基于内容图像检索的机器人障碍物检测方法

Robot obstacle detection method based on CBIR

为研究搬运机器人的视觉识别系统,提出一种基于内容图像检索的方法识别障碍物。为检测固体障碍物,从不同位置拍摄多种障碍物,保证这些图像的数量和质量,利用拍摄的图像构建一个稳健的图像数据库;利用3种不同的特征提取方法,将图像纹理作为障碍物的特征信息,实时更新障碍物的信息;进行相似度距离计算,比较检索图像与数据库中的图像距离,将固体障碍物的所有实时信息和GPS数据传输到服务器。实验结果验证了所提方法的有效性,实验结果表明,EMD距离的精准度最高,性能比其它距离计算更好。

To study the vision recognition system of the moving robot,a content-based image retrieval was proposed to identify obstacles.To detect solid obstacles,a variety of obstacles was taken from different locations to ensure the quantity and quality of these images,and a robust image database was constructed using the captured images.Three different feature extraction met-hods were adopted,and the image texture was taken as the feature information of the obstacle,meanwhile,the obstacle information was updated in real time.The similarity distance was calculated,and the distance between the retrieved image and the image in the database was compared.All real-time information of solid obstacles and GPS data was transmitted to the server.Experimental results verify the effectiveness of the proposed method.The results show that EMD distance has the highest accuracy and better performance than other distance methods.