Web tool for classification and quantification of flowers for pollinator interactions using R

Authors: SUBHASHREE, S - North Dakota State University; SUNOJ, S - North Dakota State University; IGATHINATHANE, C - North Dakota State University; Franco, Jose; MALLINGER, RACHEL - University Of Florida; Archer, David

Submitted to: American Society of Agricultural and Biological Engineers
Publication Type: Abstract Only
Publication Acceptance Date: 5/22/2018
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Agricultural intensification is the major cause of biodiversity loss since it affects the major terrestrial area in modern agriculture. There is concrete evidence of a decrease in insect pollinators because of agricultural intensification. A basic solution is to incorporate flowering plants such as cover crops in the crop rotation that is widely adopted as it possesses multiple social-economic benefits. Floral traits, such as color, shape, size, and shape, influence the pollinators visitation rate and affect plant pollination. Bee pollinator discrimination based on their floral traits among three different flower species (Phacelia, Buckwheat, and Mustard) was studied in a bee visitation study plot containing four main plots, each containing 16 subplots. The flower quantification was performed by the conventional manual method (using manual count and ruler) and a developed image processing technique. A custom designed frame with legs was used in the test plots for capturing the inflorescence images. The images were captured weekly for a duration of about three months (June – August 2017), a total of 192 sample images (4 main plots × 16 subplots × 3 replications) during one visit. An image processing algorithm in R (open source statistical and data analysis software) will be developed for classifying the flower species and quantifying the flowers coverage. The program will be developed with capabilities to (i) segment the plant area within the frame; (ii) employ projective transformation technique to correct any distortion in the captured segmented image; (iii) classify the flower species based on the color and shape; and (iv) determine the flowers coverage after classification. A simple web application will be developed using the Shiny package in R. One of the advantages of developing this web application is that it is compatible and accessible in desktops and mobile phones. This application will allow the user to upload the images to classify and quantify the flower species. Statistical analysis will be conducted to compare the difference between the manual measurements and the image analysis results; additionally, the relationship between the flower species and coverage with bee visitation rate will be determined.