Table A1.
Features of related work discussed in Section 2.
| Reference | Field of Application | Target to Detect | Sensing Device | AI Technique | Metric Used | Year |
|---|---|---|---|---|---|---|
| [7,8] | SRS | Pedestrians & vehicles | Ultrasound; magnetic field; RADAR | Fuzzy logic | ACC; TPR; FPR; Precision; AUC | 2018; 2018 |
| [10] | ITS | Traffic jams | Simulated | Bio-inspired algorithm; autonomic computing | Queue of vehicles in a traffic light | 2015 |
| [11] | ITS | Traffic jams | 5G; RFID | Microsoft Azure cognitive services | N/A | 2019 |
| [12] | ITS | Traffic jams | Smart phone | LR; bagging; AdaBoost; voting; trivial | ACC | 2017 |
| [13] | ITS | Traffic jams | Camera; security software | MLP; particle swarm | Free flow traffic | 2019 |
| [14] | ITS | Traffic jams | Telemetry | One-class SVM; logit | ACC: Recall; Precision | 2018 |
| [16] | ITS | Vehicles | Camera | AdaBoost | TPR; FDR | 2014 |
| [17] | ITS | Vehicles | Camera | AdaBoost | Detection time | 2019 |
| [18] | ITS | Vehicles | Camera | AdaBoost | TPR; FPR | 2019 |
| [19] | ITS | Vehicles | Camera | SSD MobileNet V1 model | ACC | 2019 |
| [20] | ITS | Vehicles | Camera | RF | N/A | 2019 |
| [21] | ITS | Vehicles | Camera | Naive Bayes; KNN; ANN | ACC | 2019 |
| [22] | ITS | Vehicles | 3D-LIDAR | ConvNet | ACC | 2018 |
| [23] | ITS | Vehicles | Vibration | Naive Bayes; RBFN; SVM; MLP | TPR; FPR; Precision; Recall | 2016 |
| [24] | ITS | Vehicles | Audio | CNN | ACC | 2019 |
| [25] | SRS | Vehicles | Smart phone sensors and camera | ANN; RF; KNN | RScore; TScore | 2018 |
| [26] | SRS | Vehicles | Camera | KNN | N/A | 2018 |
| [27] | SRS | Accident risk | Telemetry | LR; DT; Discriminant analysis; Naive Bayes; SVM; KNN; | ACC | 2019 |
| [28] | SRS | Accident risk | Accelerometers | SOM | FPR; Miss detection; Detection delay | 2014 |
| [29] | SRS | Pedestrians | Camera | Region-based CNN; SVM; MLP | AUC | 2019 |
| [30] | SRS | Pedestrians | Camera | KNN; SVM; ANN; DT | Performance | 2018 |
| [31] | SRS | Pedestrian | Camera | HOG based on SVM | Error rate | 2017 |
| [32] | SRS | Pedestrian | LIDAR | KNN; Naive Bayes; SVM | Error rate; AUC; Sensitivity; Specificity; Precision; ACC; F1-score | 2017 |
| [33] | SRS | Pedestrian | LIDAR | DNN; LSTM; CNN | Classification rate vs. Time to cross; Classification rate vs. Distance to cross | 2016 |
| [34] | SRS | Pedestrian | Camera | Haarcascade based on OpenCV library; HOG based on SVM; SSD based on MobileNet; YOLO based on DNN | ACC | 2019 |
| [35] | SRS | Animals | Camera; presence sensor | KNN; RF | F1-score | 2019 |
| Proposed | SRS | Vehicles | Ultrasound; magnetic field; RADAR | LR; RF; MLP; One-class SVM; LSTM; DRL | TPR; FPR; Precision; ACC; F1-score; AUC | 2020 |
ACC: accuracy; ANN: artificial neural network; AUC: area under the curve; CNN: convolutional neural network; ConvNet: deep convolutional neural network; DNN: dense neural network; DRL: deep reinforcement learning; DT: decision tree; FDR: false discovery rate; FPR: false positive rate; HOG: histogram of oriented gradients; ITS: intelligent transport system; KNN: k-nearest neighbors; logit: logistic regression linear model; LR: logistic regression; LSTM: long short-term memory; MLP: multi-layer perceptron; RBFN: radial basis function network; RF: random forest; SSD: single shot detector; SOM: self-organizing map; SRS: smart road safety; TPR: true positive rate; YOLO: you-only-look-once.