Application of Convolution Neural Network Analysis on Intra-row Weeding System for Vegetables

( I-chen Liu ),( Suming Chen ),( Chao-yin Tsai ),( Yung-huei Chang ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Weeds play an important, non-negligible role in crop cultivation because their competition for sunlight, moisture, nutrients, space and other resources directly affects the growth of crops. Application of chemical treatment on weed control will pollute the environment and agricultural products, while physical treatments is time-consuming and laborious, which leads to low efficiency. This research intends to develop an intelligent vegetable intra-row weeding system using image positioning technology to conduct physical weeding. Total of 474 cabbage images with weeds were captured in the field with camera, in which 379 of these images were used as training data, and the other 95 images were used as testing data. Through the image processing method of Convolutional Neural Network (CNN), the features were extracted and classified between identify cabbages and weeds. There were 381 cabbages in the verified images in total, only 3 of which were unidentified, with a success rate of 99.2%. No weed was identified as cabbage, and the positions of cabbages were also obtained. Field tests were conducted using this built model to identify cabbage and had good recognition rates even when weeds were more than training samples.

Development of a Sorting System for Inspection of Mushroom Mycelium Growth

( Vivian Liao ),( Suming Chen ),( Chao-yin Tsai ),( Kuang-wen Hsieh ),( Kuo-chih Tung ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Mushroom cultivation is one of the important industries in Taiwan's agriculture. In current mushroom cultivation process, which includes spawn production, substrate preparation & bottling, inoculation, spawning, scratching, fruiting, harvest and packing, required a large amount of labor. In this research, the improvement in fault inspection for spawn level and bacterial infection before scratching process was conducted. The aim of this research is to establish an inspection system using machine vision, pneumatic mechanism and Programmable Logic Controller (PLC). First, an experiment-used small lighting chamber was made to study the image processing method, which can get the development (expansion) of self-made simulated samples from its side view image. The accuracy to recognize the fault in self-made simulated samples is up to 99.9%. In real cultivation bottle samples fault inspection, it spent for about 10 seconds and can get well recognition. A prototype of sorting system has also been established to conduct the inspecting process and remove bottles with fault conditions of spawn level and bacterial infection.

Development of Phalaenopsis Flowering Quality Prediction Models

( Han-chun Hsu ),( Suming Chen ),( Chao-yin Tsai ),( Yung-huei Chang ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Phalaenopsis is an important exported flower in Taiwan and its flowering quality was reported to be correlated to its leaves’ carbohydrate contents and external traits. Phalaenopsis Sogo Yukidian 'V3' was used as the experimental samples in this research. A hand-held spectrometer and a hyperspectral system were used to build the carbohydrate content prediction models (CC model) respectively first, and then combined with leaves’ external traits obtained from hyperspectral imaging to build the flowering quality prediction model (FQ model). The advantage and application of FQ models built by both devices were compared. The results showed that the CC models built with MPLSR for glucose, fructose, sucrose, total soluble sugar, starch and total carbohydrates by hyperspectral system were all better than those by hand-held spectrometer. The results of FQ models with PLSDA and SVM also showed that the performance of hyperspectral system to discriminate quality levels was better. As for the ANN models’ results to predict the total number of flowers, the errors using both devices were all lower than 0.87. The research proved that using spectral technique to predict the interior contents in Phalaenopsis leaves or flowering quality were both feasible. Although the performance of hyperspectral imaging system is better than that of hand-held spectrometer, it was expensive and not as convenient as hand-held spectrometer. We can adopt some other ways such as using multispectral imaging system or optimize the performance of hand-held spectrometer if we want to apply the results to the Phalaenopsis industry.

Prediction of Sugar and Acidity Contents in Pineapple using Near Infrared Spectroscopy

( Bo-an Shang Kuan ),( Suming Chen ),( Chao-yin Tsai ),( Chih-hsiang Hsu ),( Ha-chun Hsu ),( Yung-huei Chang ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Pineapples are not only nutritious, but also have significant economic importance in Taiwan. Traditional technique used to determine the fruit’s quality is by hitting the fruit with palm. However, such method requires experience and is highly dependent on the condition of the fruit. More importantly, decisions made are often subjective. This research used near infrared spectroscopy to develop a fast and non-destructive method to measure the quality of pineapples. Tai-Nung No. 17 pineapple was used in this research to construct the calibration equation. During the experiments, pineapple samples are separated into three sections: top, middle, and bottom. Optical measurements are taken around each section in 90 degree increments, with a total of 131 samples. A multivariate model was then established by mathematical pre-treatments and modified partial least squared regression. Results showed strong correlation between constituents and optical spectrum, with a coefficient of determination r_c²=0.78,r_p²=0.747,SEP=1.142°Brix,RPD=1.803. Future experiment aims to increase the predictability of spectrum model by enhancing experimental methods, increasing sample size as well as minimizing experimental error.

Predicting the Anthocyanin Content of Kyoho Grapes by Nir Spectroscopy

( Si-yun Wang ),( Suming Chen ),( Chao-yin Tsai ),( Han-chun Hsu ),( Yung-huei Chang ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Kyoho grapes are in dark purple color. Anthocyanin plays a major role on the colors of berries, and contains antioxidant, anti-inflammatory, anticancer substance. The method of inspecting Anthocyanins so far is by either visual or destructive technique of using chemicals, but these methods could not carry out real-time accurate inspection. The aim of this study was to demonstrate the feasibility of non-destructive inspection in predicting Anthocyanin content by near-infrared spectroscopy. This study used the NIRS 6500 spectrometer ranged from 400 to 2500 nm. The prediction model of Anthocyanin content of Kyoho grapes was established by near-infrared spectroscopy and concentration of Anthocyanin. The results of this study showed the reliability and feasibility of near-infrared spectroscopy on predicting the Anthocyanin content of grapes. In the future, the Anthocyanin content can be quickly predicted by spectroscopic techniques.

An Application of Machine Vision on Identification of Sugarcane Nodes

( Shao-yuan Zhao ),( Suming Chen ),( Chao-yin Tsai ),( I-chen Liu ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Due to labor shortage, modern agriculture goes up on automation gradually, the planting of sugarcane is no exception. If the automatic planting machine is used, sugarcane seedlings should be prepared in advance. A sugarcane node is the main place where bud is grown from. The existing sugarcane node cutting machines rely on human judgement to determine the node locations. There are time-consuming and laborious to collect the sugarcane nodes. This study intends to use machine vision to identify sugarcane nodes for developing automatic machine. The two algorithms of R-CNN and FASTER R-CNN were used to identify sugarcane node and to compare their performance. The R-CNN algorithm is usually used for the identification of multiple targets, and its accuracy is less than FASTER R-CNN, but the processing speed is faster. In this study, 530 sugarcane photos (1300 nodes) were analyzed, 400 and 130 sugarcane photos were selected as the calibration and validation groups, respectively. The experimental results show that the processing time of the R-CNN can be completed within 0.02 sec with the identification rate of 97.9%, and the processing time and identification rate of the FASTER R-CNN are similar to those of the R-CNN. The both algorithms have good results, and can be applied to the development of automated sugarcane node cutting machines.

Study on a Wireless Environmental Monitoring System for Duck Incubator

( Rou-yan Peng ),( Suming Chen ),( Kuang-wen Hsieh ),( Chao-yin Tsai ),( Chih-hsiang Hsu ),( Jin-ming Tsai ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Taiwan's domestic development of incubators are in simple structure, at low prices, and generally used by the industry. However, high number of regional dead eggs, the inability to understand the environmental conditions of incubator and other issues occurred frequently, resulting in operation and production difficulties. This research developed a wireless environmental monitoring system to collect data and upload to the cloud. A variety of environmental sensors and Raspberry Pi 3 were used in the establishment of wireless sensing module and environmental information detection platform. The environmental information was displayed in real-time and uploaded to the cloud. Computational fluid dynamics (CFD) was also used to simulate the internal flow field inside the incubator to determine the environmental control features, and control heating and humidification equipment with a PID controller. In this study, the establishment of wireless environmental monitoring system was completed. The wireless sensing module and the environment information detection platform have also been developed and tested. The stability of wireless sensing module was 98.5%, and the CFD simulation error was under 20%, which was low enough to predict the actual field temperature changes, and can be applied to the regulation of the incubator environment.

Application of Smartphone and Cloud Server Technology with Near Infrared Spectroscopy on Sugar Content Measurement

( Pin-chih Fang ),( Suming Chen ),( Chao-yin Tsai ),( Jin-ming Tsai ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Near infrared (NIR) spectroscopy has been widely used in agricultural product inspection because it is non-destructive, fast, and easy to operate. In this research, portable near infrared light detection instrument was used to develop a smartphone interface that could analyze the sugar content of agricultural products via the cloud system. Using the android operating system smartphone as an intermediary software to control the spectrometer configuration, spectra data are transferred and calculated in the cloud system. The user interface would then display the predicted result. So far, an intermediary software has been developed that could transfer spectra data to the cloud server, and a pre-trained guavas calibration equation was used to predict the sugar content of guava. The entire process of scanning, calculating and transferring data takes approximately 10 seconds. In the future, the functionality of the user interface could be improved by allowing manipulation of data configuration during scanning, creating a more diversified user interface with faster analysis time.

The Evaluation of Drying Qualities of Tea Dryers in Taiwan

( Cheng-hou Chang ),( Suming Chen ),( Chao-yin Tsai ),( Kuo-chih Tung ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Tea drying is a necessary process of tea manufacturing to prevent deterioration, stabilize the tea quality, and to facilitate packaging and storage. Therefore, the tea dryer is an essential equipment for the tea industry; and the tea drying quality is an index to evaluate tea dryers. In this study, four types of tea dryers were used to conduct tea leaves drying experiments. Four tests of sensory evaluation, color analysis, soluble ingredients analysis and chemical composition analysis were used to evaluate the drying quality of the tea dryers. The experiment results showed that the vacuum roller dryer (VRD) in the overall tea drying qualities had a better evaluation, the next were low-humidity dryer (LHD) and multiple-layer conveyer dryer (MLD), the electric hot air dryer (HAD) was worst among all tea dryers. The comprehensive evaluation of tea drying qualities were investigated using area analysis of radar chart which were VRD (2.38) > LHD (2.15) > MLD(2.12) > HAD(2.06). Therefore, the vacuum drying with moderate stirring could improve qualities of tea drying. Based on the results of this study, tea farmers or makers can choose the proper dryer, and tea equipment manufacturers can improve and enhance the performance of tea dryers

Development of an Indoor Positioning System for Greenhouse Flying Robot

( Ming-jhe He ),( Suming Chen ),( Chao-yin Tsai ),( Chieh-yu Lin ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1

Precision cultivation is a new trend in agricultural production nowadays. The current greenhouse cultivation only uses a few of sensors to represent the overall state of the greenhouse. That couldn’t get the entire environmental conditions of greenhouse and make it difficult to implement precise cultivation in greenhouse. To utilize a quadrotor equipped with sensors to collect the environmental information in greenhouse; and locating the quadrotor is the primary task. Since the Global Positioning System cannot be used in the greenhouse, an indoor-positioning system to locate the quadrotor has to be studied. Lin (2017) has developed an indoor-positioning system which is based on UWB (Ultra-Wideband) technology but has not been able to extend to whole greenhouse. In this study, the number of anchors was increased by SPI (Serial Peripheral Interface bus) technology to extend the UWB indoor-positioning system to whole greenhouse. At present, the error of the absolute distance in the indoor positioning system is about 0.08 m. In the flight test, the quadrotor could fly on the scheduled paths and collect the environmental information.