Agricultural machinery typically requires lower limb actuation forces for operations such as treadling,pedaling and tractor based.However,limited systems exist for assessment of such forces that have ergonomic influen...Agricultural machinery typically requires lower limb actuation forces for operations such as treadling,pedaling and tractor based.However,limited systems exist for assessment of such forces that have ergonomic influence.This study,therefore developed and evaluated a single board computer integrated foot transducer(IFT)and autonomous data logging and visualization systemtomonitor dynamic lower limb exerted forces.The systemconsists of custom developed load sensors sandwiched into foot shaped units that fit operator's both feet.Stamped forces at crank angles for operations typical to pedaling while at height(above ground level)for operation representing typical treadling operations were recorded on-board amemory card and displayed on a liquid crystal display.Evaluations were conducted by imposing external loads that significantly increased(p b 0.05)the foot exerted forces.Force trends were periodic with peaks of 73,85,110.5 and 145.4 N for left foot and 41,50,131.7 and 145.4 N for right foot at loads of 10,30,50 and 70 N,respectively during pedaling operations.Similarly,the left lower actuation limb exerted forces of 139,249 and 255 N at 5,10 and 15 N of imposed loads,respectively during treadling operation.System was also evaluated for tractor operations and exerted forces ranged from 92 to 164 and 107–176 N for clutch pedal engagement at lower to higher tractor speeds on farm and tarmacadam roads,respectively.Similarly,for brake pedal engagement,such forces ranged from106 to 173 and 120–204 N on farm and tarmacadamroads.These forces varied significantly at different forward speeds.Results suggest potential of such system for foot exerted force assessments typical to agricultural machinery systems in real field.Designsmay be evaluated or reconsidered tominimizemusculoskeletal disorder risks during prolonged operations.Work-rest schedules protocols can be developed by ergonomists for safe,efficient and comfortable operations.展开更多
Rapid and accurate canopy attributes estimation is highly critical in fruit crops production management as this information can be used for canopy and crop load management as well as to develop nutrient/chemical presc...Rapid and accurate canopy attributes estimation is highly critical in fruit crops production management as this information can be used for canopy and crop load management as well as to develop nutrient/chemical prescription application maps.However,the existing ground based canopy sensing and attribute estimation methods are laborious and often involve complexity with field data collection and analysis.Manual methods can be subjec-tive as well.Therefore,this study explores aerial photogrammetry based method of tree–row–volume(TRV),leaf–wall–area(LWA),canopy volume(CV)and canopy cover(CC)esti-mation for grapevine and apple canopies.Remote sensing data was collected using a con-sumer–grade small unmanned aerial system(UAS)with an RGB imaging sensor flying at different flight altitudes i.e.,15 m(Ground sampling distance,GSD=0.45 cm pixel^(-1) at 65°sensor inclination),30 m(0.90 and 0.85 cm pixel^(-1) at 65°and 75°,respectively),45 m(1.35 and 1.27 cm pixel^(-1) at 65°and 75°,respectively)and 60 m(1.81 and 1.69 cm pixel^(-1) at 65°and 75°,respectively).Crop surface model(CSM)was derived from such data to esti-mate canopy height,width and foliage vigor,which are further used to estimate TRV,LWA,CV and CC.The ground measured and aerial imagery estimated TRV had a strong relation-ship with the data collected at the lowest GSD within grapevine canopies(R^(2)=0.77 at 0.45 cm pixel^(-1)as well as for apple canopies(R^(2)=0.82 at 0.90 cm pixel^(-1).Similar trends were observed for the LWA(R^(2)=0.77 and 0.86),CV(R^(2)=0.43 and 0.64)and CC(R^(2)=0.61 and 0.68)estimates for grapevine and apple canopies,respectively.Increasing GSD(≥0.45 cm pixel^(-1) in grapevine and≥0.90 cm pixel^(-1) in apple)resulted in a weak relationship between ground measurements and aerial imagery data-based estimates for grapevines(R^(2)=0.36)and apple canopies(R^(2)=0.39–0.78).Overall,the aerial flights with lower GSD and double grid missions with RGB imaging sensor in 65°orientation aided in the development of site–specific high–quality canopy vigor maps that can be used in precision crop inputs management related decision making.展开更多
This study evaluated corn(Zea mays var.indentata)canopy vigor and temperature variations using small unmanned aerial system(UAS)based spatiotemporal imagery data.The key objective was to develop understanding on site-...This study evaluated corn(Zea mays var.indentata)canopy vigor and temperature variations using small unmanned aerial system(UAS)based spatiotemporal imagery data.The key objective was to develop understanding on site-specific suitability of the Mid Elevation Spray Application(MESA)and Low Elevation Spray Application(LESA)sprinkler systems in irrigating corn crop and potential water as well as energy savings.Aerial data was collected through small UAS flights at 100m height above ground level and on 49,59,65,77,105,114,134 days after planting(DAP).Small UAS had 5-band multispectral and radiometric thermal imager on-board the platform.Custom image processing algorithms were developed to extract various vegetation indices(Normalized difference vegetation index[NDVI],Green NDVI[GNDVI]and Normalized difference red edge[NDRE])and canopy temperature maps from the imagery data.Two sample T-test was performed on extracted data to understand significant difference,at 5%level,within the LESA and MESA treatments.For both the irrigation techniques,the crop vigor increased in the early growth stage(49,65 DAP),peaked in the mid growth stage(77,105,114 DAP)and then decreased in the late growth stage(134 DAP).The MESA irrigated sections had higher vigor(NDVI,GNDVI,NDRE),but not significant,compared to LESA throughout the season.Similarly,the MESA irrigated areas had significantly(0.61–2.07C)cooler canopies than LESA.Such results were anticipated,in part due to the issues with the sprinkler heads used in LESA.The heads were being pulled off in the corn field,causing the weighted hose to damage the crop.A different kind of sprinkler head was used after this incident.However,some strips of corn had already been damaged.Overall,this study results confirm suitability of aerial imagery data in evaluating pertinent irrigation treatments.Additional season’s data would be needed to clearly understand which technique(LESA or MESA)is more suitable for irrigating corn crop in the central part of state of Washington.展开更多
Accurate and fast detection of abnormal hydroponic lettuce leaves is primary technology for robotic sorting.Yellow and rotten leaves are main types of abnormal leaves in hydroponic lettuce.This study aims to demonstra...Accurate and fast detection of abnormal hydroponic lettuce leaves is primary technology for robotic sorting.Yellow and rotten leaves are main types of abnormal leaves in hydroponic lettuce.This study aims to demonstrate a feasibility of detecting yellow and rotten leaves of hydroponic lettuce by machine learning models,i.e.Multiple Linear Regression(MLR),K-Nearest Neighbor(KNN),and Support Vector Machine(SVM).One-way analysis of variance was applied to reduce RGB,HSV,and L*a*b*features number of hydroponic lettuce images.Image binarization,image mask,and image filling methods were employed to segment hydroponic lettuce from an image for models testing.Results showed that G,H,and a*were selected from RGB,HSV,and L*a*b*for training models.It took about 20.25 s to detect an image with 30244032 pixels by KNN,which was much longer than MLR(0.61 s)and SVM(1.98 s).MLR got detection accuracies of 89.48%and 99.29%for yellow and rotten leaves,respectively,while SVM reached 98.33%and 97.91%,respectively.SVM was more robust than MLR in detecting yellow and rotten leaves of hydroponic.Thus,it was possible for abnormal hydroponic lettuce leaves detection by machine learning methods.展开更多
The pinto bean is one of widely consumed legume crop that constitutes over 42%of the U.S dry bean production.However,limited studies have been conducted in past to assess its quantitative and qualitative yield potenti...The pinto bean is one of widely consumed legume crop that constitutes over 42%of the U.S dry bean production.However,limited studies have been conducted in past to assess its quantitative and qualitative yield potentials.Emerging remote sensing technologies can help in such assessment.Therefore,this study evaluates the role of ground-based multispectral imagery derived vegetation indices(VIs)for irrigated the pinto bean stress and yield assessments.Studied were eight cultivars of the pinto bean grown under conventional and strip tillage treatments and irrigated at 52%and 100%of required evapotranspiration.Imagery data was acquired using a five-band multispectral imager at early,mid and late growth stages.Commonly used 25 broadband VIs were derived to capture crop stress traits and yield potential.Principal component analysis and Spearman’s rank correlation tests were conducted to identify key VIs and their correlation(rs)with abiotic stress at each growth stage.Transformed difference vegetation index,nonlinear vegetation index(NLI),modified NLI and infrared percentage vegetation index(IPVI)were consistent in accounting the stress response and crop yield at all growth stages(rs>0.60,coefficient of determination(R2):0.50–0.56,P<0.05).Ten other VIs significantly accounted for crop stress at early and late stages.Overall,identified key VIs may be helpful to growers for precise crop management decision making and breeders for crop stress response and yield assessments.展开更多
Strawberry is one of the popular fruits with numerous nutrients.The ripeness of this fruits was estimated using the hyperspectral imaging(HSI)system in field and laboratory conditions in this study.Strawberry at early...Strawberry is one of the popular fruits with numerous nutrients.The ripeness of this fruits was estimated using the hyperspectral imaging(HSI)system in field and laboratory conditions in this study.Strawberry at early ripe and ripe stageswere collected HSI data,coveredwavelength ranges from370 to 1015 nm.Spectral featurewavelengths were selected using the sequential feature selection(SFS)algorithm.Two wavelengths selected for field(530 and 604 nm)and laboratory(528 and 715 nm)samples,respectively.Then,reliability of such spectral featureswas validated based on support vectormachine(SVM)classifier.Performance of SVMclassification models had good resultswith receiver operating characteristic values for samples under both field and laboratory conditions higher than 0.95.Meanwhile,the spatial feature images were extracted from the spectral feature wavelength and the first three principal components for laboratory samples.Pretrained AlexNet convolutional neural network(CNN)was used to classify the early ripe and ripe strawberry samples,which obtained the accuracy of 98.6%for test dataset.The above results indicated real-time HSI system was promising for estimating strawberry ripeness under field and laboratory conditions,which could be a potential application technique for evaluating the harvesting time management for farmers and producers.展开更多
Heat and light stress causes sunburn to the maturing apple fruits and results in crop production and quality losses.Typically,when the fruit surface temperature(FST)rises above critical limits for a prolonged duration...Heat and light stress causes sunburn to the maturing apple fruits and results in crop production and quality losses.Typically,when the fruit surface temperature(FST)rises above critical limits for a prolonged duration,the fruit may suffer several physiological disorders including sunburn.To manage apple sunburn,monitoring FST is critical and our group at Washington State University is developing a noncontact smart sensing system that integrates thermal infrared and visible imaging sensors for real time FST monitoring.Pertinent system needs to perform in-field imagery data analysis onboard a single board computer with processing unit that has limited computational resources.Therefore,key objective of this study was to develop a novel image processing algorithm optimized to use available resources of a single board computer.Algorithm logic flow includes color space transformation,k-means++classification and morphological operators prior to fruit segmentation and FST estimation.The developed algorithm demonstrated the segmentation accuracy of 57.78%(missing error=12.09%and segmentation error=0.13%).This aided successful apple FST estimation that was 10–18C warmer than ambient air temperature.Moreover,algorithm reduced the imagery data processing time cost of the smart sensing systemfrom 87 s to 44 s using image compression approach.展开更多
Estimation of fruit size in tree fruit crops is essential for selective robotic harvesting and crop-load estimation.Machine vision systems for fruit detection and localization have been studied widely for robotic harv...Estimation of fruit size in tree fruit crops is essential for selective robotic harvesting and crop-load estimation.Machine vision systems for fruit detection and localization have been studied widely for robotic harvesting and crop-load estimation.However,only a few studies have been carried out to estimate fruit size in orchards using machine vision systems.This study was carried out to develop a machine vision system consisting of a color CCD camera and a time-of-flight(TOF)light-based 3D camera for estimating apple size in tree canopies.As a measure of fruit size,the major axis(longest axis)was estimated based on(i)the 3D coordinates of pixels on corresponding apple surfaces,and(ii)the 2D size of individual pixels within apple surfaces.In the 3D coordinates-based method,the distance between pairs of pixels within apple regions were calculated using 3D coordinates,and the maximum distance between all pixel pairs within an apple region was estimated to be the major axis.The accuracy of estimating the major axis using 3D coordinates was 69.1%.In the pixel-size-based method,the physical sizes of pixels were estimated using a calibration model developed based on pixel coordinates and the distance to pixels from the camera.The major axis length was then estimated by summing the size of individual pixels along the major axis of the fruit.The accuracy of size estimation increased to 84.8%when the pixel size-based method was used.The results showed the potential for estimating fruit size in outdoor environments using a 3D machine vision system.展开更多
Due to the illumination,complex background,and occlusion of the litchi fruits,the accurate detection of litchi in the field is extremely challenging.In order to solve the problem of the low recognition rate of litchi-...Due to the illumination,complex background,and occlusion of the litchi fruits,the accurate detection of litchi in the field is extremely challenging.In order to solve the problem of the low recognition rate of litchi-picking robots in field conditions,this study was inspired by the ideas of ResNet and dense convolution and proposed an improved feature-extraction network model named“YOLOv3_Litchi”,combining dense connections and residuals for the detection of litchis.Firstly,based on the traditional YOLOv3 deep convolution neural network and regression detection,the idea of residuals was to be put into the feature-extraction network to effectively avoid the problem of decreasing detection accuracy due to the excessive depths of the network layers.Secondly,under the premise of a good receptive field and high detection accuracy,the large convolution kernel was replaced by a small convolution kernel in the shallow layer of the network,thereby effectively reducing the model parameters.Finally,the idea of feature pyramid was used to design the network to identify the small target litchi to ensure that the shallow features were not lost and simultaneously reduced the model parameters.Experimental results show that the improved YOLOv3_Litchi model achieved better results than the classic YOLOv3_DarkNet-53 model and the YOLOv3_Tiny model.The mean average precision(mAP)score was 97.07%,which was higher than the 95.18%mAP of the YOLOv3_DarkNet-53 model and the 94.48%mAP of the YOLOv3_Tiny model.The frame frequency was 58 fps,which was higher than 29 fps of the YOLOv3_DarkNet-53 model.Compared with the classic Faster R-CNN model with the feature-extraction network VGG16,the mAP was increased by 1%,and the FPS advantage was obvious.Compared with the classic single shot multibox detector(SSD)model,both the accuracy and the running efficiency were improved.The results show that the improved YOLOv3_Litchi model had stronger robustness,higher detection accuracy,and less computational complexity for the identification of litchi in the field conditions,which should be helpful for litchi orchard precision management.展开更多
Lodging occurs when the crop canopy is too heavy for the strength of the stem and it fallsover onto the ground. This decreases crop yield and quality, and it makes harvest difficult.A research experiment was set up in...Lodging occurs when the crop canopy is too heavy for the strength of the stem and it fallsover onto the ground. This decreases crop yield and quality, and it makes harvest difficult.A research experiment was set up in a spearmint field on a center pivot with mid elevationspray application (MESA) overhead sprinklers, where the water was applied from a “midelevation” of 2 m above the ground level (AGL), and low elevation precision application(LEPA) sprinklers, where the water was emitted directly onto the soil surface through draghoses without wetting the crop canopy. Every-other span of this full-size center pivot wasconfigured with MESA and LEPA sprinklers alternatively. In 2018, imagery was collectedwith an unmanned aerial vehicle (UAV) from a cross section of this field. In 2019, a crosssection was again collected, but in addition UAV imagery was collected from marked lodgedand un-lodged areas of the field to validate the lodging detection method. These UAV-basedimagery data were captured with a ground sample distance (GSD) of 0.03 m. This researchintroduces using the texture feature, which is based on image entropy, was used to evaluate the degree of lodging. The results from 2018 showed that the average entropy of thegrayscale image from LEPA (5.5 (mean) ± 0.27 (standard deviation)) was significantly(P < 0.0001) greater than the average entropy (5.0 ± 0.25) of MESA. Also, the entropy valueextracted from the images in 2019 from the marked un-lodged locations were significantlyhigher compared to that of the lodged areas. Overall, the LEPA irrigation treatment was significantly less lodged compared to MESA. Moreover, the entropy value, or texture feature, isa viable method for estimating lodging using low altitude RGB imagery.展开更多
文摘Agricultural machinery typically requires lower limb actuation forces for operations such as treadling,pedaling and tractor based.However,limited systems exist for assessment of such forces that have ergonomic influence.This study,therefore developed and evaluated a single board computer integrated foot transducer(IFT)and autonomous data logging and visualization systemtomonitor dynamic lower limb exerted forces.The systemconsists of custom developed load sensors sandwiched into foot shaped units that fit operator's both feet.Stamped forces at crank angles for operations typical to pedaling while at height(above ground level)for operation representing typical treadling operations were recorded on-board amemory card and displayed on a liquid crystal display.Evaluations were conducted by imposing external loads that significantly increased(p b 0.05)the foot exerted forces.Force trends were periodic with peaks of 73,85,110.5 and 145.4 N for left foot and 41,50,131.7 and 145.4 N for right foot at loads of 10,30,50 and 70 N,respectively during pedaling operations.Similarly,the left lower actuation limb exerted forces of 139,249 and 255 N at 5,10 and 15 N of imposed loads,respectively during treadling operation.System was also evaluated for tractor operations and exerted forces ranged from 92 to 164 and 107–176 N for clutch pedal engagement at lower to higher tractor speeds on farm and tarmacadam roads,respectively.Similarly,for brake pedal engagement,such forces ranged from106 to 173 and 120–204 N on farm and tarmacadamroads.These forces varied significantly at different forward speeds.Results suggest potential of such system for foot exerted force assessments typical to agricultural machinery systems in real field.Designsmay be evaluated or reconsidered tominimizemusculoskeletal disorder risks during prolonged operations.Work-rest schedules protocols can be developed by ergonomists for safe,efficient and comfortable operations.
文摘Rapid and accurate canopy attributes estimation is highly critical in fruit crops production management as this information can be used for canopy and crop load management as well as to develop nutrient/chemical prescription application maps.However,the existing ground based canopy sensing and attribute estimation methods are laborious and often involve complexity with field data collection and analysis.Manual methods can be subjec-tive as well.Therefore,this study explores aerial photogrammetry based method of tree–row–volume(TRV),leaf–wall–area(LWA),canopy volume(CV)and canopy cover(CC)esti-mation for grapevine and apple canopies.Remote sensing data was collected using a con-sumer–grade small unmanned aerial system(UAS)with an RGB imaging sensor flying at different flight altitudes i.e.,15 m(Ground sampling distance,GSD=0.45 cm pixel^(-1) at 65°sensor inclination),30 m(0.90 and 0.85 cm pixel^(-1) at 65°and 75°,respectively),45 m(1.35 and 1.27 cm pixel^(-1) at 65°and 75°,respectively)and 60 m(1.81 and 1.69 cm pixel^(-1) at 65°and 75°,respectively).Crop surface model(CSM)was derived from such data to esti-mate canopy height,width and foliage vigor,which are further used to estimate TRV,LWA,CV and CC.The ground measured and aerial imagery estimated TRV had a strong relation-ship with the data collected at the lowest GSD within grapevine canopies(R^(2)=0.77 at 0.45 cm pixel^(-1)as well as for apple canopies(R^(2)=0.82 at 0.90 cm pixel^(-1).Similar trends were observed for the LWA(R^(2)=0.77 and 0.86),CV(R^(2)=0.43 and 0.64)and CC(R^(2)=0.61 and 0.68)estimates for grapevine and apple canopies,respectively.Increasing GSD(≥0.45 cm pixel^(-1) in grapevine and≥0.90 cm pixel^(-1) in apple)resulted in a weak relationship between ground measurements and aerial imagery data-based estimates for grapevines(R^(2)=0.36)and apple canopies(R^(2)=0.39–0.78).Overall,the aerial flights with lower GSD and double grid missions with RGB imaging sensor in 65°orientation aided in the development of site–specific high–quality canopy vigor maps that can be used in precision crop inputs management related decision making.
基金This project was funded in part by State ofWashingtonWater Research Center and USDA National Institute for Food and Agriculture Projects#WNP00745 and WNP0839.Authors would like to acknowledge help of Rajeev Sinha,Haitham Bahlol and Azeem Khan in completion of this study.
文摘This study evaluated corn(Zea mays var.indentata)canopy vigor and temperature variations using small unmanned aerial system(UAS)based spatiotemporal imagery data.The key objective was to develop understanding on site-specific suitability of the Mid Elevation Spray Application(MESA)and Low Elevation Spray Application(LESA)sprinkler systems in irrigating corn crop and potential water as well as energy savings.Aerial data was collected through small UAS flights at 100m height above ground level and on 49,59,65,77,105,114,134 days after planting(DAP).Small UAS had 5-band multispectral and radiometric thermal imager on-board the platform.Custom image processing algorithms were developed to extract various vegetation indices(Normalized difference vegetation index[NDVI],Green NDVI[GNDVI]and Normalized difference red edge[NDRE])and canopy temperature maps from the imagery data.Two sample T-test was performed on extracted data to understand significant difference,at 5%level,within the LESA and MESA treatments.For both the irrigation techniques,the crop vigor increased in the early growth stage(49,65 DAP),peaked in the mid growth stage(77,105,114 DAP)and then decreased in the late growth stage(134 DAP).The MESA irrigated sections had higher vigor(NDVI,GNDVI,NDRE),but not significant,compared to LESA throughout the season.Similarly,the MESA irrigated areas had significantly(0.61–2.07C)cooler canopies than LESA.Such results were anticipated,in part due to the issues with the sprinkler heads used in LESA.The heads were being pulled off in the corn field,causing the weighted hose to damage the crop.A different kind of sprinkler head was used after this incident.However,some strips of corn had already been damaged.Overall,this study results confirm suitability of aerial imagery data in evaluating pertinent irrigation treatments.Additional season’s data would be needed to clearly understand which technique(LESA or MESA)is more suitable for irrigating corn crop in the central part of state of Washington.
基金the Science and Technology Program in Yulin City of China(CXY-2020-076,CXY-2019-129)Youth Science and Technology Nova Program in Shaanxi Province of China(2021KJXX-94)+1 种基金Key Research and Development Program of Shaanxi(2021NY-135)Recruitment Program of High-End Foreign Experts of the State Administration of Foreign Experts Affairs,Ministry of Science and Technology,China(G20200027075)。
文摘Accurate and fast detection of abnormal hydroponic lettuce leaves is primary technology for robotic sorting.Yellow and rotten leaves are main types of abnormal leaves in hydroponic lettuce.This study aims to demonstrate a feasibility of detecting yellow and rotten leaves of hydroponic lettuce by machine learning models,i.e.Multiple Linear Regression(MLR),K-Nearest Neighbor(KNN),and Support Vector Machine(SVM).One-way analysis of variance was applied to reduce RGB,HSV,and L*a*b*features number of hydroponic lettuce images.Image binarization,image mask,and image filling methods were employed to segment hydroponic lettuce from an image for models testing.Results showed that G,H,and a*were selected from RGB,HSV,and L*a*b*for training models.It took about 20.25 s to detect an image with 30244032 pixels by KNN,which was much longer than MLR(0.61 s)and SVM(1.98 s).MLR got detection accuracies of 89.48%and 99.29%for yellow and rotten leaves,respectively,while SVM reached 98.33%and 97.91%,respectively.SVM was more robust than MLR in detecting yellow and rotten leaves of hydroponic.Thus,it was possible for abnormal hydroponic lettuce leaves detection by machine learning methods.
基金This work was supported in part by USDA National Institute for Food and Agriculture Projects WNP00745,WNP00839 and from the Feed the Future Innovation Lab for Climate-Resilient Beans Project#AID-OAA-A-13-00077.We also thank Dr.Lynden Porter,Dr.Manoj Karkee,Mr.Encarnacion Rivera and Mr.Treva Anderson for their technical support.
文摘The pinto bean is one of widely consumed legume crop that constitutes over 42%of the U.S dry bean production.However,limited studies have been conducted in past to assess its quantitative and qualitative yield potentials.Emerging remote sensing technologies can help in such assessment.Therefore,this study evaluates the role of ground-based multispectral imagery derived vegetation indices(VIs)for irrigated the pinto bean stress and yield assessments.Studied were eight cultivars of the pinto bean grown under conventional and strip tillage treatments and irrigated at 52%and 100%of required evapotranspiration.Imagery data was acquired using a five-band multispectral imager at early,mid and late growth stages.Commonly used 25 broadband VIs were derived to capture crop stress traits and yield potential.Principal component analysis and Spearman’s rank correlation tests were conducted to identify key VIs and their correlation(rs)with abiotic stress at each growth stage.Transformed difference vegetation index,nonlinear vegetation index(NLI),modified NLI and infrared percentage vegetation index(IPVI)were consistent in accounting the stress response and crop yield at all growth stages(rs>0.60,coefficient of determination(R2):0.50–0.56,P<0.05).Ten other VIs significantly accounted for crop stress at early and late stages.Overall,identified key VIs may be helpful to growers for precise crop management decision making and breeders for crop stress response and yield assessments.
基金This researchwas supported by National Natural Science Foundation of China(Nos.31701325,31671632)This work was also funded by China Scholarship Council(No.201709135004)Post-doctor Fund of Jiangsu Province.
文摘Strawberry is one of the popular fruits with numerous nutrients.The ripeness of this fruits was estimated using the hyperspectral imaging(HSI)system in field and laboratory conditions in this study.Strawberry at early ripe and ripe stageswere collected HSI data,coveredwavelength ranges from370 to 1015 nm.Spectral featurewavelengths were selected using the sequential feature selection(SFS)algorithm.Two wavelengths selected for field(530 and 604 nm)and laboratory(528 and 715 nm)samples,respectively.Then,reliability of such spectral featureswas validated based on support vectormachine(SVM)classifier.Performance of SVMclassification models had good resultswith receiver operating characteristic values for samples under both field and laboratory conditions higher than 0.95.Meanwhile,the spatial feature images were extracted from the spectral feature wavelength and the first three principal components for laboratory samples.Pretrained AlexNet convolutional neural network(CNN)was used to classify the early ripe and ripe strawberry samples,which obtained the accuracy of 98.6%for test dataset.The above results indicated real-time HSI system was promising for estimating strawberry ripeness under field and laboratory conditions,which could be a potential application technique for evaluating the harvesting time management for farmers and producers.
基金This project was funded in part by NSF/USDA-NIFA Cyber Physical Systems and USDA-NIFA WNP0745 projects.The author extends their gratitude to Dr.Sindhuja Sankaran and Mr.Chongyuan Zhang of Washington State University for their assistance in completion of this study.
文摘Heat and light stress causes sunburn to the maturing apple fruits and results in crop production and quality losses.Typically,when the fruit surface temperature(FST)rises above critical limits for a prolonged duration,the fruit may suffer several physiological disorders including sunburn.To manage apple sunburn,monitoring FST is critical and our group at Washington State University is developing a noncontact smart sensing system that integrates thermal infrared and visible imaging sensors for real time FST monitoring.Pertinent system needs to perform in-field imagery data analysis onboard a single board computer with processing unit that has limited computational resources.Therefore,key objective of this study was to develop a novel image processing algorithm optimized to use available resources of a single board computer.Algorithm logic flow includes color space transformation,k-means++classification and morphological operators prior to fruit segmentation and FST estimation.The developed algorithm demonstrated the segmentation accuracy of 57.78%(missing error=12.09%and segmentation error=0.13%).This aided successful apple FST estimation that was 10–18C warmer than ambient air temperature.Moreover,algorithm reduced the imagery data processing time cost of the smart sensing systemfrom 87 s to 44 s using image compression approach.
基金supported in part by the USDA’s Hatch and Multistate Project Funds(Accession Nos.1005756 and 1001246)。
文摘Estimation of fruit size in tree fruit crops is essential for selective robotic harvesting and crop-load estimation.Machine vision systems for fruit detection and localization have been studied widely for robotic harvesting and crop-load estimation.However,only a few studies have been carried out to estimate fruit size in orchards using machine vision systems.This study was carried out to develop a machine vision system consisting of a color CCD camera and a time-of-flight(TOF)light-based 3D camera for estimating apple size in tree canopies.As a measure of fruit size,the major axis(longest axis)was estimated based on(i)the 3D coordinates of pixels on corresponding apple surfaces,and(ii)the 2D size of individual pixels within apple surfaces.In the 3D coordinates-based method,the distance between pairs of pixels within apple regions were calculated using 3D coordinates,and the maximum distance between all pixel pairs within an apple region was estimated to be the major axis.The accuracy of estimating the major axis using 3D coordinates was 69.1%.In the pixel-size-based method,the physical sizes of pixels were estimated using a calibration model developed based on pixel coordinates and the distance to pixels from the camera.The major axis length was then estimated by summing the size of individual pixels along the major axis of the fruit.The accuracy of size estimation increased to 84.8%when the pixel size-based method was used.The results showed the potential for estimating fruit size in outdoor environments using a 3D machine vision system.
基金This work was financially supported by the National Natural Science Foundation of China(Grant No.32071912,No.61863011,No.31701325,No.31571568,No.31570180)the Guangzhou Science and Technology Project(Grant No.202002020016,No.202102080337)+4 种基金the Natural Science Foundation of Guangdong Province(Grant No.2018A030313330,No.2020A1515010793)the Second Batch of Industry-Education Cooperation Collaborative Projects in 2019,Ministry of Education(Grant No.201902062040)the Guangzhou Key Laboratory of Intelligent Agriculture(Grant No.201902010081)the Project of Rural Revitalization Strategy in Guangdong Province(Grant No.2020KJ261)the Applied Science and Technology Special Fund Project,Meizhou,China(Grant No.2019B0201005).
文摘Due to the illumination,complex background,and occlusion of the litchi fruits,the accurate detection of litchi in the field is extremely challenging.In order to solve the problem of the low recognition rate of litchi-picking robots in field conditions,this study was inspired by the ideas of ResNet and dense convolution and proposed an improved feature-extraction network model named“YOLOv3_Litchi”,combining dense connections and residuals for the detection of litchis.Firstly,based on the traditional YOLOv3 deep convolution neural network and regression detection,the idea of residuals was to be put into the feature-extraction network to effectively avoid the problem of decreasing detection accuracy due to the excessive depths of the network layers.Secondly,under the premise of a good receptive field and high detection accuracy,the large convolution kernel was replaced by a small convolution kernel in the shallow layer of the network,thereby effectively reducing the model parameters.Finally,the idea of feature pyramid was used to design the network to identify the small target litchi to ensure that the shallow features were not lost and simultaneously reduced the model parameters.Experimental results show that the improved YOLOv3_Litchi model achieved better results than the classic YOLOv3_DarkNet-53 model and the YOLOv3_Tiny model.The mean average precision(mAP)score was 97.07%,which was higher than the 95.18%mAP of the YOLOv3_DarkNet-53 model and the 94.48%mAP of the YOLOv3_Tiny model.The frame frequency was 58 fps,which was higher than 29 fps of the YOLOv3_DarkNet-53 model.Compared with the classic Faster R-CNN model with the feature-extraction network VGG16,the mAP was increased by 1%,and the FPS advantage was obvious.Compared with the classic single shot multibox detector(SSD)model,both the accuracy and the running efficiency were improved.The results show that the improved YOLOv3_Litchi model had stronger robustness,higher detection accuracy,and less computational complexity for the identification of litchi in the field conditions,which should be helpful for litchi orchard precision management.
文摘Lodging occurs when the crop canopy is too heavy for the strength of the stem and it fallsover onto the ground. This decreases crop yield and quality, and it makes harvest difficult.A research experiment was set up in a spearmint field on a center pivot with mid elevationspray application (MESA) overhead sprinklers, where the water was applied from a “midelevation” of 2 m above the ground level (AGL), and low elevation precision application(LEPA) sprinklers, where the water was emitted directly onto the soil surface through draghoses without wetting the crop canopy. Every-other span of this full-size center pivot wasconfigured with MESA and LEPA sprinklers alternatively. In 2018, imagery was collectedwith an unmanned aerial vehicle (UAV) from a cross section of this field. In 2019, a crosssection was again collected, but in addition UAV imagery was collected from marked lodgedand un-lodged areas of the field to validate the lodging detection method. These UAV-basedimagery data were captured with a ground sample distance (GSD) of 0.03 m. This researchintroduces using the texture feature, which is based on image entropy, was used to evaluate the degree of lodging. The results from 2018 showed that the average entropy of thegrayscale image from LEPA (5.5 (mean) ± 0.27 (standard deviation)) was significantly(P < 0.0001) greater than the average entropy (5.0 ± 0.25) of MESA. Also, the entropy valueextracted from the images in 2019 from the marked un-lodged locations were significantlyhigher compared to that of the lodged areas. Overall, the LEPA irrigation treatment was significantly less lodged compared to MESA. Moreover, the entropy value, or texture feature, isa viable method for estimating lodging using low altitude RGB imagery.