WEED COVER ON AND BETWEEN CORN ROWS IMPLICATIONS FOR REAL-TIME WEED DETECTION

2010 
The spatial distribution of weeds in crop fields is heterogeneous. Therefore, limiting herbicides application to weed infested areas would lead to economical and environmental benefits. For real-time spot treatments, sensors detecting weed patches are needed. Vegetation sensors could be used in the inter-rows to trigger herbicide spraying on both rows and inter-rows if weed cover on and between the crop rows is uniform. To verify this hypothesis, weed cover on and between corn rows was evaluated using photographs acquired in corn fields at the 3 to 5 leaf stage. A one hectare plot was sampled in 2004, 2005 and 2007 at one location and nine one hectare plots were sampled in corn fields dispersed across the province of Quebec (Canada) in 2008. All fields were planted in corn under conventional tillage (75 cm row spacing). A segmentation algorithm was used to isolate vegetation pixels. Samples for the analysis consisted of 23 x 750 mm strips free of corn plants and covering three regions: undisturbed inter-row (UIR), corn row and inter-row compacted by tractor and/or seeders wheel during the seeding process (WIR). Repeated Anova measures indicated that weed cover on the undisturbed inter-row was generally lower than on the CR or WIR (p<0.001). No significant difference in weed cover was observed between CR and WIR. A presence/absence contingency table showed that 13-15% of samples had no weed pixels on the WIR while pixels were present on the row, indicating that 13-15% of weeds located on the CR would be missed if detection was based on the WIR.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    15
    References
    0
    Citations
    NaN
    KQI
    []