Towards leveraging the role of machine learning and artificial intelligence in precision agriculture and smart farming

In recent years, modern spectral analysis techniques, such as ultraviolet–visible (UV-vis) spectroscopy, mid-infrared (MIR) spectroscopy, near-infrared (NIR) spectroscopy, Raman spectroscopy, terahertz (THz) spectroscopy, nuclear magnetic resonance (NMR) spectroscopy, laser-induced breakdown spectroscopy (LIBS), etc., have experienced rapid development and have been widely applied in various fields such as agricultural, food, pharmaceutical, petroleum, chemical industry, tobacco, environmental protection and medical science. A remarkable feature of all these techniques is to extract useful chemical information from the spectral data as detailed as possible with the aid of chemometric methods with the aim of significantly improving both robustness and accuracy of analytical results. Under the general background of the development in artificial intelligence, big data, cloud computing, and other technologies, the emergence of novel idea, approaches, and strategies endows chemometrics with a new vitality. Chemometrics has become the research focuses and hotspots in various fields, especially in the field of spectral analysis. This article reviewed various chemometric methods applied in modern spectral analysis in recent ten years, especially from the perspective of practicability, including spectral pre-processing, wavelength (variable) selection, data dimension reduction, quantitative calibration, pattern recognition, calibration transfer, calibration maintenance, and multispectral data fusion. More importantly, future trends in chemometric methods in the field of spectral analysis was also prospected in this article. It is sincerely expected that this summary and review could give specialists and scholars in the fields of spectroscopy and chemometrics certain inspiration to accelerate modern spectral analysis techniques booming evolution. 

Recent advances of chemometric calibration methods in modern spectroscopy: Algorithms, strategy, and related issues

Plant diseases pose a major challenge in the agricultural sector, which affects plant development and crop productivity. Sugarcane farming is a highly organized part of farming. Owing to the desirable condition for sugarcane cultivation, India stands among the second largest producers of sugarcane over the globe. At the same time, sugarcane gets easily affected by multifarious diseases which significantly influence crop productivity. The recently developed computer vision (CV) and deep learning (DL) models with an effective design can be employed for the detection and classification of diseases in sugarcane plant. The disease detection in sugarcane plant is not accurate in the existing techniques. This paper presents a quantum behaved particle swarm optimization based deep transfer learning (QBPSO-DTL) model for sugarcane leaf disease detection and classification which produces high accuracy. The proposed QBPSO-DTL method is designed and trained for the prediction of diseased leaf images. The proposed QBPSO-DTL technique encompasses the design of optimal region growing segmentation to determine the affected regions in the leaf image. In addition, the SqueezeNet model is employed as a feature extractor and the deep stacked autoencoder (DSAE) model is applied as a classification model. Finally, the hyperparameter tuning of the DSAE model is carried out by using the QBPSO algorithm. For demonstrating the enhanced outcomes of the QBPSO-DTL approach, a wide range of experiments were implemented and the results ensured the improvements of the QBPSO-DTL model.

/pdf/quantum-behaved-particle-swarm-optimization-based-deep-2j14h73n.pdf

Quantum Behaved Particle Swarm Optimization-Based Deep Transfer Learning Model for Sugarcane Leaf Disease Detection and Classification

Banana cultivation is one of the main agricultural elements in India, while the common problem of cultivation is that the crop has been influenced by several diseases, while the pest indications have been needed for discovering the infections initially for avoiding the financial loss to the farmers. This problem will affect the entire banana productivity and directly affects the economy of the country. A hybrid convolution neural network (CNN) enabled banana disease detection, and the classification is proposed to overcome these issues guide the farmers through enabling fertilizers that have to be utilized for avoiding the disease in the initial stages, and the proposed technique shows 99% of accuracy that is compared with the related deep learning techniques.

/pdf/banana-plant-disease-classification-using-hybrid-2z2kfhig.pdf

Banana Plant Disease Classification Using Hybrid Convolutional Neural Network

Rice is considered one the most important plants globally because it is a source of food for over half the world’s population. Like other plants, rice is susceptible to diseases that may affect the quantity and quality of produce. It sometimes results in anywhere between 20–40% crop loss production. Early detection of these diseases can positively affect the harvest, and thus farmers would have to be knowledgeable about the various disease and how to identify them visually. Even then, it is an impossible task for farmers to survey the vast farmlands on a daily basis. Even if this is possible, it becomes a costly task that will, in turn, increases the price of rice for consumers. Machine learning algorithms fitted to drone technology combined with the Internet of Things (IoT) can offer a solution to this problem. In this paper, we propose a Deep Convolutional Neural Network (DCNN) transfer learning-based approach for the accurate detection and classification of rice leaf disease. The modified proposed approach includes a modified VGG19-based transfer learning method. The proposed modified system can accurately detect and diagnose six distinct classes: healthy, narrow brown spot, leaf scald, leaf blast, brown spot, and bacterial leaf blight. The highest average accuracy is 96.08% using the non-normalized augmented dataset. The corresponding precision, recall, specificity, and F1-score were 0.9620, 0.9617, 0.9921, and 0.9616, respectively. The proposed modified approach achieved significantly better results compared with similar approaches using the same dataset or similar-size datasets reported in the extant literature.

/pdf/deep-learning-utilization-in-agriculture-detection-of-rice-3d3rn4ul.pdf

Deep Learning Utilization in Agriculture: Detection of Rice Plant Diseases Using an Improved CNN Model

Deep learning is a branch of artificial intelligence. In recent years, with the advantages of automatic learning and feature extraction, it has been widely concerned by academic and industrial circles. It has been widely used in image and video processing, voice processing, and natural language processing. At the same time, it has also become a research hotspot in the field of agricultural plant protection, such as plant disease recognition and pest range assessment, etc. The application of deep learning in plant disease recognition can avoid the disadvantages caused by artificial selection of disease spot features, make plant disease feature extraction more objective, and improve the research efficiency and technology transformation speed. This review provides the research progress of deep learning technology in the field of crop leaf disease identification in recent years. In this paper, we present the current trends and challenges for the detection of plant leaf disease using deep learning and advanced imaging techniques. We hope that this work will be a valuable resource for researchers who study the detection of plant diseases and insect pests. At the same time, we also discussed some of the current challenges and problems that need to be resolved.

/pdf/plant-disease-detection-and-classification-by-deep-learning-kg0yu97w6a.pdf

Plant Disease Detection and Classification by Deep Learning—A Review

Nondestructive prediction and visualization of total flavonoids content in Cerasus Humilis fruit during storage periods based on hyperspectral imaging technique

Soluble solids content (SSC) is one of the most important quality attributes affecting the taste and maturity of fresh fruit. In this study, with the cerasus humilis fruit as the research object, a prediction model of soluble solid content (SSC) in cerasus humilis (CH) is established based on visible / near-infrared spectroscopy to explore a nondestructive testing method of the interior quality of CH. The visible / nearinfrared spectral info (350-2500nm) of 160 CHs was collected to extract the reflection spectrum, establishing the linear model (PLSR) and non-linear model (LS-SVM) of CH’s spectral info and SSC. The prediction performance and stability of the model were justified using several statistical indicators namely correlation coefficient of the prediction set (Rp), the root mean square error of the prediction set (RMSEP), and the residual predictive deviation (RPD) index. Results showed that multiplicative scatter correction (MSC) was proved to be the best preprocessing method, UVE-CARS was the optimal method of dimension reduction, the quantities of characteristic wavelengths was 10 and the optimal model was UVE-CARS-PLSR, in which Rc is 0.8995, Rp is 0.8579, RMSEC is 0.8897, RMSEP is 0.9059, and RPD is 1.8766, indicating that the redundant data of the original spectrum can be reduced, the wavelength dimensions can be reduced, valid info can be retained and data processing can be simplified as UVE-CARS extracts characteristic wavelengths. Reference and theoretical basis are provided in this research for future research and development of portable detector and online sorting detection of CH internal quality. 摘要 可溶性固形物含量(SSC)是评价鲜果口感和成熟度的重要品质指标之一。本研究以欧李果实为研究对 象,基于可见/近红外光谱技术结合化学计量学方法建立欧李果中 SSC 含量的预测模型,探究欧李果内部品质 的快速无损伤检测方法。采集 160 个欧李果的可见/近红外光谱信息(350~2500nm),建立欧李果光谱信息 和 SSC 的线性模型(偏最小二乘回归算法)和非线性模型(最小二乘支持向量机)预测模型,通过预测集相关系数 (Rp)、预测集均方根误差(RMSEP)和剩余预测偏差(RPD)等指标来评价模型的预测性能及稳定性。结果表明, 多元散射校正为最佳预处理方法,最佳降维方法为 UVE-CARS,特征波长个数为 10,最优模型为 UVECARS-PLS,其中 Rp 为 0.8579,RMSEP 为 0.9059,RPD 为 1.8766。说明 UVE-CARS 提取特征波长可减 少原始光谱的冗长数据,降低波长维数,保留有效信息,简化数据处理。本研究为欧李果内部品质后续便携式 检测仪和在线分选检测研究提供了参考和理论基础。 INTRODUCTION Cerasus humilis (Bge.) Sok. (CH) is a kind of rosaceae cherry. It is usually grown in sun-slope sandy land and mountain shrubs, or cultivated in gardens. The fruit ripens around August 20 every year. CH is usually found in the northern regions of Yellow River. CH has a strong reticular root system and is droughtresistant, making it not only fixate soil but also regulate ecosystem, and it is also known as the “fruit rich in calcium” for its high calcium content, and is the third generation exclusive in China. CH’s seed kernels are the main source of Yu Li Ren (Semen Pruni), an herb known for the idea of “homology of medicine and 1 Bin Wang, As. Ph.D. Stud. Eng.; Junlin He, Prof. Ph.D. Eng.; Shujuan Zhang, Prof. Ph.D. Eng.; 2 Lili Li, As. Ph.D. Stud. Eng. Vol. 61, No. 2 / 2020 INMATEH – 252 food”. CH pulp can be consumed or made into juice, fruit wine, vinegar and preserved dried fruits, and it is good for health. The CH products have a unique taste and rich aroma and of high nutritional value, so it is known as a “super fruit”, and honoured as one of the three high-end fruits with American blueberry and Russian sea-buckthorn. The internal qualities of CH are abundant. Soluble solid content (SSC) and titratable acidity (TA) will affect its taste and nutrition when it comes to its quality assessment, and they serve as the measurement standards and important indicators of CH’s maturity. Regular testing of CH’s internal qualities is destructive, complicated and time-consuming, and its tissues are often damaged, affecting its sales and edibility and lowering its values (Guo et al., 2010). Near-internal spectroscopy has been successfully applied by domestic and overseas researchers to the fruit internal quality determination in recent years. There are in-depth studies on apples, mangos, tomatoes, kiwis, peaches, pears, strawberries and fresh dates. Ar et al., (2019) used near-infrared spectroscopy to predict the possibilities of internal qualities of persimmons such as SSC, Vc, total acid and hardness. The results showed that MSC is the optimal preprocessing method. With the establishment of a PLS calibration model, the optimal factor quantity of the SSC, Vc, total acid and hardness of persimmons is 17, 16, 12 and 12, respectively. Parpinello et al., (2013) studied a detection method, which combines nearinfrared ray (NIR) measurement and glucose analysis, and the partial least squares (PLS) model based on cross-validation serves as the main statistical parameter where the prediction set determination coefficient is 0.82 and the standard error of prediction is 0.83%°Brix. Purwanto et al., (2015) used near-infrared spectroscopy to predict the SSC and acidity of the mango species known as “Gedong Gincu”. The results showed that different preprocessing methods play critical roles in terms of establishing accurate models of mango internal quality prediction. Maniwara et al., (2014) used visible light and short-wave near-infrared spectroscopy to establish a PLSR prediction model for the indicators of soluble solids content, titratable acid content, ascorbic acid content, ethanol concentration, and peel hardness of passion fruit. Studies showed that PLSR prediction model proves to have the best prediction performance on the SSC in passion fruit, and the prediction correlation coefficient is 0.923. Escribano et al., (2017) collected the NI spectra of sweet cherries within the wavelength of 729 to 975 nm, and established a PLS prediction model of SSC for sweet cherries under two temperature conditions. The results showed that the determination coefficient (R2) of SSC calibration set is 0.922 and 0.946 and the standard error is 0.612% and 0.792% when the temperature is 0C and 23C, respectively. Yu et al., (2017) tested the SSC in grapes based on NIR and RC, RMSEP and RMSEC of PLS are 0.83, 0.76 and 0.84 by using the orthogonal test. Sun et al., (2018) used visible / nearinfrared semi-transmission spectroscopy to explore the influence of being unpeeled (complete) and peeled on the SSC detection accuracy of navel oranges. Studies showed that peel imposes significant impact on the SSC detection accuracy under the 5% confidence level. The correlation coefficient and root mean square error of the prediction set of the optimal PLS of the SSC in unpeeled and peeled navel oranges are 0.888 and 0.456% / 0.944 and 0.324%, respectively. Zhang et al., (2011) established the relations among visible light, near-infrared diffuse reflection spectrum (Vis/NIR) and the soluble tannin content in persimmons. The results showed that first derivative and detrending algorithms are the optimal preprocessing method, and the modified partial least squares (MPLS) demonstrated better prediction performance of the soluble tannin content in astringent persimmon, of which the RCV, RP2, RMSECV and RMSEP are 0.7227, 0.6785, 0.148 and 0.1763, respectively. The studies above show that it is feasible to use NIR to detect the internal qualities of fruits, but there is no research on the internal quality detection of CHs based on Vis/NIR spectroscopy. With cerasus humilis “Nongda 5” as the research object, a prediction model of the SSC in Cerasus Humilis (CH) is established based on visible / near-infrared spectroscopy to explore a nondestructive testing method of the internal quality of CH, hoping to achieve a CH SSC prediction model of good stability and high prediction accuracy. ASD Field Spee3 spectrometer has been employed to collect DRS data, and a spectral preprocessing method has been selected optimally. Four dimension-reducing methods, which are PLSR regression coefficients (RC), competitive adaptive reweighted sampling (CARS), and successive projections algorithm (SPA) and uninformative variable elimination (UVE), are used to extract the characteristic wavelengths, and different prediction models are built combined with PLSR and LS-SVM. They provide technical support for rapid, nondestructive, low-cost, and large-scale grading detection research of CH qualities. Vol. 61, No. 2 / 2020 INMATEH – 253 MATERIALS AND METHODS Cerasus humilis samples Sampling was conducted on August 15, 2019. The sampling site was the Jinzhong Agricultural HighTech Industrial Demonstration Zone Base in Taigu County, Shanxi Province, China (112°29'E, 37°23'N), and the variety was “Nongda 5”, sample growth state as shown in Fig. 1. Samples are of consistent maturity, shape and have no damage in order to minimize the influence of individual differences on experiment results. They were placed in a low-temperature fresh-keeping box which was transported to the laboratory the same day. The surface was wiped. Before the data acquisition, all samples were individually numbered and they were placed in an environment where temperature is 25°C and relative humidity is 20% for 6 hours to prevent the temperature from affecting the spectra and qualities. Fig. 1 Growth state of cerasus humilis A high-precision electronic balance (FA1004N, Shanghai) and a Vernier Caliper (Mitutoyo, Japan) were used to weigh and measure the weight and diameters of each sample. Table 1 shows the statistics of 160 samples. Table 1 Statistics of 160 samples Sample parameters Min. Max. Mean Standard deviation Variable coefficient (%) Diameter [mm] 15.48 28.66 24.72 3.62 14.64 Weight [g] 7.47 16.18 11.39 1.76 15.45 Vis/NIR collection A FieldSpec3 analytical spectral device (ASD, USA) was used to collect the VIS/NIR data of samples. The interval of spectral data is 1 nm, the number of scans is 30 times, the resolution is 3.5 nm, and the wavelength range is 350-2500 nm. Diffuse reflection was employed to sample the spectra. Each sample was scanned 3 times at an interval of 120 degrees above the equator, and its average value was taken as the final spectral data. The spectral experiment plat

Nondestructive testing of soluble solids content in cerasus humilis using visible / near-infrared spectroscopy coupled with wavelength selection algorithm

In order to realize the rapid and non-destructive detection of fresh Cerasus Humilis’ (CH) classification, and promote the deep-processing of post-harvest fresh fruit and improve market competitiveness, this study proposed a nonlinear identification method based on genetic algorithm (GA) optimized back propagation (BP) neural network of different varieties of fresh CH fruit. “Nongda-4”, “Nongda-5”, and “Nongda-7” fresh CH fruit were selected as research objects to collect their visible/near-infrared spectral data dynamically. The original spectra were preprocessed by moving smoothing (MS) and standard normal variate (SNV) methods, for the characteristic wavelengths were extracted with four dimension-reducing methods, namely principal components analysis (PCA), competitive adaptive reweighed sampling (CARS), CARS-mean impact value (CARS-MIV), and random frog (RF) algorithm. Finally, the BP prediction models were established based on full-spectrum and characteristic wavelengths. At the same time, the GA optimization was used to optimize the initial weight and threshold of the BP neural network and compared with the partial least squares’ discrimination analysis (PLS-DA) linear model. Through comparing the MS (7)+SNV was proved to be the best preprocessing method, the CARS-MIV-GA-BP model had the best discriminant accuracy, the prediction set accuracy was 98.76%, of which the variety “Nongda-4” and “Nongda-5” recognition rate were 100%, the variety “Nongda-7” recognition rate was 96.29%. The results show that the GA can effectively optimize the initial weights and threshold randomization of the BP neural network, improve the discrimination accuracy of CH varieties, and the CARS-MIV algorithm can effectively reduce the number of input nodes of the BP neural network model, simplify the structure of BP neural network. This study provides a new theoretical basis for the detection of fresh CH fruit classification.

On-line detection of cerasus humilis fruit based on vis/nir spectroscopy combined with variable selection methods and ga-bp model

The accurate segmentation of apple leaf disease spots is the key to identifying the classification of apple leaf diseases and disease severity. Therefore, a DeepLabV3+ semantic segmentation network model with an actors spatial pyramid pool module (ASPP) was proposed to achieve effective extraction of apple leaf lesion features and to improve the apple leaf disease recognition and disease severity diagnosis compared with the classical semantic segmentation network models PSPNet and GCNet. In addition, the effects of the learning rate, optimizer, and backbone network on the performance of the DeepLabV3+ network model with the best performance were analyzed. The experimental results show that the mean pixel accuracy (MPA) and mean intersection over union (MIoU) of the model reached 97.26% and 83.85%, respectively. After being deployed into the smartphone platform, the detection time of the detection system was 9s per image for the portable and intelligent diagnostics of apple leaf diseases. The transfer learning method provided the possibility of quickly acquiring a high-performance model under the condition of small datasets. The research results can provide a precise guide for the prevention and precise control of apple diseases in fields.

/pdf/diagnosis-and-mobile-application-of-apple-leaf-disease-ui2jcprd.pdf

Bin Wang

Papers

Plant Disease Detection and Classification by Deep Learning—A Review

Nondestructive prediction and visualization of total flavonoids content in Cerasus Humilis fruit during storage periods based on hyperspectral imaging technique

Nondestructive testing of soluble solids content in cerasus humilis using visible / near-infrared spectroscopy coupled with wavelength selection algorithm

On-line detection of cerasus humilis fruit based on vis/nir spectroscopy combined with variable selection methods and ga-bp model

Diagnosis and Mobile Application of Apple Leaf Disease Degree Based on a Small-Sample Dataset