OpenCV（九）--文字扫描OCR识别-Toy模板网

这篇具有很好参考价值的文章主要介绍了OpenCV（九）--文字扫描OCR识别。希望对大家有所帮助。如果存在错误或未考虑完全的地方，请大家不吝赐教，您也可以点击"举报违法"按钮提交疑问。

步骤：边缘检测+计算轮廓+变换+OCR

def show(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

img = cv2.imread('tip.png')
show('img', img)

opencv ocr,OpenCV,opencv,计算机视觉,人工智能

img = cv2.resize(img, (500, int((500 * img.shape[0])/img.shape[1])))
# 预处理
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (5, 5), 0)
# 边缘检测
edged = cv2.Canny(gray, 75, 200)
show('edged', edged)

opencv ocr,OpenCV,opencv,计算机视觉,人工智能

# 轮廓处理
cnts = cv2.findContours(edged.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0]
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:5]  # reverse=True表示倒序(从大到小)

for c in cnts:
    # 计算轮廓近似
    peri = cv2.arcLength(c, True)
    approx = cv2.approxPolyDP(c, 0.02*peri, True)
    # 如果近似为一个矩形
    if len(approx) == 4:
        screenCnt = approx
        break

cv2.drawContours(img, [screenCnt], -1, (0, 255, 0),2)
show('img', img)

opencv ocr,OpenCV,opencv,计算机视觉,人工智能

# 透视变换
def four_point_transform(img, pts):
    p = sorted(pts, key=lambda pts: pts[0, 1], reverse=False)
    p = sorted(p, key=lambda p: p[0, 0], reverse=False)
    tl, bl, tr, br = [i[0] for i in p]
    # 计算边长度
    wA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    wB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    mw = max(int(wB), int(wA))+1

    hA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    hB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    mh = max(int(hB), int(hA))+1

    # 变换后对应坐标位置
    dst = np.array([
        [0, 0],
        [mw-20, 0],
        [mw-20, mh-20],
        [0, mh-20]
    ], dtype="float32")
    # 计算变换矩阵
    rect = np.zeros((4,2), dtype="float32")
    rect[0] = tl.astype(float).tolist()
    rect[1] = tr.astype(float).tolist()
    rect[2] = br.astype(float).tolist()
    rect[3] = bl.astype(float).tolist()
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(img, M, (mw, mh))
    return warped

warped = four_point_transform(orig, screenCnt)
show('warped', warped)
# 二值处理
warped = cv2.cvtColor(warped, cv2.COLOR_BGR2GRAY)
ref = cv2.threshold(warped, 180, 255, cv2.THRESH_BINARY)[1]
show('ref', ref)

opencv ocr,OpenCV,opencv,计算机视觉,人工智能

OCR识别

import pytesseract
from PIL import Image
cv2.imwrite("ocr.png", ref)
# 下载地址：https://github.com/UB-Mannheim/tesseract/wiki
# tesseract.exe的路径
pytesseract.pytesseract.tesseract_cmd = 'C:\Program Files\Tesseract-OCR\\tesseract.exe'
# 指定训练集的路径
tessdata_dir_config = r'--tessdata-dir "C:\Program Files\Tesseract-OCR\tessdata"'
test =pytesseract.image_to_string(Image.open("ocr.png"))
print(test)

识别结果：
opencv ocr,OpenCV,opencv,计算机视觉,人工智能

完整代码文章来源地址https://www.toymoban.com/news/detail-799945.html

import cv2  # opencv读取的格式是BGR
import matplotlib.pyplot as plt
import numpy as np

def show(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

img = cv2.imread('tip.png')
# show('img', img)
ratio = img.shape[0] / 500.0
orig = img.copy()

img = cv2.resize(img, (500, int((500 * img.shape[0])/img.shape[1])))
# 预处理
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (5, 5), 0)
# 边缘检测
edged = cv2.Canny(gray, 75, 200)
# show('edged', edged)
# 轮廓处理
cnts = cv2.findContours(edged.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)[0]
cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:5]  # reverse=True表示倒序(从大到小)

for c in cnts:
    # 计算轮廓近似
    peri = cv2.arcLength(c, True)
    approx = cv2.approxPolyDP(c, 0.02*peri, True)
    # 如果近似为一个矩形
    if len(approx) == 4:
        screenCnt = approx
        break

cv2.drawContours(img, [screenCnt], -1, (0, 255, 0), 2)
# show('img', img)

# 透视变换
def four_point_transform(img, pts):
    p = sorted(pts, key=lambda pts: pts[0, 1], reverse=False)
    p = sorted(p, key=lambda p: p[0, 0], reverse=False)
    tl, bl, tr, br = [i[0] for i in p]
    # 计算边长度
    wA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
    wB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
    mw = max(int(wB), int(wA))+1

    hA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
    hB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
    mh = max(int(hB), int(hA))+1

    # 变换后对应坐标位置
    dst = np.array([
        [0, 0],
        [mw-20, 0],
        [mw-20, mh-20],
        [0, mh-20]
    ], dtype="float32")
    # 计算变换矩阵
    rect = np.zeros((4,2), dtype="float32")
    rect[0] = tl.astype(float).tolist()
    rect[1] = tr.astype(float).tolist()
    rect[2] = br.astype(float).tolist()
    rect[3] = bl.astype(float).tolist()
    M = cv2.getPerspectiveTransform(rect, dst)
    warped = cv2.warpPerspective(img, M, (mw, mh))
    return warped

warped = four_point_transform(orig, screenCnt)
# show('warped', warped)
# 二值处理
warped = cv2.cvtColor(warped, cv2.COLOR_BGR2GRAY)
ref = cv2.threshold(warped, 180, 255, cv2.THRESH_BINARY)[1]
show('ref', ref)
import pytesseract
from PIL import Image
cv2.imwrite("ocr.png", ref)
# 下载地址：https://github.com/UB-Mannheim/tesseract/wiki
# tesseract.exe的路径
pytesseract.pytesseract.tesseract_cmd = 'C:\Program Files\Tesseract-OCR\\tesseract.exe'
# 指定训练集的路径
tessdata_dir_config = r'--tessdata-dir "C:\Program Files\Tesseract-OCR\tessdata"'
test =pytesseract.image_to_string(Image.open("ocr.png"))
print(test)

到了这里，关于OpenCV（九）--文字扫描OCR识别的文章就介绍完了。如果您还想了解更多内容，请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章，希望大家以后多多支持TOY模板网！