Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
|
8 years ago | |
---|---|---|
Font | 8 years ago | |
cache | 8 years ago | |
dataset | 8 years ago | |
hyperlpr | 8 years ago | |
model | 8 years ago | |
templates | 8 years ago | |
.gitignore | 8 years ago | |
README.md | 8 years ago | |
batch.py | 8 years ago | |
benchmark.py | 8 years ago | |
upload.py | 8 years ago | |
wxpy_uploader.py | 8 years ago |
HyperLPR是一个基于Python的使用深度学习针对对中文车牌识别的实现,与开源的EasyPR相比,它的检测速度和鲁棒性和多场景的适应性都要好于EasyPR。
step1. 使用opencv 的 HAAR cascade 检测车牌大致位置
step2. Extend 检测到的大致位置的矩形区域
step3. 使用类似于mser的方式的多级二值化+ransac拟合车牌的上下边界
step4. 使用CNN regression回归车牌左右边界
step5. 使用基于纹理场的算法进行车牌校正倾斜
step6. 使用CNN滑动窗切割字符
step7. 使用CNN识别字符
from hyperlpr import pipline as pp
import cv2
image = cv2.imread("filename")
image,res = pp.SimpleRecognizePlate(image)
高性能开源中文车牌识别框架
C++ Java C Python CMake other