| In recent years,with the development of intelligent algorithm,building information model(BIM),deep learning and other technologies,the traditional construction mode has been unable to meet the requirements of production development,and the integrated construction of Intelligent Machinery arises at the historic moment.Integrated construction is a huge and complex system,but there are few intelligent theoretical researches related to integrated construction.Therefore,based on ant colony algorithm,object detection algorithm based on deep learning,knowledge base and other intelligent theory,this paper studies the integrated construction.Intelligent algorithm is widely used in various fields because of its high accuracy and strong adaptability.In this paper,an improved ant colony algorithm(MACO)is proposed.Four functions are selected to verify the optimization performance of MACO.Meanwhile,the optimization performance of MACO is compared with that of standard GA,ACO,PSO,AFSA and QGA.By finding the maxima and minima of the four benchmark functions,it is proved that the accuracy and robustness of MACO are obviously better than other algorithms,so it can be used to improve the K-means clustering algorithm,and the improved MACO clustering algorithm is applied to the anchor frame clustering of CM-YOLO.Target detection algorithm is the key algorithm for intelligent machinery to realize cooperative operation and obstacle avoidance.By adjusting the network structure,replacing the activation function,and applying MACO clustering algorithm to adaptive anchor frame extraction,cm-yolov5 algorithm with higher accuracy and faster speed is obtained.In order to verify the effectiveness of the algorithm,1500 images of excavators,bulldozers,rollers,trucks and workers are collected and stored in adas.We use image annotation tools to label five kinds of objects in adas,and transform the label file into the format that CM-YOLOv5 can recognize.CM-YOLOv5 is trained and learned based on the data set and its label file,and a target detection model suitable for construction machinery recognition is obtained.The training time is reduced by 30%compared with that before the improvement.The network structure is lighter and the accuracy is higher.The real-time detection can be realized,and the accurate identification and positioning can be carried out under the bad conditions of many targets mutual occlusion,insufficient light,small targets and low resolution,which can be further applied to engineering practice.Before drawing up the integrated construction plan,the earthwork volume is calculated based on the construction map,and the cases in the knowledge base are matched.Through the scanning aerial photography of a construction site by Dajiang surveying and mapping UAV,and based on the modeling software,367 images are converted into high-precision three-dimensional digital model,the total average error is 6.22 cm,and the centimeter level positioning is realized.Compared with the traditional method,the 3D model is more efficient,more intuitive and effective,and realizes the engineering digitization.The information of equipment,operation object and operation environment in each process link is analyzed,and the construction process is established.According to the rule language,the case information of construction process knowledge base under multiple working conditions is transformed.The integrated construction technology mainly includes the construction object and the construction technology of dump truck,bulldozer,excavator and roller.By establishing the ontology concept tree model of construction technology,the information processing is standardized and simplified. |
