Character identification method and device

1. A character recognition method, comprising:

obtaining a target image to be analyzed which contains a character;

inputting the target image into a pre-trained deep neural network to determine a feature map corresponding to a character region of the target image; and

performing character recognition on the feature map corresponding to the character region by the deep neural network to obtain the character contained in the target image;

wherein the deep neural network is obtained by training with sample images, a result of labeling character regions in the sample images, and characters contained in the sample images;

wherein during the training of the deep neural network, a vector for adjusting character regions is trained with character regions of irregular shapes in the sample images;

wherein the step of determining a feature map corresponding to a character region of the target image comprises:

determining candidate regions in the target image according to a preset segmentation rule;

adjusting a position and/or shape of each candidate region according to the trained vector;

extracting features of the candidate regions to obtain feature maps corresponding to respective candidate regions; and

recognizing a feature map containing a character from the feature maps corresponding to respective candidate regions, and determining the recognized feature map as the feature map corresponding to the character region of the target image.

2. The method according to claim 1 , wherein the step of determining a feature map corresponding to a character region of the target image comprises:

extracting a feature of the target image to obtain a feature map corresponding to the target image; and

analyzing the feature map corresponding to the target image at pixel-level to recognize a region containing a character, and determining a feature map corresponding to the recognized region as the feature map corresponding to the character region in the target image.

3. The method according to claim 1 , wherein the deep neural network comprises at least a convolutional neural network, a recurrent neural network, a classifier and a sequence decoder; and wherein the step of performing character recognition on the feature map corresponding to the character region by the deep neural network to obtain the character contained in the target image comprises:

extracting a feature of the character region at character-level by the convolutional neural network;

extracting a context feature of the character region by the recurrent neural network; and

performing classification and recognition on the extracted feature map by the classifier and the sequence decoder to obtain the character contained in the target image.

4. The method according to claim 1 , wherein the training process of the deep neural network comprises:

obtaining sample images, a result of labeling character regions in the sample images, and characters contained in the sample image; and

training the deep neural network with the sample images, the result of the labeling character regions in the sample images and the characters contained in the sample images.

5. A character recognition device, comprising:

a first obtaining module, configured for obtaining a target image to be analyzed which contains a character;

a determining module, configured for inputting the target image into a pre-trained deep neural network to determine a feature map corresponding to a character region of the target image;

a recognizing module, configured for performing character recognition on the feature map corresponding to the character region by the deep neural network to obtain the character contained in the target image;

a determining sub-module, configured for determining candidate regions in the target image according to a preset segmentation rule;

a first extracting sub-module, configured for extracting features of the candidate regions to obtain feature maps corresponding to respective candidate regions;

a first recognizing sub-module, configured for recognizing a feature map containing a character from the feature maps corresponding to respective candidate regions, and determining the recognized feature map as the feature map corresponding to the character region of the target image; and

an adjusting module, configured for adjusting a position and/or shape of each candidate region according to the trained vector;

wherein the deep neural network is obtained by training with sample images, a result of labeling character regions in the sample images, and characters contained in the sample images;

wherein during the training of the deep neural network, a vector for adjusting character regions is trained with character regions of irregular shapes in the sample images.

6. The device according to claim 5 , wherein the determining module comprises:

a second extracting sub-module, configured for extracting a feature of the target image to obtain a feature map corresponding to the target image; and

a second recognizing sub-module, configured for analyzing the feature map corresponding to the target image at pixel-level to recognize a region containing a character, and determining a feature map corresponding to the recognized region as the feature map corresponding to the character region in the target image.

7. The device according to claim 5 , wherein the deep neural network comprises at least a convolutional neural network, a recurrent neural network, a classifier and a sequence decoder; and wherein the recognizing module comprises:

a third extracting sub-module, configured for extracting a feature of the character region at character-level by the convolutional neural network;

a fourth extracting sub-module, configured for extracting a context feature of the character region by the recurrent neural network; and

a third recognizing sub-module, configured for performing classification and recognition on the extracted feature map by the classifier and the sequence decoder to obtain the character contained in the target image.

8. The device according to claim 5 , wherein the device further comprises:

a second obtaining module, configured for obtaining sample images, a result of labeling character region of the sample images, and characters contained in the sample image; and

a training module, configured for training the deep neural network with the sample images, the result of the labeling character regions of the sample images and the characters contained in the sample images.

9. An electronic device, comprising:

a processor, a memory, a communication interface and a bus, wherein

the processor, the memory and the communication interface are connected and communicate with each other via the bus;

the memory stores executable program codes; and

the processor executes a program corresponding to the executable program codes by reading the executable program codes stored in the memory to carry out the character recognition method according to claim 1 .

10. A non-transitory storage medium, wherein the storage medium is configured to store executable program codes that, when executed, performs the character recognition method according to claim 1 .