Histogram based thresholding methods in crop rows detection

Many processes in agriculture require high degree of precision and they must be repeated for several hours. This can be very exhausting for human labor and because of that there is a need for automation of such processes. Crop rows detection methods play important role in automation of many processes in agriculture like planting, fertilization, plant protection and harvesting. All of these processes can easily be automated if position of crop rows is correctly detected. Most of state of the art methods for crop rows detection are based on image processing. One of the most important image preprocessing stage is image thresholding. Thresholding error directly affects crop rows detection accuracy and because of that it is very important to use proper thresholding method. In this paper we presented several histogram based thresholding methods including Otsu thresholding method, Double Otsu thresholding method, Minimum error thresholding method and Minimum cross-entropy thresholding method. All methods are tested on crop rows images captured in the Croatian region of Slavonia. For all captured images an expert created ground truth images which are used for methods error calculation and analysis.