Progress 03/09/17 to 09/30/21
Outputs
Target Audience:The target audience for the final year of the project was researchers, youths, and college graduates. Changes/Problems:COVID-19 related challenges were present but we were still able to go out to the field and collect data needed for analysis. Supply chain issues did impact our ability to keep desired equipment ready in a timely manner. What opportunities for training and professional development has the project provided?Two graduate students, 1 technician, and 1 visiting professional were engaged in the reported research. More than 15 undergraduates students were additionally provided training on UAV operation as part of a summer REEU project. How have the results been disseminated to communities of interest?Mostly through peer review publications, oral and poster presentations at professional conferences. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
Progress achieved under major goals is given below: Goal (1) a) A Corn Disease and Severity (CD&S) dataset has been created that consists of 511, 524, and 562, field acquired raw images, corresponding to three common foliar corn diseases, namely Northern Leaf Blight (NLB), Gray Leaf Spot (GLS), and Northern Leaf Spot (NLS), respectively. For training disease identification models, half of the imagery data for each disease was annotated using bounding boxes and also used to generate 2343 additional images through augmentation using three different backgrounds. For severity estimation, an additional 515 raw images for NLS were acquired and categorized into severity classes ranging from 1 (resistant) to 5 (susceptible). Overall, the CD&S dataset consisted of 4455 total images comprising of 2112 field images and 2343 augmented images. The dataset was made available via GitHub for use for global use by those interested in training and testing various deep learning algorithms for disease identification and severity estimation. The link to data is given below: Dataset_Original JPG files (.jpg) https://osf.io/s6ru5/ Dataset_Annotated JPG & Text files (.jpg, .txt) https://osf.io/s6ru5/ Dataset_Black_Background JPG files (.jpg) https://osf.io/s6ru5/ Dataset_White_Background JPG files (.jpg) https://osf.io/s6ru5/ Dataset_No_Background PNG files (.png) https://osf.io/s6ru5/ Dataset_Severity JPG files (.jpg) https://osf.io/s6ru5 b. Research was conducted and reported on accurately locating disease outbreak sites and track the severity of estimation in corn fields. Such an effort is a first step in developing an effective disease management system. The research community has been increasingly relying on images acquired by Unmanned Aerial Vehicles (UAVs) and/or handheld sensors instead of using a manual scouting approach to develop an improved disease management system. However, most of the reported studies have used publicly available datasets consisting of images acquired under the uniform indoor settings. We conducted multiple UAV flights over different corn fields resulting in a collection of approximately 60,000 images. Three neural networks namely VGG16, ResNet50, and InceptionV3 were used to train the image classification models for diseased area identification, disease type identification, and severity estimation. The UAV acquired images were split into 250 x 250 pixels to identify the presence of diseased patches within fields with almost perfect accuracies. Different disease types were identified by acquiring a handheld sensor images for Northern Leaf Blight (NLB), Gray Leaf Spot (GLS), and Northern Leaf Spot (NLS) and accuracies up to 98.34% was obtained. Three disease severity levels (low, medium, and high) of NLS was estimated with accuracies up to 94.02%. Finally, an object-detection algorithm, YOLOv4, was used for identifying and locating multiple instances of disease lesions within the NLB, GLS, and NLS handheld images with a mean Average Precision (mAP) score of 57.93%. The study provided promising results for realizing a working system. Goal (2) Different edge devices and DL models are under evaluation for developing smart ground and aerial robots for identifying real-time, foliar diseases iin corn. The research is still under progress.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Ahmad, Aanis1, D. Saraswat, A. Etienne1, Varun Aggarwal1, and B. Hancock. 2021. Performance of Deep Learning Models for Classifying and Detecting Common weeds in Corn and Soybean Production Systems. Computers and Electronics in Agriculture, 184. https://doi.org/10.1016/j.compag.2021.106081
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Ahmad, Aanis1, D. Saraswat, A. Gamal, and G.S. Johal. 2021. Comparison of Deep Learning Models for Corn Disease Identification, Tracking, and Severity Estimation Using Images Acquired From UAV-Mounted and Handheld Sensors. (2021 ITSC Best Paper). https://elibrary.asabe.org/abstract.asp?JID=5&AID=52429&CID=virt2021&T=1
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Ahmad, Aanis1, D. Saraswat, A. Gamal, and G.S. Johal. 2021. CD&S Dataset: Handheld Imagery Dataset Acquired Under Field Conditions for Corn Disease Identification and Severity Estimation. https://export.arxiv.org/abs/2110.12084
|
Progress 10/01/19 to 09/30/20
Outputs
Target Audience:Precision agriculture and big data research and extension personel. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?Research meetings were held with collaborating scientists using virtual media throughout the data collection period and afterwards. What do you plan to do during the next reporting period to accomplish the goals?Analyze the data, continue field data collection during 2021 production season, and use the analyzed data to develop outputs such as conference presentations, web-based tools, research manuscripts, extension publications, new grant proposals etc.
Impacts
What was accomplished under these goals?
New collaborations were initiated with plant pathologists to monitor and identify corn diseases from multiple research plots using UAS based sensors. Data was also collected from multiple plots to identify phentoypic traits for corn. For analyzing these datasets, new studies were initiated to assess the performance of various machine learning models on publicly available datasets. The data collected during 2020 production session is under analysis and is expected to lead to several outputs such as conference presentations, web-based tools, research manuscripts, extension publications, new grant proposals etc.
Publications
- Type:
Conference Papers and Presentations
Status:
Submitted
Year Published:
2020
Citation:
Sura, H., D. Saraswat, and A. Etienne. 2020. Performance of Deep Convolution Neural Network MODELS for Assessing Classification Accuracy of Weed Images Acquired from Publicly Available Datasets. In Virtual ICPA Seminar Series, September 10.
Etienne, A. and D. Saraswat. 2020. Deep Learning Based Object Detection System for Identifying Weeds Using UAS Imagery. Trans. Of ASABE (Under Review)
Ahmad, D. Saraswat, V. Aggarwal, A. Etienne, and B. Hancock. 2020. Performance of Deep Learning Models for Classifying and Detecting Common Weeds in Corn and Soybean Production Systems. Computers and Electronics in Agriculture (Under Review)
|
Progress 10/01/18 to 09/30/19
Outputs
Target Audience:The target audience reached during the reporting period included research professionals, county extension educators, and4-H youths. A number of presentations and hands-on demonstration was conducted for the intended stakeholders during the year under report. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?The project provided an opportunity to network with colleagues from other land-grant universities with similar interests during the annual meeting of the S-1042 organized at the Mississippi State University. How have the results been disseminated to communities of interest?The results of new discovries have been communicated through presentations and conference publications. What do you plan to do during the next reporting period to accomplish the goals?I will continue to work on the objectives identified during the first year and have hired a new graduate student to explore newer algorithms for making full use of UAS imagery for agricultural decision making.
Impacts
What was accomplished under these goals?
(1) A number of flight missions were conducted for acquiring weed images from tresearch plots located in three Purdue Agricultural research Centers. Several deep learning models were developed and compared for assessing classification and identification accuracy of weeds at different growth stages. A manuscript is under preparation. 2) Participated in the development of an online app named CONTxT (https://openatk.com/CONTxT). The app was launched in Spring 2019 and used by Purdue extension agents for data collection during the rest of the year. Several demonstrations for 4-H participants were organized during 2019 summer to apprise them with latest UAS regulations and applications in agriculture.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Aaron Etienne, Dharmendra Saraswat, "Machine learning approaches to automate weed detection by UAV based sensors", in Autonomous Air and Ground Sensing Systems for Agricultural Optimization and Phenotyping IV, J. Alex Thomasson; Mac McKee; Robert J. Moorhead, Editors, Proceedings of SPIE Vol. 11008 (SPIE, Bellingham, WA 2019), 110080R.
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2019
Citation:
Aaron Etienne, Dharmendra Saraswat (2019). Weed Location and Type Identification Using UAV Imagery and Deep Learning, ASABE Paper No. 1901363. St. Joseph, MI: ASABE.
- Type:
Other
Status:
Published
Year Published:
2019
Citation:
Aanis Ahmad, Dharmendra Saraswat, Aaron Etienne, and Ben Hancock (2019). Evaluation of Machine Learning Models for Classifying and Detecting Common Weeds in Corn and Soybean. SURF Presentation, Purdue University.
|
Progress 10/01/17 to 09/30/18
Outputs
Target Audience:The target audience for workshops organized during the year included 4-H students, county extension educators, professionals from academia and industry and college students. Changes/Problems:None at this time. What opportunities for training and professional development has the project provided?The project provided opportunities to strengthen network with other faculty at land-grant universities and research professionals from federal laboratories to come together for finalizing UAV data collection protocol in a record time. How have the results been disseminated to communities of interest?The results of experiences were disseminated through a new app named UAS User Log and a number of workshops that included presentations and hands-on demonstrations for a variety of stakeholders. What do you plan to do during the next reporting period to accomplish the goals?I will continue to mentor my graduate student in analyzing data collected during the 2018 production season, provide workshops to stakeholders for disseminating latest research information, continue brainstorming for updating UAS User Log app with multi-state colleagues, and explore competitive funding opportunities.
Impacts
What was accomplished under these goals?
The progress made under the two identified goals is given below: A multi-rotor, unmanned aerial system (UAS), M600 was flown at varied altitude above ground and with a variety of sensors (visible, near infra red and thermal) from early March until third week of October to acquire images for corn and soybean field experiments at four Purdue Agriculture Research Centers. The flight missions were planned to obtain images of weeds at different growth stages and at different times of the production season under mid-west conditions for answering research questions. The data is being catalogued and machine learning model is under development at the time of reporting. The project provided insights into challenges and opportunities while dealing with commercially available several brands of multispectral and another commerical product that combinesthermal sensor with a RGB sensor. A multistate group formed during the first meeting of S-1069 group met in Texas A & M University campus in March, 2018 and a protocol for documenting UAV mission was deliberated and finalized. The protocol was developed in the form of an app that was released in May 2018 for use by researchers and professionals in streamlining data collection throughout the country.
Publications
- Type:
Other
Status:
Accepted
Year Published:
2018
Citation:
Saraswat, D; D.E. Martin; L.R. Khot and S. Murray. 2018. UAS User Log. Purdue University. https://data.nal.usda.gov/dataset/uas-user-log
|
Progress 03/09/17 to 09/30/17
Outputs
Target Audience:The target audience reached during the reporting period included research professionals, county extension educators, and 4-H youths. A number of presentations and hands-on demonstration was conducted for the intended stakeholders during the year. Changes/Problems:The commercially available sensors are not fully charaterized and mid-west weather conditions require new methods to research mitigating the effects of wind and bi-directional reflectance on images acquired using UAS. Efforts to handle these challenges will remain focus of research inthe future. What opportunities for training and professional development has the project provided?The project providied an opportunities to network with other faculty nd professsional in other land-grant universities with similar interests. As a result of visiting for the inaugural meeting for UAS group, i was able to find new collaborators and submitted a new research proposal for funding consideration to Agricultural and Food Reearch Initiative. How have the results been disseminated to communities of interest?The results of experiences learnt have been communicated through presentations and hands-on demonstrations organized for various stakeholders. What do you plan to do during the next reporting period to accomplish the goals?I will continue to work on the objectives identified during the first year and expect to greatly benefit as one of my new graduate student hire is interested to focushis master's research on exploring the use of UAS for answering various issues related to production agriculture.
Impacts
What was accomplished under these goals?
1) A multi-rotor, unmanned aerial system (UAS) was used throughout the summer to acquire images for various field experiments at one of the Purdue Agricultural Center. The flight missions were planned to provide training to research staff in proper handling of UAS and sensors and experience the importance of factors that determine the quality of images acquired during the mission. The effort got a boost with Purdue administration providing additional funding to the Center for acquiring a fixed wing unit UAS along with a multispectral camera. A number fo missions were also conducted with the newly acquired UAS and camera combo to develop an understanding about system's limitations. 2) A multistate group was formed during the first meeting of UAS group in North Carolina and development of mission documentation prootocol was initiated. This protocol once finalized is likely going to help in streamlining data collection throughout the country.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2017
Citation:
Leiva, J.N., J. Robbins, D. Saraswat, Y. She and R. Ehsani. 2017. Evaluating Remotely Sensed Plant Count Accuracy with Differing UAS Altitudes, Differing Canopy Separations and Ground Covers. Journal of Applied Remote Sensing: CIGR Journal. 11(3):036003-1 to 036003-15.
- Type:
Journal Articles
Status:
Published
Year Published:
2017
Citation:
Leiva, J.N., J. Robbins, D. Saraswat, Y. She and R. Ehsani. 2016. Effect of Plant Canopy Shape and Flowers on Plant Count Accuracy Using Remote Sensing Imagery. Agricultural Engineering International: CIGR Journal. 18(2):73-82.
|
|