| The country’s modernization is inseparable from the normal operation of power equipment.As the country vigorously promotes infrastructure construction,a large number of construction vehicles are engaged in construction operations in scenarios that include transmission lines,under this circumstance,accidents involving the destruction of transmission lines caused by the brutal construction of construction vehicles frequently occur.Traditional transmission line protection methods such as manual inspection and real-time monitoring cannot detect hidden dangers in time and are prone to missed inspections.In order to solve the shortcomings of traditional methods,this paper combines deep learning technology to propose an engineering vehicle detection and tracking algorithm for transmission line scenarios.Aiming at the problem of low accuracy of the YOLOv3 detection algorithm on the constructed engineering vehicle data set,a three-point improvement method is proposed.First,improve the feature fusion layer for detecting small targets and introduce the SPP module to enhance the detection performance of the YOLOv3 algorithm;use the k-means++ clustering algorithm to retrieve the size of the prior frame;use the GIoU idea to improve the original loss function and speed up the frame regression.Finally,experiments show that the mean Average Precision of the improved algorithm is increased by 6.5% compared to the original algorithm.In view of the problem that the SORT tracking algorithm is prone to mark jumps when the tracking target is occluded,the matching stage of the algorithm is chosen to be improved.Add HSV color histogram similarity calculation to strengthen the algorithm’s ability to match tracking targets.Experiments show that the accuracy of multi-target tracking of the improved algorithm is increased by 5.4% compared with the original algorithm.According to the above algorithm and PyQt5,a set of engineering vehicle detection and tracking experiment system with interface interaction function is designed.The experiment proves that the algorithm in this paper can achieve the purpose of detecting and tracking engineering vehicles. |