Target object identification method and device

Abstract

The invention discloses a target object identification method and device, and relates to the technical field of super deep learning. The specific implementation mode of the method comprises the following steps: acquiring current priorities of a plurality of to-be-detected target objects included in a vehicle end image; wherein the current priority is determined according to the driving data of the vehicle; determining a plurality of preprocessing areas of the vehicle end image according to a target input scale and a current priority of a pre-trained image detection model; wherein the preprocessing area comprises one or more to-be-detected target objects; and according to the plurality of preprocessing areas, processing the vehicle end image, and taking a processing result as input of an image detection model to identify a target object. According to the embodiment, more target objects with higher priorities can be covered in a limited detection area, and extra vehicle end resources are not consumed, so that multi-area detection of the target objects is realized, the user experience is improved, and the driving safety and the vehicle end service stability are guaranteed.

Description

technical field [0001] The present invention relates to the technical field of image processing, and in particular, to a method and device for identifying a target object. Background technique [0002] Image identification refers to the use of deep learning algorithms to process, analyze and understand images to identify objects and objects in various patterns. [0003] In the existing vehicle driving process, in order to identify the target object (for example, a traffic light) in the field of view, the vehicle-end server usually determines the area of ​​interest, adjusts the area of ​​interest to the scale required for model detection through the resize function, and then conducts the model. detection. [0004] However, under the limited detection area of ​​the model, if there are multiple target objects in the field of view, the existing cropping methods cannot take into account, which may result in the loss of the target; if the method of increasing the batch size is us...

AI Technical Summary

Problems solved by technology

[0004] However, under the limited detection area of ​​the model, if there are multiple target objects in the field of view, the existing cropping method cannot be considered, which may cause target loss; if the method of increasing the batch size is used for detection, the vehicle end that needs to be occupied The larger the resources, the greater the load pressure on the car-side server, the response delay, etc., the poor user experience, and even the crash of the car-side server in severe cases, causing security incidents

Images