Using fast fourier transform for weed detection in corn fields

Agriculture plays one of the most important roles in economy and therefore lowering costs and improving quality of agricultural products is highly demanded. Controlling weeds is one of the most important and also expensive labors in agriculture which can be automated using robotic cultivators. These robots should be armed with a digital camera which uses a method to classify between weeds and crops based on captured image and then remove the weed by spraying herbicide accurately on the weed, cutting with blades or damaging with electric shock devices. Any of these methods reduce herbicide usage which also protects environment from side-effects of these chemical substances. In this article a method is proposed which utilizes fast Fourier transform and leaf edge density to classify between crop and weed leaves in corn fields in real-time. This method is based on specific shapes of these leaves and leaf vein structures. Testing the method on a sample set of corn field images showed more than 92% accuracy in detecting weed plants. The resulting application is finally compiled to a dynamic linked library (dll) and used in a graphical user interface (GUI) to be used further by a cultivator robot in a real field.