| Summary: | 碩士 === 國立臺灣大學 === 生物產業機電工程學研究所 === 89 === A machine vision system was developed for the guidance and operation of an automated guided vehicle in agricultural environment. Vanishing points/line provided the information of guidance. And the field operation and parameters were determined by the species and growth stages of crops based on their texture.
When an automated guided vehicle drives on a field in which crops grew along parallel lines, the vanishing points of the parallel lines could be served as the target of navigation. A slanting camera due to the vehicle driving on the rough surface of the field would lead to inaccurate calculation of vanishing points. A star-shaped pattern, which consists of three sets of parallel lines, was used to obtain pattern vanishing line for the calibration of the camera orientation. Based on the information, the navigation direction of the vehicle could be determined within the error of 4 degrees.
For the recognition of crop environment, crop texture changes as the distance of crops from camera due to perspective effect. Therefore, inverse perspective transformation was introduced to diminish the distortion. Back propagation neural network (BPNN) incorporated with gray level co-occurrence matrix was used to distinguish different kinds and growth stages of crops. The accuracy for the recognition of 22 different crops was up to 95% approximately.
|
|---|
