Method and electronic device for identifying size of measurement target object
Provided are a method and electronic device for identifying a size of a measurement target object. The method includes imaging a reference object, which is a reference for identifying the size of the measurement target object, to acquire a reference object image, imaging the measurement target object to acquire a target object image, fusing the acquired reference object image and the acquired target object image, and inputting the fused reference object image and target object image to a first neural network model to acquire size information of the measurement target object from the first neural network model.
1. A method of identifying a size of a measurement target object by an electronic device having at least one camera, a processor, and a memory, the method comprising:
capturing, by the at least one camera, a reference object image of a reference object, which is a reference for identifying the size of the measurement target object;
capturing, by the at least one camera, a target object image of the measurement target object;
inputting, by the processor, the target object image to a first Convolutional Neural Network (CNN)-based model;
acquiring, by the processor, a target object mask image generated by the first CNN-based model on the basis of a target object area in the target object image;
fusing, by the processor, the target object mask image and the reference object image, to generate a fused image;
inputting, by the processor, the fused image to a second CNN-based neural network, which is different from the first CNN-based model;
acquiring, by the processor, size information of the measurement target object from the second CNN-based neural network.
2. The method of claim 1 , wherein the first CNN-based model is configured to identify an object area including each of the at least one measurement target object in the target object image, which is input to the first CNN-based model and includes the at least one measurement target object, generate an object box including the object area, identify a type of the measurement target object represented by the object box, and binarize an image input to the first CNN-based model on the basis of the object area to generate the target object mask image.
3. The method of claim 2 , wherein the second CNN-based neural network is configured to acquire the target object mask image and the reference object image, compare the acquired target object mask image and the acquired reference object image, and output information on the type and size of the measurement target object, which is represented by the object box in the target object image including the measurement target object, on the basis of a comparison result.
4. The method of claim 1 , wherein the first CNN-based model and the second CNN-based neural network are trained in advance on the basis of training data which is generated using a preset computer aided design (CAD) model and a camera characteristic model including a preset camera characteristic parameter in a virtual environment for virtually imaging the measurement target object and the reference object.
5. The method of claim 4 , wherein the training data is generated using the CAD model and the camera characteristic model while changing domain information which varies on the basis of a path along which virtual light is reflected by the measurement target object or the reference object.
6. The method of claim 1 , further comprising acquiring information on a type of the measurement target object and an area of the measurement target object in the target object image from the second CNN-based neural network.