Objective  To solve the existing problems of noise interference and ineffective detection in the target area during navel orange recognition in agricultural environment.

Method  A navel orange recognition method based on wavelet transform and Otsu threshold denoising was proposed. Firstly, we chose a better contrast and established the YCbCr model with color space helpful for image segmentation. Then we designed a navel orange detection algorithm based on Otsu threshold denoising and reduced noise interference in the segmentation region of navel orange. Finally we proposed a circle filling method based on the center of mass to determine the position of navel orange in the image, and the detection result was displayed in the original image.

Result  The recognition rates of cyan and orange navel oranges were 87.10% and 94.18%, respectively. The recognition rates of navel orange were 92.96% and 90.15% respectively under direct light and backlight, and 90.82% and 93.18% respectively under occluded and unoccluded situation. The total recognition rate was 92.07%.

Conclusion  The method has strong environmental adaptability, and is suitable for the identification and processing of navel orange images under different occlusion, light and skin color conditions in agricultural environment.