A CNN-based handwritten English document recognition method
A recognition method and English technology, applied in the field of computer vision, to achieve the effect of improving recognition accuracy, saving training time, and rapid network structure convergence
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0040] A CNN-based handwritten English document recognition method, see figure 1 , the method includes the following steps:
[0041] 101: Obtain a data set composed of handwritten English letters and punctuation marks, and construct a training sample set and a test sample set based on the data set;
[0042] 102: Construct an 8-layer convolutional neural network, including 5 convolutional layers and 3 fully connected layers, and the output of the last fully connected layer is sent to a softmax layer with 59 output vectors;
[0043] 103: Using overlapping Pooling, a Pooling layer is composed of Pooling unit grids with an interval of s pixels, each grid has a z*z size proximity relationship, all located in the center of the Pooling unit, s<z; yes Convolution, downsampling, and pooling operations are performed on each pixel of the input image to obtain the size of the feature map of each layer;
[0044] 104: Input the training sample set, extract character features, and perform ...
Embodiment 2
[0052] The following is combined with specific examples, calculation formulas, Figure 1-Figure 5 The scheme in Example 1 is further introduced, see the following description for details:
[0053] 201: Obtain a sample set of English letters and punctuation marks;
[0054] First, orthorectify the handwritten English document image; then use the projection method [1] Carry out character segmentation on the measured English handwritten document: first perform the horizontal projection of the image to obtain the projection histogram in the horizontal direction. According to the projection histogram, divide each line of the image in the handwritten English document; The text image cut out by the line is vertically projected to obtain the projection histogram in the vertical direction, and a single character image is segmented according to the principle of dichotomy [2] .
[0055] Among them, the handwritten English letters and related punctuation texts of 30 people were collecte...
Embodiment 3
[0086] Combine below Figure 4-Figure 5 , and Table 1 and Table 2, the scheme in Embodiment 1 and 2 is further introduced, see the following description for details:
[0087] The data set includes: 52 English letters in upper and lower case, 6 commonly used punctuation marks (",.?!:), and space bar, basically covering all the characters that may appear in the recognition of handwritten English documents, a total of 59 characters. Each person writes each character 5 times, 4 of which are selected as the training sample set, and the remaining 1 is used as the test sample set I, and finally 6790 training sample sets and 1180 test sample sets are obtained.
[0088] The data set does not need to go through complicated preprocessing steps, and steps such as binarization, noise removal, and tilt correction are omitted. The image size and type are unified to 320*320*3unit8, which helps to improve the stability of training samples. The constructed neural network uses the VGG model to ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More