Image polar radius distribution for seed orientation adjustment

Seed orientation is an important factor in high-speed automatic seeders used for sowing crops such as garlic cloves and corn kernels. However, obtaining accurate seed orientation angles can be challenging due to high-frequency noise points and false top features in seed images, which can negatively impact the orientation effect. In this paper, we introduce a directional adjustment method based on the polar radius distribution map to address this issue. Our method involves two main steps. Firstly, we use the morphological opening operation to process speckles and some seed apex parts and use the region marker to obtain the original polar radius distribution curve signal. Secondly, we use the wavelet packet transform to analyze the original polar radius distribution curve signal, which yields a direction rotation angle after analysis and calculation and enables the determination of seed orientation. Our experimental results show that our method can better identify the germination location at the top of the seed, and the calculated rotation angle is more conducive to seed direction adjustment than the rotation angle directly determined by the maximum polar radius in the polar radius vector. This is due to the advantage of higher frequency resolution of wavelet packet analysis, which enables us to obtain a polar radius distribution curve signal with low noise, better smoothness, and more realistic reflection of the edge characteristics of the seed contour boundary. The proposed method provides an effective solution for accurately adjusting the orientation of garlic cloves and corn kernels and potentially other crops that require automation technology. Our key finding is that using the polar radius distribution map and wavelet packet transform can yield better results in seed orientation adjustment, by removing high-frequency noise points and false top features and improving the accuracy of seed rotation angle calculation.