Abstract

Youchun Ding , Du Chen and Shumao Wang

For combine harvester, automatically and promptly recognizing the cut and uncut edge of guideline is the key technology to realize autonomous navigation based on vision machine. One novel method termed wavelet image rotation and projection algorithm has been established. It includes the following essential steps. Firstly, 23 scale of wavelet transforming was selected because of intrinsic characteristics of multiscale and multiresolution of wavelet transforming, and the low frequency wavelet image was obtained which was used to rotate transform to computer column average values. To achieve the best match of rough-division and elaborate-division, and to reduce processing time, the initial given rotation angle was 5°, and the final rotation angle was 1° based on former rotating transform. Then the rotation and projection matrix R including the column average values and the enumeration angles was achieved. Finally, the difference in rotation and projection matrix came from R ’s row vectors’ difference operation to locate extremum values, the heading angle and lateral position can be found out. The method result shows the accuracy of recognizing cut and uncut line, achieve the effect with 95.6%, the average processing time is 109.3 ms per frame, and the algorithm owns the characteristics of anti-color-disturbance, independent on the whole brightness and high robustness. The method can also be applied to detect the field end line.