.313 °Brix, 5.176 and 16.613 respectively. This method can be extended to the online analysis of other spherical fruits.

Development and evaluation of an online grading system for pinto beans using machine vision and artificial neural network

The design of an intelligent system for qualitative evaluation of beans product impurities is the most important step necessary to make a bean sorting machine. In this research, a real-time system of pinto beans sorting (from red, white, and damaged beans, and stones) was designed and developed by combining image processing and artificial neural networks (ANNs). In total, six parameters were selected from the statistical characteristics of beans for classification of pinto beans from other beans and stones. Several ANN classifiers each with different number of neurons in the hidden layer were trained to determine the optimal structure. Optimal topology of ANN classifier was 6-12-8-2. In the first step the offline system was evaluated. The correct classification rate for pinto, white, red, and damaged beans and stones were 86.27, 100, 100, 54.9 and 65.3%, respectively. The average accuracy of offline method was 81.2%. The corresponding MSE were calculated as 0.05, 0.059, 0.013, 0.099 and 0.096, respectively. The accuracy of the online sorting system for pinto bean from others was 97.87%. The results showed that the designed system combined with ANN technique had acceptable efficiency in pinto been grading.

Auto-sorting commonly recovered plastics from waste household appliances and electronics using near-infrared spectroscopy

Abstract The recycling of plastics from Waste electrical and electronic equipment (WEEE) was constrained by the mix of types. Near-infrared (NIR) spectroscopy is suitable for polymer detection, and it is a rapid, non-destructive analysis method that can be applied to automatic on-line sorting system. The NIR spectra of four commonly recovered WEEE plastics, which are polypropylene (PP), polystyrene (PS), acrylonitrile butadiene styrene (ABS), and acrylonitrile butadiene styrene/polycarbonate (ABS/PC) blend, was collected. The flame-retardant ABS showed difference from ABS in NIR spectra. Three classification methods, which are spectral angle mapper (SAM), partial least squares discriminant analysis (PLS-DA) and linear discriminant analysis combined with principal component analysis (PCA-LDA), was tested. And the classification models trained on the virgin plastics have been compared with the models trained on the WEEE plastic to evaluate how these methods perform under limited training data. PLS-DA is one of the most widely used classification method in spectral data analysis, but it had unsuccessful prediction when the training set only included virgin plastics. But the overall prediction accuracy over 99% could be achieved by the other two whether the training set was the spectra of virgin plastics or WEEE plastics. In general, NIR spectroscopy has the competency of separating WEEE plastics. Finally, an automatic on-line sorting system was designed specifically for the large plastic segments from household appliances and electronics.

Rapid and non-destructive evaluation of seed quality of Chinese fir by near infrared spectroscopy and multivariate discriminant analysis

Chinese fir (Cunninghamia lanceolata (Lamb) Hook) is a promising timber species for planting in the sub-tropical ecosystem. However, the germination rate of Chinese fir seed lots is low, and viability determination by conventional methods (germination test and tetrazolium test) is time-consuming and destructive. Therefore, the aim of this study was to investigate the potential of Near Infrared (NIR) spectroscopy for rapidly and non-destructively determining viability of Chinese fir seeds. A total of 450 seeds was first identified as viable, empty, or dead (n = 150 seeds per seed lot fraction) by digital X-ray and their viability was further confirmed by germination test. NIR reflectance spectra were collected from single seeds using XDS Rapid Content Analyzer from 780 to 2500 nm. Classification models were developed on calibration set (n = 300 seeds) by Orthogonal Projection to Latent Structures-Discriminant Analysis (OPLS-DA) using the entire or selected NIR regions, and the fitted models were validated using external test set (n = 150 seeds). The model’s ability to recognize members (sensitivity) while rejecting non-members (specificity) of a given class was 100% for viable and dead seeds and 98% for empty seeds, attesting to the model’s robustness. When the model was fitted in the shorter NIR region (780–1100 nm), the sensitivity and specificity reached 100% for all seed lot fractions. The mean classification accuracy was 99% in the full and longer NIR regions and 100% in the shorter NIR region. The spectral differences among seed lot fractions could be attributed to differences in seed coat chemical composition in empty and dead seeds, presumably tannin content, and major storage reserves in viable seeds, notably fatty acids, proteins, and carbohydrate. It can be concluded that NIR spectroscopy has a great potential to evaluate Chinese fir seed quality, which could be applied in the development of an on-line sorting system.

초분광 영상기술을 이용한 수박종자 발아여부 비파괴 선별기술 개발

Abstract Seed viability is one of the most important parameters that is directly related with seed germination performance and seedling emergence. In this study, a hyperspectral imaging (HSI) system having a range of 1000–2500 nm was used to classify viable watermelon seeds from nonviable seeds. In order to obtain nonviable watermelon seeds, a total of 96 seeds were artificially aged by immersing the seeds in hot water (25°C) for 15 days. Further, hyperspectral images for 192 seeds (96 normal and 96 aged) were acquired using the developed HSI system. A germination test was performed for all the 192 seeds in order to confirm their viability. Spectral data from the hyperspectral images of the seeds were extracted by selecting pixels from the region of interest. Each seed spectrum was averaged and preprocessed to develop a classification model of partial least square discriminant analysis (PLS-DA). The developed PLS-DA model showed a classification accuracy of 94.7% for the calibration set, and 84.2% for the validation set. The results demonstrate that the proposed technique can classify viable and nonviable watermelon seeds with a reasonable accuracy, and can be further converted into an online sorting system for rapid and nondestructive classification of watermelon seeds with regard to viability.

가시광 및 근적외선 투과분광법을 이용한 감염 씨감자 온라인 선별시스템 개발

가시광 및 근적외선 투과분광법을 이용한 감염 씨감자 온라인 선별시스템 개발

Near Infrared Spectroscopy as non-destructive method for sorting viable, petrified and empty seeds of <i>Larix sibirica</i>

Ledeb. is one of the promising timber species for planting in the boreal ecosystem; but poor seed lot quality is the major hurdle for production of sufficient quantity of planting stocks. Here, we evaluated the potential of Near Infrared (NIR) Spectroscopy for sorting viable and non-viable seeds, as the conventional sorting technique is inefficient. NIR reflectance spectra were collected from single seeds, and discriminant models were developed with Orthogonal Projections to Latent Structure – Discriminant Analysis (OPLS-DA). The computed model predicted the class membership of filled-viable, empty and petrified seeds in the test set with 98%, 82% and 87% accuracy, respectively. When two-class OPLS-DA model was fitted to discriminate viable and non-viable (empty and petrified seeds combined), the predicted class membership of test set samples was 100% for both classes. The origins of spectral differences between non-viable (petrified and empty) and viable seeds were attributed to differences in seed moisture content and storage reserves. In conclusion, the result provides evidence that NIR spectroscopy is a powerful non-destructive method for sorting non-viable seeds of ; thus efforts should be made to develop on-line sorting system for large-scale seed handling.Larix sibirica Larix sibirica

Study on an automatic sorting system for Date fruits

In the present study, a machine vision based, online sorting system was developed, the aim being to sort Date fruits (Berhee CV.) based at different stages of maturity, namely Khalal, Rotab and Tamar to meet consumers’ demands. The system comprises a conveying unit, illumination and capturing unit, and sorting unit. Physical and mechanical features were extracted from the samples provided, and the detection algorithm was designed accordingly. An index based on color features was defined to detect Date samples. Date fruits were fed on a conveyor belt in a row. When they were at the center of the camera’s field of view, a snapshot was taken, the image was processed immediately and the maturity stage of the Date was determined. When the Date passed the sensor, positioned at the end of the conveyor belt, a signal was sent to the interface circuit and an appropriate actuator, driven by a step motor, was actuated, leading the Date toward an appropriate port. For validation of proposed system performance, entire samples were again sorted by experts visually. Detection rate of the system for Tamar and Khalal was satisfactory. Although the detection rate was insufficient for the Rotab stage, there was no a significant difference between system accuracy and that obtained by the experts. The speed of image processing system was 0.34s. System capacity was 15.45kg/h.

Modeling of Dropping Time of Kiwi Fruit in Water

In this study, the dropping time of Hayward kiwi fruit was theoretically formulated and then determined experimentally using water column. Some effective characters of kiwi fruit on rising time were determined using standard methods. The best model for dropping time of Hayward kiwi fruit was found as: with R2 = 0.96. The difference of fruit density minus the density of water and volumes of fruit had major effect on dropping time of kiwi fruit. The fruit shape had only a small effect on it. It was concluded that on online sorting system; dropping time has potential to remove poor quality fruit from kiwi fruit.

영상처리를 이용한 고구마 자동 선별시스템 개발

Grading and sorting an indeterminate form of agricultural products such as sweet potatoes and potatoes are a labor intensive job because its shape and size are various and complicate. It costs a great deal to sort sweet potato in an indeterminate forms. There is a great need for an automatic grader fur the potatoes. Machine vision is the promising solution for this purpose. The optical indices for qualifying weight and appearance quality such as shape, color, defects, etc. were obtained and an on-line sorting system was developed. The results are summarized as follows. Sorting system combined with an on-line inspection device was composed of 5 sections, human inspection, feeding, illumination chamber, image processing & control, and grading & discharging. The algorithms to compute geometrical parameters related to the external guality were developed and implemented for sorting the deformed sweet potatoes. Grading accuracy by image processing was <TEX>$96.4\%$</TEX> and the processing capacity was 10,800 pieces per hour.

Pilot scale evaluation of sorting technologies for CCA treated wood waste.

Two sorting technologies including a chemical stain method and an x-ray fluorescence technique were investigated for separating chromated copper arsenate (CCA) treated wood from other wood types in the wood waste stream. Stains were investigated in both laboratory and field settings. Studies included specially mixed solutions with chrome azurol, 1-(2-pyridylazo)-2-naphthol (PAN) and rubeanic acid chemicals. X-ray fluorescence was tested in the laboratory using a commercially available x-ray fluorescence spectrometer. Laboratory scale experiments showed that both technologies were able to detect CCA treated wood in mixtures of treated wood and untreated wood, with detection limits on the order of 3 to 5% CCA. Results from field experiments at construction and demolition facilities indicate that although the chemical stains can be effectively used to identify CCA treated wood waste in field settings, their use will be limited to sorting relatively small wood waste piles due to increased labor and time needed for processing the wood waste. Operational parameters for sorting using x-ray fluorescence technology were established. These parameters concluded that arsenic was the most sensitive metal for analysis, analysis time was less than 2 seconds per wood sample, and the maximum separation distance between the sample and the x-ray probe was 2.5 cm. X-ray technology shows considerable promise for separating large quantities of CCA-treated wood from other wood types in the field using an on-line sorting system.

On-line multi-unit sorting with resolution of superposition potentials

An on-line sorting system for multi-unit records is described which solves the problem of the resolution of superposition potentials. A segment of record up to 17 sec long, sampled at 8.3 kc/sec, is continuously stored on digital tape. A backwards and forwards scanning display facility is used by the operator to select up to four units representative of those he desires to analyse. A high-speed scanning operation compares all of the subsequent units in the record with the representative “templates”. Well matching units are automatically accepted and update their appropriate templates. For dissimilar units (usually superposition potentials) a first-order resolution into components is attempted and, if this is unsuccessful, their locations are marked and their “contemporary” templates are stored. After this scan, the systematically recalled superposition potentials are resolved by the operator using two special display facilities. An editing scheme whereby the original recording may be searched for suspected errors ensures error-free interval-time plots. A record of 10 sec duration, in which three different units are to be analysed, might require a total analysis time of 60 sec. The program, written for a DEC PDP-12 with an 8K memory and an Extended Arithmetic Element, is available from the author. In principle, the system is adaptable for any comparable laboratory computer with digital tape or disc systems.