Hamed Ramezani - Former M.Sc. student, Department of Biosystems Engineering, Faculty of Agriculture, Shahid Chamran University of Ahvaz, Ahvaz, Iran.
Hassan Masoudi - Assistant professor, Department of Biosystems Engineering, Faculty of Agriculture, Shahid Chamran University of Ahvaz, Ahvaz, Iran.
Hassan ZakiDizaji - Assistant professor, Department of Biosystems Engineering, Faculty of Agriculture, Shahid Chamran University of Ahvaz, Ahvaz, Iran.
Gholamreza Akbarizadeh - Associate professor, Department of Electrical Engineering, Faculty of Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran

Computer vision unit is used as the main part in automatic guidance systems of vehicles for guiding and safety purposes. In this study, a specific image processing algorithm was designed to develop a machine vision system to be used in agricultural tractors as a human detection system. At first, the input image from a camera installed in front of tractor was divided into small blocks and then unneeded sections of the image were deleted. Then, human was separated from the background, using magnitude gradient and watershed techniques. Finally, using subtraction of human from the background in each frame and comparing consecutive frames, the human was searched and identified. To evaluate the designed algorithm performance in different conditions, a factorial experiment was executed with a completely randomized design. The experimental factors were tractor forward speed at three levels (۱۰, ۲۰, and ۳۰ km/h) and lighting condition at three levels (morning, noon, and evening). The results of the variance analysis showed that the effect of forward speed and lighting conditions on human detection accuracy was significant at ۱% probability level. However, the dual interaction effects of lighting conditions and forward speed were not significant. The proposed algorithm had high accuracy at ۱۰ km/h and its performance in the morning was better than other times. Therefore, development of visual system is possible using the proposed image processing algorithm.

Agricultural tractors, Machine vision, Magnitude gradient, MATLAB software, Watershed technique