Intelligent safe driving solution based on multi-source information fusion

At present, the existing safe driving monitoring system technologies, such as passive technology (based on vehicle information) and active technology (based on biosensors and visual sensors), cannot directly detect drivers, and are prone to false positives and high costs. Therefore, based on deep learning, hardware acceleration, and multi-sensor technology, this paper studies the intelligent safe driving solution of multi-source information fusion, improves and integrates the existing Retinaface and Retinanet solutions, proposes a line of sight direction classification algorithm based on feature points and a fatigue index algorithm based on multi-scale features, and develops a set of Alcohol, GPS and other multi-sensor fusion vehicle monitoring FPGA module is equipped with a hardware accelerator based on convolutional neural network. A visual application program based on MVVM and a wearable device module based on the multi-sensor fusion of temperature, blood oxygen, and heart rate are developed. This scheme can realize the integration of passive technology and active technology and has the function of a collaborative analysis of vehicle data, driver health data, and attitude information. It can cover a variety of scenes at the same time, with high transmission security and greatly improved accuracy.