Accident avoidance and prediction system using adaptive probabilistic threshold monitoring technique

Abstract Identifying abnormal occasions in an observed area has been of the significant applications in Wireless Sensor Networks (WSNs) and the Internet of Things. Accidents and property harm can be maintained a strategic distance from if precise alarms are informed on time. In conventional checking techniques, a predefined threshold is given, and a signal is activated when the sensor perusing surpasses this threshold. This Single Threshold based Monitoring (STM) experiences the substandard nature of detected data, bringing about numerous false alarms. This work proposes an Adaptive Probabilistic Threshold Monitoring (APTM) technique for WSNs, where a signal is activated if the likelihood of the checked esteem being more significant than a predefined threshold (α) is more impressive than time delay (τ). The tight upper limits of the likelihood that controlled sum is more significant than the predetermined threshold are given. As indicated by the breaking points, probabilistic threshold-based algorithms for conglomeration checking are proposed. Broad execution assessment shows the adequacy of the proposed algorithms—the proposed design centers on observing the driver's level of diversion by checking optical parameters, health condition, driving example of the driver. It additionally screens street and activity conditions, the sudden entry of animal on the roadways. By a broad experimental assessment utilizing original dataset, the proposed algorithms beat the STM strategy in term of false alarm rate MSE is 2db.