Image-based onsite object recognition for automatic crane lifting tasks

The construction industry is suffering from aging workers and frequent accidents, as well low productivity. Automation robotics regarded a promising approach for enhancing the development of industry, automatic operation cranes, an important aspect construction, attracting increasing attention. However, due to complexity dynamics sites, it difficult cranes automatically recognize locate lifting objects (e.g., precast facades partitions) on site. To solve this problem, image-based automated onsite object recognition developed in study. This fusion Faster-R-CNN (Region-based Convolutional Neural Network), Canny, Hough Transformation, Endpoint clustering analysis Vertex-based Determining Model, uniquely with exact pose extract its features centroid coordinates, size, color). Based extracted features, can be retrieved database IFC (Industry Foundation Classes) format BIM (Building Information Modeling) obtain more crane. It shown field experiment that workable has potential support cranes. contributes basic promotes rapid automation robotics.