Joe Mari Maja

Huanglongbing (Citrus Greening) Detection Using Visible, Near Infrared and Thermal Imaging Techniques

This study demonstrates the applicability of visible-near infrared and thermal imaging for detection of Huanglongbing (HLB) disease in citrus trees. Visible-near infrared (440–900 nm) and thermal infrared spectral reflectance data were collected from individual healthy and HLB-infected trees. Data analysis revealed that the average reflectance values of the healthy trees in the visible region were lower than those in the near infrared region, while the opposite was the case for HLB-infected trees. Moreover, 560 nm, 710 nm, and thermal band showed maximum class separability between healthy and HLB-infected groups among the evaluated visible-infrared bands. Similarly, analysis of several vegetation indices indicated that the normalized difference vegetation index (NDVI), Vogelmann red-edge index (VOG) and modified red-edge simple ratio (mSR) demonstrated good class separability between the two groups. Classification studies using average spectral reflectance values from the visible, near infrared, and thermal bands (13 spectral features) as input features indicated that an average overall classification accuracy of about 87%, with 89% specificity and 85% sensitivity could be achieved with classification models such as support vector machine for trees with symptomatic leaves.

Real-time obstacle avoidance algorithm for visual navigation

A real-time obstacle avoidance algorithm is presented for autonomous mobile robots equipped with a CCD as its only sensing modality. The approach uses segmentation technique to segregate ground from other fixtures. It uses a simple computation for the threshold value. The processing speed of the algorithm is fast enough that it can avoid some active obstacles. The algorithm was tested in various lighting conditions in an indoor environment and is remarkably robust. Results of the different thresholding techniques are also presented in conjunction with our computation of the threshold value.

Determining Machine Efficiency Parameters for a Citrus Canopy Shaker Using Yield Monitor Data

A yield monitoring system was used to collect yield data from a commercial citrus canopy shaker during the 2008 harvesting season. A computer algorithm was developed to preprocess the yield data before estimating field efficiency (Ef) and field machine index (FMI) measures of the mechanical harvesting equipment. Total time of the harvesting operation was calculated and divided into primary and support functions, which corresponded to the effective harvesting time and machine time losses, respectively. Time losses related to row-end turning were determined using an algorithm based on linear regression and geometrical methods. Each component of the harvesting operation was then expressed as a percentage of total field time. It was observed that FMI varied from 80% to 98% with 4% standard deviation. Turning time varied from 3% to 8% of the total operational time. Further data analysis showed an exponential relationship between FMI and row-end turning time with R2=0.97. It was also observed that the actual travel distance and the effective time of operation have linear relationships with the theoretical distance of operation with R2=0.96 and 0.93, respectively.

Comparison of two multiband cameras for use on small UAVs in agriculture

This study evaluates the applicability of two multiband cameras used with a small unmanned aerial vehicle (UAV) for stress detection in citrus orchards. The aerial images were acquired using both cameras at UAV flying altitudes of 30, 60, and 90 m were processed to extract histogram distributions of green normalized difference vegetative index as feature datasets. Support vector machine based classification results revealed that the high resolution camera with near infrared (670–750 nm) and green bands was better in detecting healthy and unhealthy citrus trees. The highest average overall classification accuracy of 91±7% (mean ± standard deviation) was obtained using feature datasets of high resolution camera images acquired at an altitude of 60 m.

An Automated Tine Control System for Tractor Drawn Citrus Canopy Shakers

The overall goal of this study was to develop and evaluate an automated control system to adjust the movement of the shaking tines of a tractor drawn citrus mechanical harvesting machine. This system could potentially minimize tree injuries and help with ease of use and increase productivity of citrus mechanical harvesting operation. In citrus groves with significant variability in tree size, trees can be seriously damaged during harvesting by manually controlled mechanical harvesters because it is difficult for the operators to look back all the time and adjust the position of the tines. Two ultrasonic sensors were used to measure the distance of the tines from the canopy edge. A cable positioning sensor was used to report the location of tines. A program was written for this automated control system using LabVIEW and two systems were tested in a grove located in Florida. An RTK GPS was used to measure the location of the tines with respect to the tree canopy. The georeferenced data obtained from the GPS were overlaid on the aerial image of the test plots using GIS software. Using a Factorial design with mixed-effects, two systems were compared based on the mean values of the error term. There were significant differences between the treatments (the average distances between canopy boundary and manual run, canopy boundary and auto-run1, and canopy boundary and auto-run2). Based on the results obtained in this study, both automated control systems seem to be promising but need some improvements.

Yield Monitoring System for Citrus Mechanical Harvester

Yield monitoring system is one of the important components in precision agriculture. Unfortunately, there is no commercial yield monitoring system available for citrus mechanical harvesters to date. Most of the large equipment manufacturer investments are focus on developing the efficiency of their machine but has little investment or none at all in research and development of yield monitoring system for these high value crops. A yield monitoring system using impact plate has been developed that uses the impact of the fruits to correlate to its mass. The system performs well in both laboratory and field tests but due to the impact created by the oranges falling on the plate, the durability of the system is less. A non-contact method of collecting yield data will be very desirable to either augment the performance of the current system and eventually replaced it.

Real-Time, Variable-Depth Tillage for Managing Soil Compaction in Cotton Production

Cotton root growth is often hindered in the Southeastern U.S. due to the presence of root-restricting soil layers. Tillage must be used to temporarily remove this compacted soil layer to allow root growth to depths needed to sustain plants during periods of drought. However, the use of a uniform depth of tillage may be an inefficient use of energy due to the varying depth of this root-restricting layer. Therefore, the objective of this project was to develop and test equipment for controlling tillage depth “on-the-go” to match the soil physical parameters, and to determine the effects of site-specific tillage on soil physical properties, energy requirements, and plant responses in cotton production. Site-specific tillage operations reduced fuel consumption by 45% compared to conventional constant-depth tillage. Only 20% of the test field required tillage at recommended depth of 38-cm deep for Coastal Plain soils. Cotton taproot length in the variable-depth tillage plots was 96% longer than those in the no-till plots (39 vs. 19.8 cm). Statistically, there was no difference in cotton lint yield between conventional and the variable-depth tillage. Deep tillage (conventional or variable-rate) increased cotton lint yields by 20% compared to no-till.

Experimental approach to detect water stress in ornamental plants using sUAS-imagery

Efficiency in irrigation management is crucial to optimize water use in agriculture. A good irrigation strategy requires accurate and reliable measurements of crop water status that provide dynamic data and timely spatial information. However, this is not feasible with time-consuming manual measurements, which are also prone to cumulative errors due to subjective estimations. Ornamental horticulture crops offer challenges for applying small unmanned aircraft systems (sUAS) technology due to the relatively small area of production and its diversity of plant species. sUAS can operate on demand at low flight height and to carry a wide range of sensors allows capturing the variation of plant traits over time, making it a timely alternative to ground-based data collection in nursery systems. This research evaluated the potential of sUAS-based images to estimate crop water status under three different irrigation regimes. sUAS-imagery of experimental plots was acquired in August 2017 using several multispectral sensors. Container-grown ornamental plants used in the study were Cornus, Hydrangea, Spiraea, Buddleia and Physocarpus. An algorithm based on the object-based image analysis (OBIA) paradigm was applied to retrieve spectral information from each individual plant. Preliminary one-way analysis of variance (ANOVA) identified water stressed and non-stressed plants from data of each study sensor, although spectral separation was higher when information from the sensors was combined. Our results revealed the potential of the sUAS to monitor water status in container-grown ornamental plants, although further analysis is needed to explore vegetation indices and data analysis algorithms.

Predicting cotton yield of small field plots in a cotton breeding program using UAV imagery data

One of the major criteria used for advancing experimental lines in a breeding program is yield performance. Obtaining yield performance data requires machine picking each plot with a cotton picker, modified to weigh individual plots. Harvesting thousands of small field plots requires a great deal of time and resources. The efficiency of cotton breeding could be increased significantly while the cost could be decreased with the availability of accurate methods to predict yield performance. This work is investigating the feasibility of using an image processing technique using a commercial off-the-shelf (COTS) camera mounted on a small Unmanned Aerial Vehicle (sUAV) to collect normal RGB images in predicting cotton yield on small plot. An orthonormal image was generated from multiple images and used to process multiple, segmented plots. A Gaussian blur was used to eliminate the high frequency component of the images, which corresponds to the cotton pixels, and used image subtraction technique to generate high frequency pixel images. The cotton pixels were then separated using k-means cluster with 5 classes. Based on the current work, the calculated percentage cotton area was computed using the generated high frequency image (cotton pixels) divided by the total area of the plot. Preliminary results showed (five flights, 3 altitudes) that cotton cover on multiple pre-selected 227 sq. m. plots produce an average of 8% which translate to approximately 22.3 kgs. of cotton. The yield prediction equation generated from the test site was then use on a separate validation site and produced a prediction error of less than 10%. In summary, the results indicate that a COTS camera with an appropriate image processing technique can produce results that are comparable to expensive sensors.