Development and validation of a neural network for the automated detection of horn flies on cattle

Authors: PSOTA, ERIC - University Of Nebraska; LUC, EMILY - University Of Tennessee; PIGHETTI, GINA - University Of Tennessee; SCHNEIDER, LIESEL - University Of Tennessee; TROUT FRYXELL, REBECCA - University Of Tennessee; Keele, John; Kuehn, Larry

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/24/2020
Publication Date: 1/1/2021
Citation: Psota, E.T., Luc, E.K., Pighetti, G.M., Schneider, L.G., Trout Fryxell, R.T., Keele, J.W., Kuehn, L.A. 2021. Development and validation of a neural network for the automated detection of horn flies on cattle. Computers and Electronics in Agriculture. 180. Article 105927. https://doi.org/10.1016/j.compag.2020.105927.
DOI: https://doi.org/10.1016/j.compag.2020.105927

Interpretive Summary: Horn flies feed on the blood of cattle causing annual losses of $1.5 billion annually in the United States. Horn flies spread bacteria causing mastitis in cattle and, potentially, bacterial contamination of beef and dairy products. Insecticides kill horn flies but are not satisfactory because flies develop resistance to insecticide and the insecticides kill beneficial non-target insects such as honeybees and dung beetles. There is genetic variation in resistance of cows (the host) to horn flies. Two approaches to horn fly control that minimize insecticidal use are integrated pest management and genetic selection of cattle for horn fly resistance. Both approaches require low cost counting of horn flies on individual animals. Because of the large numbers of flies on some animals (> 1,000), manual counting of flies on animals or on digital images is not cost effective. We developed a computer vision system that accurately detects and counts flies on individual cows. Coupled with digital images of registered cattle our system can exploit the known pedigree of those animals to produce a national cattle evaluation for horn fly resistance while at the same time facilitating the identification of antagonisms and synergies between horn fly resistance and production traits. Furthermore, our computer vision system supports precision application of insecticide while minimizing fly resistance to insecticide and off target effects on beneficial insects. Low cost horn fly counting is a prerequisite to the development of any new alternatives to insecticides for controlling horn flies. Development of control strategies requires quantification of horn flies to know if the strategy is effective.

Technical Abstract: When the number of horn flies that blood feed on cattle exceeds the economic threshold, they can adversely affect the health and wellbeing of their hosts. Excessive horn fly burdens also lead to reduced weight gain and, consequently, diminished profits for livestock producers. Effective management and treatment require reliable surveillance methods for estimating the degree of horn fly burden (i.e., counting the number of flies on cattle). Traditionally, these estimates are obtained through human visual estimation, either in-person or by counting images on a photo; however, these methods are costly both in terms of time and labor and are prone to subjectivity and bias. In contrast, automated methods can expedite the counting process and remove subjectivity and bias. To this end, a 2-stage method is presented here that uses computer vision and deep learning to identify the location of flies in digital images. The first stage segments the salient cow from all other parts of the image to remove flies on neighboring cattle from consideration. The second stage processes full-resolution patches of the original image and produces a heat map at the location of flies in the images. The method was trained on a set of 375 human-annotated images and tested on 120 images, where significant variation was observed amongst the human scorers. Counting results are compared to four separate human scorers and demonstrate that the neural network produces consistent results and that the method is reliable. Thus, the developed method can be used for monitoring changes in horn fly populations over time by anyone and allows for increased rigor and repeatability. An examination of individual images where the method was closest to and farthest from the human counts provides valuable insights regarding photographic processes that lead to success and failure.