An alarm method for dangerous objects in the blind area of ​​automobiles based on deep learning

Abstract

The invention proposes a method for alarming dangerous objects in blind spots of automobiles based on deep learning. The invention includes a deep learning-based dangerous object alarm system in the blind area of ​​an automobile. The method of the present invention trains the network model; uses a plurality of cameras to separately collect the images of the blind spots of the driving car, and transmits them to the corresponding processor; the corresponding processor inputs the pre-processed image into the trained network model, and detects the The category, confidence level and position coordinates of the dangerous object; calculate the horizontal distance from the dangerous object to the corresponding camera; The location coordinates are transmitted to the voice converter respectively, and a danger warning voice is generated, which is broadcast by the car audio. The invention achieves the effect of real-time alarm in the blind area, and effectively reduces accidents caused by blocking small objects such as children and small animals in the blind area.

Description

technical field [0001] The invention belongs to the field of computer vision technology and the field of intelligent driving assistance, and in particular relates to a method for alarming dangerous objects in blind spots of automobiles based on deep learning. Background technique [0002] The proportion of road traffic accidents is increasing year by year, and the driver's subjective judgment error is an important factor in the occurrence of accidents, but there are many factors of the vehicle itself that affect the driver's judgment. On the one hand, due to the design of the car itself, there are many blind spots on the way of driving, especially when turning, there will be A / B / C pillar blind spots, front / rear blind spots, rearview mirror blind spots, etc., and the driver cannot pass through the rearview mirror. When seeing dangerous objects in the blind spot, even if the car is equipped with rearview mirrors and some supplementary blind-sight mirrors, due to the limitation...

AI Technical Summary

Problems solved by technology

Mainstream target detection methods such as Faster R-CNN, R-FCN, SSD, etc. However, in the trade-off between accuracy and detection speed, these methods are more or less deficient. In the automotive application that requires strict accuracy and speed It is limited, especially the test effect on small targets is not ideal

Images