Improvement and optimization of special robot vision guidance algorithm based on oblique convolution bilateral filtering

DOI:	10.11809/bqzbgcxb2024.04.036
Keywords:	skewed convolution; bilateral filtering; mobile robot; unstructured environment; wavelet decomposition
Abstract:	Mobile robots can effectively reduce human labor behavior while ensuring personal safety. Robots have been widely used in transportation, rescue, industrial automation and other fields. In this paper, a visual guided image processing and obstacle height evaluation method was proposed for special obstacle jumping robots to operate in unstructured scenes mainly in planes such as towns and factories. In the process of image preprocessing, in order to better preserve the image edges with specific orientation, a skew convolution spatial convolution kernel with Gaussian distribution was introduced on the basis of the two sided filtering algorithm, which was improved to skew convolution two sided filtering, and compared with other edge preserving filtering algorithms. The experimental results show that when the SNR of different algorithms is consistent, the average PFOM of oblique convolution bilateral filtering algorithm was higher than that of guided filtering and weighted least squares filtering. When σ=0.02, the PFOM of oblique convolution bilateral filtering is 12.18 % higher than that of guided filtering and 4.4 % higher than that of weighted least squares filtering. The obstacle crossing robot was directly aligned with the step at a height close to 100 mm, 150 mm and 200 mm under different lighting conditions, and the success rate of obstacle crossing was 100 % under the guidance of the visual system.
Issue:	Vol. 45 No. 4 (2024)
Published:	2024-04-30
PDF HTML