Object recognition apparatus for selecting neural network models according to environment and method thereof

摘要

The present invention relates to a device for recognizing an object selecting a neural network model according to an environment. According to the present invention, the device comprises: an image acquisition unit which acquires an image from a camera capturing the environment outside the vehicle; a brightness value calculation unit which calculates a brightness value of the current time point from the captured image; an environment value calculation unit which trains a classification model by using images of rain, snow and fog, classifies the current environmental state by applying the image of the current point of time to the classification model completed with training, and calculates the grade of the intensity of the classified environmental state; a storage unit which stores a plurality of neural network models having different number of layers; a control unit which calculates the environmental index by using the rating value for the brightness of the current time and the rating value for the environmental state, and selects the number of layers corresponding to the calculated environmental index; a neural network model selection unit which selects a neural network model corresponding to the number of the selected layers among the plurality of neural network models; and an object recognition unit which recognizes an object located in the front by using the selected neural network model. According to the present invention, the device for recognizing an object can reduce power consumption by using a neural network model with fewer layers in good weather conditions, and can increase the recognition rate of the object by using a neural network model with a large number of layers in bad weather conditions.