To improve the performance of the forest fire smoke detection model and achieve a better balance between detection accuracy and speed, an improved YOLOv4 detection model (MoAm-YOLOv4) that combines a lightweight netwo...To improve the performance of the forest fire smoke detection model and achieve a better balance between detection accuracy and speed, an improved YOLOv4 detection model (MoAm-YOLOv4) that combines a lightweight network and attention mechanism was proposed. Based on the YOLOv4 algorithm, the backbone network CSPDarknet53 was replaced with a lightweight network MobilenetV1 to reduce the model’s size. An attention mechanism was added to the three channels before the output to increase its ability to extract forest fire smoke effectively. The algorithm used the K-means clustering algorithm to cluster the smoke dataset, and obtained candidate frames that were close to the smoke images;the dataset was expanded to 2000 images by the random flip expansion method to avoid overfitting in training. The experimental results show that the improved YOLOv4 algorithm has excellent detection effect. Its mAP can reach 93.45%, precision can get 93.28%, and the model size is only 45.58 MB. Compared with YOLOv4 algorithm, MoAm-YOLOv4 improves the accuracy by 1.3% and reduces the model size by 80% while sacrificing only 0.27% mAP, showing reasonable practicability.展开更多
With new sources of big data,it is increasingly possible to practically implement advanced freight forecasting models including activity-based and truck touring models.Such models improve upon traditional trip-based a...With new sources of big data,it is increasingly possible to practically implement advanced freight forecasting models including activity-based and truck touring models.Such models improve upon traditional trip-based approaches by capturing freight behaviors sensitive to transportation policy and infrastructure changes.A persistent challenge with the use of big data in this context is the ability to generalize a set of representative behaviors to serve as the basis for model calibration and validation from anonymized data depicting the complex behaviors of the population.To address this challenge,we present a two-stage methodology to extract unique and representative freight activity patterns from passively collected truck Global Positioning System(GPS)data.The first stage involved a heuristicbased approach to derive a set of stop and trip characteristics from large-streams of GPS pings.The second stage employed data mining and machine learning techniques to discern common freight activity patterns from the set of defined features.The resulting activity pattern profiles,defined as chains of activities and their trajectories over time and space,allow us to maintain the anonymity of the trucks included in the GPS dataset while providing high-resolution travel profiles-a necessary condition for most data sharing agreements between public agencies and private data providers.These activity patterns serve as the critical,and currently missing,data needed to calibrate and validate advanced freight forecasting models.With more advanced forecasting models reflective of observed freight behaviors,we will be able to evaluate a wider spectrum of policy and infrastructure scenarios more accurately.展开更多
文摘To improve the performance of the forest fire smoke detection model and achieve a better balance between detection accuracy and speed, an improved YOLOv4 detection model (MoAm-YOLOv4) that combines a lightweight network and attention mechanism was proposed. Based on the YOLOv4 algorithm, the backbone network CSPDarknet53 was replaced with a lightweight network MobilenetV1 to reduce the model’s size. An attention mechanism was added to the three channels before the output to increase its ability to extract forest fire smoke effectively. The algorithm used the K-means clustering algorithm to cluster the smoke dataset, and obtained candidate frames that were close to the smoke images;the dataset was expanded to 2000 images by the random flip expansion method to avoid overfitting in training. The experimental results show that the improved YOLOv4 algorithm has excellent detection effect. Its mAP can reach 93.45%, precision can get 93.28%, and the model size is only 45.58 MB. Compared with YOLOv4 algorithm, MoAm-YOLOv4 improves the accuracy by 1.3% and reduces the model size by 80% while sacrificing only 0.27% mAP, showing reasonable practicability.
文摘With new sources of big data,it is increasingly possible to practically implement advanced freight forecasting models including activity-based and truck touring models.Such models improve upon traditional trip-based approaches by capturing freight behaviors sensitive to transportation policy and infrastructure changes.A persistent challenge with the use of big data in this context is the ability to generalize a set of representative behaviors to serve as the basis for model calibration and validation from anonymized data depicting the complex behaviors of the population.To address this challenge,we present a two-stage methodology to extract unique and representative freight activity patterns from passively collected truck Global Positioning System(GPS)data.The first stage involved a heuristicbased approach to derive a set of stop and trip characteristics from large-streams of GPS pings.The second stage employed data mining and machine learning techniques to discern common freight activity patterns from the set of defined features.The resulting activity pattern profiles,defined as chains of activities and their trajectories over time and space,allow us to maintain the anonymity of the trucks included in the GPS dataset while providing high-resolution travel profiles-a necessary condition for most data sharing agreements between public agencies and private data providers.These activity patterns serve as the critical,and currently missing,data needed to calibrate and validate advanced freight forecasting models.With more advanced forecasting models reflective of observed freight behaviors,we will be able to evaluate a wider spectrum of policy and infrastructure scenarios more accurately.
