Human eyes are prone to fatigue when identifying obstacles for a long time,resulting in large error in obstacle identification.Therefore,a machine vision based obstacle identification method for live working robots in distribution net-works is proposed.The camera carried by the live working robot accurately captures continuous frame images during distri-bution network operations.Using multi-scale Harris subpixel corner detection algorithm,image feature points are extracted and matched using an improved k-d tree nearest neighbor search algorithm to obtain matching feature point pairs.Establish a robot motion transformation model,and combine adaptive filtering and threshold segmentation to suppress the background information of the image after motion compensation,extract obstacle targets,and complete obstacle recognition.Experi-ments show that this method can effectively collect images during distribution network operations,and the feature point matching effect is good.The obstacle recognition results are consistent with the actual situation.
machine visiondistribution networklive workingrobotobstacle identificationfeature point extraction
万方数据
维普
