The spatial distribution of weeds is aggregated most of the time in crop fields. Site-specific management of weeds could result in economical and environmental benefits due to herbicide saving. Real time spot-spraying of weeds has an advantage of not requiring time and resources for mapping prior to the spraying operation. However, to implement this technique, sensors are needed for automatic detection of weeds. Up to now, the biggest challenge for the development of weed sensors is the discrimination between crop plants and weeds. One way to avoid discrimination between crop and weed is to sense the inter-row, where only weeds are present. To compare the weed infestation of the row and the inter-row, ground pictures of corn field were acquired at the growth stage 3 to 5 leaves of corn. Images were then treated to keep only vegetation pixels by color segmentation. From these images, three regions were compared: the inter-row compacted by tractor/seeder wheels during the seeding process, the corn row (where there was no corn) and the undisturbed inter-row. The results of this study show that there is a phenomenon on the compacted inter-row and on the corn row that promotes the emergence of weeds by 38 % on average compared to the undisturbed inter-row. However, no significant difference was detected on some plots, Moreover, when linearly (contiguous samples) comparisons were made of the three regions, it appears that sensing the compacted inter-row would result in over-estimating the row infestation 15 % of the time, under-estimate 12 % of the time, and for the undisturbed inter-row it would over-estimate by 12 % and under-estimate by 18 %. In conclusion, according to our data, it would be more conservative to use a vegetation sensor in the compacted inter-row with an expectation of 12% weed escape.