Feasibility and reliability of agricultural crop height measurement using the laser sensor array

Crop height measurement is widely used to analyze and estimate the overall crop condition and the amount of biomass production.Not only is manual measurement on a large scale time-consuming but also it is not practical.Besides,advanced equipment is available but they require technical skills and are not reasonable for smallholders.This article investigates the feasibility of a simple and low-cost measurement system that can monitor crops height of paddy rice and wheat using laser technology.After designing and fabricating,this system was tested and evaluated in both laboratory and farm sections.In the laboratory,paddy rice height was measured,and in the field section,the height detection system measured wheat height.The results showed that the coefficient of determination(R3)between manual measurement and height detection system measurement for paddy rice was 0.96 and for wheat was 0.85.Besides,there was no significant difference between the two datasets at the level of 5%.Hence,this system can be a useful and accurate tool to monitor crops height in different growing steps.

External defects and severity level evaluation of potato using single and multispectral imaging in near infrared region Author links open overlay panel

Non-invasive potato defects detection has been demanded for sorting and grading purpose.Researches on the classification of the defects has been available,however,investigation on the severity level calculation is limited.For the detection of the common scab,it has been found that imaging in the infrared region provide an interesting characteristic that could distinguish defected area to normal area.Thus,investigations on this wavelength range is interesting to add more knowledge and for applications.In this research,the multispectral image has been obtained and investigated especially at three wavelengths(950,1150,1600 nm).Image pre-processing and pseudo-color conversion techniques were explored to enhance the contrast between defects,normal background skin area and soil deposits.Results show that external defects,such as common scab and some mechanical damage types,appear brighter in the near infrared region,especially at 1600 nm against the normal skin background.It has been found that pseudo-color images conversion provides more information regarding type if surface characteristics compared to grayscale single imaging.Image segmentation using pseudo-color images after multiplication operation pre-processing could be used for common scab and mechanical damage detection excluding soil deposits with a Dice Sorensen coefficient of 0.64.In addition,image segmentation using single image at 1600 nm shown relatively better results with Dice Sorensen coefficient of 0.72 with note that thick soil deposits will also be segmented.Defect severity level evaluation had an R2 correlation of 0.84 against standard measurements of severity.

Tea picking point detection and location based on Mask-RCNN

The accurate identification,detection,and segmentation of tea buds and leaves are important factors for realizing intelligent tea picking.A tea picking point location method based on the region-based convolutional neural network(R-CNN)Mask-RCNN is proposed,and a tea bud and leaf and picking point recognition model is established.First,tea buds and leaf pictures are collected in a complex environment,the Resnet50 residual network and a feature pyramid network(FPN)are used to extract bud and leaf features,and preliminary classification and preselection box regression training-performed on the feature maps through a regional proposal network(RPN).Second,the regional feature aggregation method(RoIAlign)is used to eliminate the quantization error,and the feature map of the preselected region of interest(ROI)is converted into a fixed-size feature map.The output module of the model realizes the functions of classification,regression and segmentation.Finally,through the output mask image and the positioning algorithm the positioning of the picking points of tea buds and leaves is determined.One hundred tea tree bud and leaf pictures in a complex environment are selected for testing.The experimental results show that the average detection accuracy rate reaches 93.95%and that the recall rate reaches 92.48%.The tea picking location method presented in this paper is more versatile and robust in complex environments.

Plant trait estimation and classification studies in plant phenotyping using machine vision - A review

Today there is a rapid development taking place in phenotyping of plants using non-destructive image based machine vision techniques.Machine vision based plant phenotyping ranges from single plant trait estimation to broad assessment of crop canopy for thousands of plants in the field.Plant phenotyping systems either use single imaging method or integrative approach signifying simultaneous use of some of the imaging techniques like visible red,green and blue(RGB)imaging,thermal imaging,chlorophyll fluorescence imaging(CFIM),hyperspectral imaging,3-dimensional(3-D)imaging or high resolution volumetric imaging.This paper provides an overview of imaging techniques and their applications in the field of plant phenotyping.This paper presents a comprehensive survey on recent machine vision methods for plant trait estimation and classification.In this paper,information about publicly available datasets is provided for uniform comparison among the state-of-the-art phenotyping methods.This paper also presents future research directions related to the use of deep learning based machine vision algorithms for structural(2-D and 3-D),physiological and temporal trait estimation,and classification studies in plants.

An improved binocular localization method for apple based on fruit detection using deep learning

Apple picking robot is now being developed as an alternative to hand picking due to a great demand for labor during apple harvest season.Accurate detection and localization of target fruit is necessary for robotic apple picking.Detection accuracy has a great influence on localization results.Although current researches on detection and localization of apples using traditional image algorithms can obtain good results under laboratory conditions,it is difficult to accurately detect and locate objects in natural field with complex environments.With the rapid development of artificial intelligence,accuracy of apple detection based on deep learning has been significantly improved.Therefore,a deep learningbased method was developed to accurately detect and locate the position of fruit.For different localization methods,binocular localization is a widely used localization method for its bionic principle and lower equipment cost.Hence,this paper proposed an improved binocular localization method for apple based on fruit detection using deep learning.First,apples of binocular images were detected by Faster R-CNN.After that,a segmentation based on chromatic aberration and chromatic aberration ratio was applied to segment apple and background pixels in bounding box of detected fruit.Furthermore,template matching with parallel polar line constraint was used to match apples in left and right images.Finally,two feature points on apples were selected to directly calculate three dimensional coordinates of feature points with the binocular localization principle.In this study,Faster R-CNN achieved an AP of 88.12%with an average detection speed of 0.32 s for an image.Meanwhile,standard deviation and localization precision of depth of two feature points on each apple were calculated to evaluate localization.Results showed that the average standard deviation and the average localization precision of 76 groups of datasets were 0.51 cm and 99.64%,respectively.Results indicated that the proposed improved binocular localization method is promising for fruit localization。

A computer vision system for early detection of anthracnose in sugar mango(Mangifera indica)based on UV-A illumination

The present work describes the development of a computer vision system for the early detection of anthracnose in sugar mango based on Ultraviolet A illumination(UV-A).Anthracnose,a disease caused by the fungus Colletotrichum sp,is commonly found in the fruit of sugar mango(Mangifera indica).It manifests as surface defects including black spots and is responsible for reducing the quality of the fruit.Consequently,it decreases its commercial value.In more detail,this study poses a system that begins with image acquisition under white and ultraviolet illumination.Furthermore,it proposes to analyze the Red,Green and Blue color information(R,G,B)of the pixels under two types of illumination,using four different methods:RGB-threshold,RGB-Linear Discriminant Analysis(RGBLDA),UV-LDA,and UV-threshold.This analysis produces an early semantic segmentation of healthy and diseased areas of the mango image.The results showed that the combination of the linear discriminant analysis(LDA)and UV-A light(called UV-LDA method)in sugar mango images allows early detection of anthracnose.Particularly,this method achieves the identification of the disease one day earlier than by an expert with respect to the scale of anthracnose severity implemented in this work.

Implementation of drone technology for farm monitoring&pesticide spraying:A review

The world receives more than 200 thousand people in a day and it is expected that the total world population will reach 9.6 billion by the year 2050.This will result in extra food demand,which can only be met from enhanced crop yield.Therefore,modernization of the agricultural sector becomes the need of the hour.There are many constraints that are responsible for the low production of crops,which can be overcome by using drone technology in the agriculture sector.This paper presents an analysis of drone technologies and their modifications with time in the agriculture sector in the last decade.The application of drones in the area of crop monitoring,and pesticide spraying for Precision Agriculture(PA)has been covered.The work done related to drone structure,multiple sensor development,innovation in spot area spraying has been presented.Moreover,the use of Artificial Intelligent(AI)and deep learning for the remote monitoring of crops has been discussed.

Research and application on corn crop identification and positioning method based on Machine vision

Weeds that grow among crops are undesirable plants and have adversely affected crop growth and yield.Therefore,the study explores corn identification and positioning methods based on machine vision.The ultra-green feature algorithm and maximum betweenclass variance method(OTSU)were used to segment maize corn,weeds,and land;the segmentation effect was significant and can meet the following shape feature extraction requirements.Finally,the identification and positioning of corn were achieved by morphological reconstruction and pixel projection histogram method.The experiment reveals that when a weeding robot travels at a speed of 1.6 km/h,the recognition accuracy can reach 94.1%.The technique used in this study is accessible for normal cases and can make a good recognition effect;the accuracy and real-time requirements of robot recognition are improved and reduced the calculation time.

Deep learning based classification of sheep behaviour from accelerometer data with imbalance

Classification of sheep behaviour from a sequence of tri-axial accelerometer data has the potential to enhance sheep management.Sheep behaviour is inherently imbalanced(e.g.,more ruminating than walking)resulting in underperforming classification for the minority activities which hold importance.Existing works have not addressed class imbalance and use traditional machine learning techniques,e.g.,Random Forest(RF).We investigated Deep Learning(DL)models,namely,Long Short Term Memory(LSTM)and Bidirectional LSTM(BLSTM),appropriate for sequential data,from imbalanced data.Two data sets were collected in normal grazing conditions using jaw-mounted and earmounted sensors.Novel to this study,alongside typical single classes,e.g.,walking,depending on the behaviours,data samples were labelled with compound classes,e.g.,walking_-grazing.The number of steps a sheep performed in the observed 10 s time window was also recorded and incorporated in the models.We designed several multi-class classification studies with imbalance being addressed using synthetic data.DL models achieved superior performance to traditional ML models,especially with augmented data(e.g.,4-Class+Steps:LSTM 88.0%,RF 82.5%).DL methods showed superior generalisability on unseen sheep(i.e.,F1-score:BLSTM 0.84,LSTM 0.83,RF 0.65).LSTM,BLSTM and RF achieved sub-millisecond average inference time,making them suitable for real-time applications.The results demonstrate the effectiveness of DL models for sheep behaviour classification in grazing conditions.The results also demonstrate the DL techniques can generalise across different sheep.The study presents a strong foundation of the development of such models for real-time animal monitoring.

Method for wheat ear counting based on frequency domain decomposition of MSVF-ISCT

Wheat ear counting is a prerequisite for the evaluation of wheat yield.A wheat ear counting method based on frequency domain decomposition is proposed in this study to improve the accuracy of wheat yield estimation.The frequency domain decomposition of wheat ear image is completed by multiscale support value filter(MSVF)combined with improved sampled contourlet transform(ISCT).Support Vector Machine(SVM)is the classic classification and regression algorithm of machine learning.MSVF based on this has strong frequency domain filtering and generalization ability,which can effectively remove the complex background,while the multi-direction characteristics of ISCT enable it to represent the contour and texture information of wheat ears.In order to improve the level of wheat yield prediction,MSVF-ISCT method is used to decompose the ear image in multiscale and multi direction in frequency domain,reduce the interference of irrelevant information,and generate the sub-band image with more abundant information components of ear feature information.Then,the ear feature is extracted by morphological operation and maximum entropy threshold segmentation,and the skeleton thinning and corner detection algorithms are used to count the results.The number of wheat ears in the image can be accurately counted.Experiments show that compared with the traditional algorithms based on spatial domain,this method significantly improves the accuracy of wheat ear counting,which can provide guidance and application for the field of agricultural precision yield estimation.

Texture-based latent space disentanglement for enhancement of a training dataset for ANN-based classification of fruit and vegetables

The capability of Convolutional Neural Networks(CNNs)for sparse representation has significant application to complex tasks like Representation Learning(RL).However,labelled datasets of sufficient size for learning this representation are not easily obtainable.The unsupervised learning capability of Variational Autoencoders(VAEs)and Generative Adversarial Networks(GANs)provide a promising solution to this issue through their capacity to learn representations for novel data samples and classification tasks.In this research,a texture-based latent space disentanglement technique is proposed to enhance learning of representations for novel data samples.A comparison is performed among different VAEs and GANs with the proposed approach for synthesis of new data samples.Two different VAE architectures are considered,a single layer dense VAE and a convolution based VAE,to compare the effectiveness of different architectures for learning of the representations.The GANs are selected based on the distance metric for disjoint distribution divergence estimation of complex representation learning tasks.The proposed texture-based disentanglement has been shown to provide a significant improvement for disentangling the process of representation learning by conditioning the random noise and synthesising texture rich images of fruit and vegetables。

Peanut drying:Effects of various drying methods on drying kinetic models,physicochemical properties,germination characteristics,and microstructure

The current work aims to explore the suitable drying technique for peanut pods which can be used for seeds or edible peanuts.Four drying methods,namely naturally-open sun drying as the control check(CK),hot air drying(HAD),pulsed vacuum drying(PVD),and radio frequency combined hot air drying(RF-HAD),were employed to dry peanut pods,and their effects on the nutritional quality attributes in terms of protein,fat,fatty acid contents,etc.,germination characteristics,microstructure,color,texture,acid value and peroxide value of peanuts were explored.Mathematical models of peanuts drying with four drying methods were also established.According to the statistical parameters including the determination coefficient(R^(2))、root mean square error(RMSE)and chi-square value(v^(2)),theWeibull model was best for predicting the moisture ratio change kinetics of peanuts during its four drying processes.There were significant differences in physicochemical indexes of peanut by different drying methods(p<0.05).Fat and oleic acid contents under RF-HAD were significantly higher than those by the other three drying methods.Compared with the naturally-open sun drying,RF-HAD reduced drying time by 76.70%and the microstructure of RF-HAD peanuts produced larger and more cavities.The RF-HAD kept better comprehensive nutritional quality,but the germination rate was reduced by 27.80%.PVD could maintain good nutritional quality and germination rate among these mechanical drying technologies.However,PVD had a longer drying time of 9.5 h than RF-HAD and HAD,and the microstructure of pulsed vacuum dried peanuts showed dense structure and less cavity.Hot air-dried peanut kernel held the highest protein(28.75%),fatty acids contents(26.11%)and germination rate(88.00%).However,peanut kernel dried by HAD showed poor qualities,such as high acid value,peroxide value and large color changes.These findings indicated RF-HAD was a promising drying technique for edible peanuts regarding the higher drying rate and better-quality preservation,while HAD was suitable for peanut seeds drying as it could well protect the germination rate.

Numerical investigation for effects of natural light and ventilation on 3D tomato body heat distribution in a Venlo greenhouse

Maintaining suitable temperature level around tomato in the greenhouse is essential for the high-quality production.However,in summer,the temperature level around the tomato is usually unclear except using a high-precision temperature imager.To solve this problem,thermal performance of 3D(three-dimensional)tomato model built based on SolidWorks was investigated by the computational fluid dynamics(CFD)simulations.To assess the effect of temperature distribution around the tomato,a simplified 3D tomato numerical model was firstly validated by a set of field measurement data.The light intensity and indoor ventilation were regarded as the mainly environment factors in the Venlo greenhouse,thermal stratification around tomatoes at different time of day was further studied.The numerical results illustrated the different temperature distribution around tomato body under different radiation intensity.It was found that ventilation could obviously adjust the temperature gradient around the tomato,and alleviate high temperature effect particularly in summer.Suitable ventilation could create a suitable thermal environment for the tomato growth.This study clearly demonstrated 3D temperature distribution around tomatoes,which is beneficial to provide the reference for accurate detection of 3D tomato temperature and appropriate thermal environment design.

Classification of weed seeds based on visual images and deep learning

Weeds are mainly spread by weed seeds being mixed with agricultural and forestry crop seeds,grain,animal hair,and other plant products,and disturb the growing environment of target plants such as crops and wild native plants.The accurate and efficient classification of weed seeds is important for the effective management and control of weeds.However,classification remains mainly dependent on destructive sampling-based manual inspection,which has a high cost and rather low flux.We considered that this problem could be solved using a nondestructive intelligent image recognition method.First,on the basis of the establishment of the image acquisition system for weed seeds,images of single weed seeds were rapidly and completely segmented,and a total of 47696 samples of 140 species of weed seeds and foreign materials remained.Then,six popular and novel deep Convolutional Neural Network(CNN)models are compared to identify the best method for intelligently identifying 140 species of weed seeds.Of these samples,33600 samples are randomly selected as the training dataset for model training,and the remaining 14096 samples are used as the testing dataset for model testing.AlexNet and GoogLeNet emerged from the quantitative evaluation as the best methods.AlexNet has strong classification accuracy and efficiency(low time consumption),and GoogLeNet has the best classification accuracy.A suitable CNN model for weed seed classification could be selected according to specific identification accuracy requirements and time costs of applications.This research is beneficial for developing a detection system for weed seeds in various applications.The resolution of taxonomic issues and problems associated with the identi-fication of these weed seeds may allow for more effective management and control.

Non-invasive sensing techniques to phenotype multiple apple tree architectures

Tree fruit architecture results from combination of the training system and pruning and thinning processes across multiple growth and development years.Further,the tree fruit architecture contributes to the light interception and improves tree growth,fruit quality,and fruit yield,in addition to easing the process of orchard management and harvest.Currently tree architectural traits are measured manually by researchers or growers,which is labor-intensive and time-consuming.In this study,the remote sensing techniques were evaluated to phenotype critical architectural traits with the final goal to assist tree fruit breeders,physiologists and growers in collecting architectural traits efficiently and in a standardized manner.For this,a consumer-grade red–green–blue(RGB)camera was used to collect apple tree side-images,while an unmanned aerial vehicle(UAV)integrated RGB camera was programmed to image tree canopy at 15 m above ground level to evaluate multiple tree fruit architectures.The sensing data were compared to ground reference data associated with tree orchard blocks within three training systems(Spindle,V-trellis,Biaxis),two rootstocks(‘WA 38 trees grafted on G41 and M9-Nic29)and two pruning methods(referred as bending and click pruning).The data were processed to extract architectural features from ground-based 2D images and UAV-based 3D digital surface model.The traits extracted from sensing data included box-counting fractal dimension(DBs),middle branch angle,number of branches,trunk basal diameter,and tree row volume(TRV).The results from ground-based sensing data indicated that there was a significant(P<0.0001)difference in DBs between Spindle and V-trellis training systems,and correlations between DBs with tree height(r=0.79)and total fruit yield per unit area(r=0.74)were significant(P<0.05).Moreover,correlations between average or total TRV and ground reference data,such as tree height and total fruit yield per unit area,were significant(P<0.05).With the reported findings,this study demonstrated the potential of sensing techniques for phenotyping tree fruit architectural traits.

Modelling reference evapotranspiration using gene expression programming and artificial neural network at Pantnagar,India

Evapotranspiration is an essential component of the hydrological cycle that is of particular interest for water resource planning.Its quantification is helpful in irrigation scheduling,water balance studies,water allocation,etc.Modelling of reference evapotranspiration(ET0)using both gene expression programming(GEP)and artificial neural network(ANN)techniques was done using the daily meteorological data of the Pantnagar region,India,from 2010 to 2019.A total of 15 combinations of inputs were used in developing the ET0 models.The model with the least number of inputs consisted of maximum and minimum air temperatures,whereas the model with the highest number of inputs consisted of maximum air temperature,minimum air temperature,mean relative humidity,number of sunshine hours,wind speed at 2mheight and extra-terrestrial radiation as inputs and with ET0 as the output for all the models.All the GEP models were developed for a single functional set and pre-defined genetic operator values,while the best structure in each ANN model was found based on the performance during the testing phase.It was found that ANN models were superior to GEP models for the estimation purpose.It was evident from the reduction in RMSE values ranging from 2%to 56%during training and testing phases in all the ANN models compared with GEP models.The ANN models showed an increase of about 0.96%to 9.72%of R2 value compared to the respective GEP models.The comparative study of these models with multiple linear regression(MLR)depicted that the ANN and GEP models were superior to MLR models.

Piezoelectric atomizer in aeroponic systems:A study of some fluid properties and optimization of operational parameters

The piezoelectric method to atomize fluids in aeroponic systems is known as a new application.In the current study,the effects of four independent variables on the misting rate were investigated.These variables were Dosage of Chemical Fertilizer(DCF),the Height of Solution above Piezoelectric Ceramic(HSPC),Piezoelectric Location at the Bottom of the Container(PLBC),and the Container Geometry(CG).Physical properties such as fluid density,viscosity,surface tension,Total Dissolved Solids(TDS),and Electrical Conductivity(EC)of the solutions were measured.Moreover,optimization was performed to determine suitable ranges of independent variables to get the optimum value of the misting rate.Results showed that the HSPC,PLBC,and CG significantly affect the total misting rate(P<1%).Besides,the DCF,HSPC,and CG have significant effects on surface evaporation at a probability level of 1%.At the temperature of 25℃and for different concentrations of fertilizer,the density,viscosity,surface tension,TDS,and EC varied between 1002-1006.5 kg/m^(3),1.009-1.395 mPas,0.07182-0.07298 N/m,253-3397 mg/L,502-6795 mS/cm,respectively.Total misting rate ranged between 0.499 and 5.720 g/min.In the outlet of the mist generation container,mass flow rate was in the range of 59.895 to 65.116 g/min.Also,theoretically calculated mist average droplet size was in the range of 2.90 to 2.92 mm for different fertilizer doses.The present study is the first research on the factors affecting the ultrasonic mist production rate in an aeroponic system.For future developments,effect of voltage,frequency,and using horn should be investigated.Besides,similar research should be done in the field of gas-phase bioreactors(mist bioreactors).

A machine vision-intelligent modelling based technique for in-line bell pepper sorting

The uniformity of appearance attributes of bell peppers is significant for consumers and food industries.To automate the sorting process of bell peppers and improve the packaging quality of this crop by detecting and separating the not likable low-color bell peppers,developing an appropriate sorting system would be of high importance and influence.According to standards and export needs,the bell pepper should be graded based on maturity levels and size to five classes.This research has been aimed to develop a machine vision-based system equipped with an intelligent modelling approach for in-line sorting bell peppers into desirable and undesirable samples,with the ability to predict the maturity level and the size of the desirable bell peppers.Multilayer perceptron(MLP)artificial neural networks(ANNs)as the nonlinear modelswere designed for that purpose.TheMLP modelswere trained and evaluated through five-fold cross-validation method.The optimum MLP classifier was compared with a linear discriminant analysis(LDA)model.The results showed that the MLP outperforms the LDA model.The processing time to classify each captured image was estimated as 0.2 s/sample,which is fast enough for in-line application.Accordingly,the optimum MLP model was integrated with a machine vision-based sorting machine,and the developed system was evaluated in the in-line phase.The performance parameters,including accuracy,precision,sensitivity,and specificity,were 93.2%,86.4%,84%,and 95.7%,respectively.The total sorting rate of the bell pepper was also measured as approximately 3000 samples/h.

Sensitivity analysis of the DehumReq model to evaluate the impact of predominant factors on dehumidification requirement of greenhouses in cold regions

In this study,the sensitivity of a novel dehumidification requirement model(DehumReq)is analyzed to evaluate the effect of the predominant factors on the dehumidification needs of the greenhouses.The hourly dehumidification demand and sensitivity coefficient(SC)are estimated for three different seasons:warm(July),mild(May),and cold(November),by using the local sensitivity analysis method.Based on SC values,the solar radiation,air exchange,leaf area index(LAI),and indoor setpoints(temperature,relative humidity(RH),and water vapor partial pressure(WVPP))have significant impact on the dehumidifi-cation needs,and the impact varies from season to season.Most parameters have a higher SC in summer,whereas solar radiation and LAI have a higher SC in mild season.The dehumidification load increases 4 times of its base value with increasing solar radiation by 200 W/m^(2),and the highest LAI(10)caused 5 times increment of the load.The changing of WVPP from its base value(1.5 kPa)to maximum(2.9 kPa)reduces the load 70%in summer.Air exchange was found to be the most crucial parameter because it is the main dehumidification approach that has a large range and is easily adjustable for any greenhouses.Sufficient air exchange by ventilation or infiltration will reduce the dehumidification load to zero in May and November and minimizes it to only nighttime load in July.For the other parameters,higher ambient air RH and indoor air speed will result in higher the dehumidification load;whereas higher inner surface condensation will result in lower dehumidifi-cation load.The result of this study will assist in the selection of the most efficient moisture control strategies and techniques for greenhouse humidity control.

A new approach to learning and recognizing leaf diseases from individual lesions using convolutional neural networks

Leaf disease recognition using image processing and deep learning techniques is currently a vibrant research area.Most studies have focused on recognizing diseases from images of whole leaves.This approach limits the resulting models’ability to estimate leaf disease severity or identify multiple anomalies occurring on the same leaf.Recent studies have demonstrated that classifying leaf diseases based on individual lesions greatly enhances disease recognition accuracy.In those studies,however,the lesions were laboriously cropped by hand.This study proposes a semi-automatic algorithm that facilitates the fast and efficient preparation of datasets of individual lesions and leaf image pixel maps to overcome this problem.These datasets were then used to train and test lesion classifier and semantic segmentation Convolutional Neural Network(CNN)models,respectively.We report that GoogLeNet’s disease recognition accuracy improved by more than 15%when diseases were recognized from lesion images compared to when disease recognition was done using images of whole leaves.A CNN model which performs semantic segmentation of both the leaf and lesions in one pass is also proposed in this paper.The proposed KijaniNet model achieved state-of-the-art segmentation performance in terms of mean Intersection over Union(mIoU)score of 0.8448 and 0.6257 for the leaf and lesion pixel classes,respectively.In terms of mean boundary F1 score,the KijaniNet model attained 0.8241 and 0.7855 for the two pixel classes,respectively.Lastly,a fully automatic algorithm for leaf disease recognition from individual lesions is proposed.The algorithm employs the semantic segmentation network cascaded to a GoogLeNet classifier for lesion-wise disease recognition.The proposed fully automatic algorithm outperforms competing methods in terms of its superior segmentation and classification performance despite being trained on a small dataset.