Evaluating Texture Modelling Techniques to Determine Stand Establishment and Plant Populations in Corn.

Early detection of crop emergence and plant health enables producers to make timely management decisions or implement alternative production strategies if needed. Collecting information on stand establishment, early season plant health and other abiotic factors affecting crop development requires an effective scouting program by producers. This gives farmers information needed to estimate potential yield, and also provides documented evidence that can be given to insurance companies if problems arise. Current field-based assessment methods are labor intensive, costly, and provide limited information. Whereas manual field assessment provides more conclusive information, it too becomes costly and less economical as field size increases and as a result scouting frequency diminishes. Timely information regarding stand counts, plant heights, and health at specific growth stages is vital to make educated management decisions. Automated scouting programs however, could be an effective means to accumulating and managing crop data. Automatically estimating early stand counts and frequently evaluating crop density on a large number of acres provides tremendous value to a producer. To accomplish this, we advocate employing an Unmanned Aerial System (UAS) capable of collecting geo-referenced high spatial resolution imagery. Such a system can collect high resolution imagery at a fraction of time and cost as that of manual assessment. We have collected over 10 acres of imagery using Robota Triton and Precisionhawk Lancaster platforms fitted with visible and near infra-red cameras. The images were collected at altitudes of 150 feet, 200 feet and 400 feet with ground resolution of approximately 0.5 to 2 inches. Images were mosaicked to form an orthomosaic. Initial pilot experiments with template matching algorithms for stand counts and texture based analysis for density estimation show encouraging results. Texture modelling techniques were investigated to map three different densities (Low, Medium and High) on a corn field by using visible imagery collected using UAS.