Assessment of Wild Pig Damage in Mississippi Crop Fields
Wild pigs are a significant threat to cropland in the Midsouth, and can cause damage from crop planting through maturity. Estimating economic loss due to wild pigs is a complex problem due to both sampling error and lack of visual investigation. This is especially challenging in the interior of fields with tall-growing crops–e.g. corn. Quantifying the extent of wild pig damage in fields, rather than merely noting that damage is present, is more informative for estimating total losses.
Researchers at MSU conducted experiments in five production corn fields in Bolivar, Leflore, and Sunflower Counties in Mississippi in 2016 with the objective of assessing the ability of a small unmanned aerial system (UAS) to estimate the area of wild pig damage in these fields. Specific goals were to 1) reliably automate detection of wild pig damage, and 2) accurately estimate the area of damage.
Damaged areas were differentiated from healthy areas by differences in image texture. When corn is damaged, it changes the uniform texture within the image. This is detectable with computer algorithms, and can be used to generate classified maps which show the areas of damage. Methodology and equipment used in the study, plus the results, are reported in an article titled “Quantifying damage from wild hogs with small unmanned aerial systems” that was published in Vol. 42 of the Wildlife Society Bulletin in May 2018 (see below link). A summary of the results follow.
• The premise for the study is that cost-effective, reliable techniques for discovery and measurement of wild pig-related damage are a necessity for determining if damage is significant enough to warrant remedial action by a producer.
• Healthy crop structure exhibited a uniform texture pattern which differed from that of damaged areas. This difference in image texture is the basis for computer-automated detection of wild pig damage.
• A substantial portion of damage to corn fields occurs immediately after planting when wild pigs root up recently planted corn seed. Their consumption of these seeds prior to emergence resulted in a flush of weeds in these areas that were barren of corn plants. Later-season damage consisted of trampled corn stalks. Both instances resulted in image textures that differed from that of healthy corn.
• Overall classification accuracies for pig damage were between 65% and 78%. When the classifier was incorrect, it was more likely to label damaged areas as healthy rather than labeling healthy areas as damaged.
• In summary, the methods used in the study did automate detection, but underestimated the area of wild pig damage in the studied corn fields.
• The approach used in this study provided an objective technique to quantify wild pig damage to a crop field and removed the potential bias from self-reporting by landowners and human observers. The approach additionally allowed for whole-field sampling rather than subsampling of small areas within a field.
The above results, even with identified shortcomings of the technology, indicate that UAS’s represent a new option for assessment of wild pig damage to crops. As with all new technology, it is likely that the identified shortcomings will be reduced or eliminated with continued research. Much of the error in classification can be traced to underlying issues with image mosaicking, which can easily be resolved in future efforts by utilizing more ground reference points within a field that better align single image frames in the mosaic.
One area not addressed in the above research is the use of a UAS to detect wild pig activity at night so that immediate action can be taken to eradicate these predators at the location of their activity. Currently under FAA regulation, night flight is not permissible unless the UAS operator has been granted a waiver. To obtain that waiver, the operator must possess a valid remote pilot certificate and the UAS aircraft must be equipped with additional lights to increase its visibility to the operator as well as manned aircraft.
Reference source: Samiappan et al., Quantifying Damage from Wild Pigs with Small Unmanned Aerial Systems, Wildlife Soc. Bull., Vol., 42(2), p. 304-309, 2018.