Method and method for monitoring blind zone of vehicle

Abstract

The present invention discloses a learning method of a CNN (Convolutional Neural Network) for monitoring one or more blind zones of a reference vehicle. The learning method includes the following steps: a learning device instructs a detector to output class information and location information on a monitored vehicle in a training image; instructs a cue information extracting layer to output cue information on the monitored vehicle by using the outputted information, and enables the FC layer for blind zone confirmation to use the clue information or the value obtained by processing the clue information to perform a neural network operation to determine whether the monitored target vehicle is located in the blind zone, and carries out backpropagation on the blind zone loss value obtained bycomparing the result of the judgment with its corresponding 1st GT to learn the parameters of the FC layer for blind zone confirmation, and carries out backpropagation on the vehicle detection loss value obtained by comparing the class information with the corresponding 2nd GT to learn the parameters of the detector.

Description

technical field [0001] The present invention relates to a CNN (Convolutional Neural Network, Convolutional Neural Network) learning method for monitoring more than one blind spot of a reference vehicle, and in more detail, relates to a method for monitoring the blind spots of the reference vehicle. The learning method and testing method of CNN and the learning device and testing device utilizing it, wherein, the learning method comprises: step a: when inputting a training image corresponding to at least one video image captured by the reference vehicle, Make the detector of the reference vehicle output the category information and position information about the observation object vehicle included in the training image; Step b: make the clue information extraction layer use the category information and the location information about the observation object vehicle The location information is used to perform a predetermined operation to output one or more clue information about t...

AI Technical Summary

Problems solved by technology

[0010] However, there is a problem that it is necessary to separately design the logic for judging whether a vehicle is in a blind spot based on a vehicle detector used in a blind spot monitoring system using an existing visual sensor.

[0011] In addition, after designing the vehicle detector used in the blind spot monitoring system using the existing visual sensor, it is necessary to design the logic of judging whether the vehicle is in the blind spot according to the output characteristics of the designed vehicle detector, so it is necessary to develop the blind spot monitoring system. Aspects that consume a lot of time

Images