【OpenCV】 ⚠️實戰⚠️ 銀行卡卡号讀取

• ​​概述​​
• ​​預處理​​

• ​​代碼​​
• ​​模闆預處理​​
• ​​銀行卡預處理​​

• ​​計算輪廓​​

• ​​代碼​​
• ​​模闆輪廓​​
• ​​銀行卡輪廓​​

• ​​其他程式​​
• ​​主函數​​

• ​​代碼​​
• ​​數字分割​​
• ​​最終結果​​

概述

今天帶大家使用我們之前學會的知識來實作銀行卡卡号讀取. 代碼分為四個部分: 主函程式, 預處理, 計算輪廓, 其他程式.

預處理

通過灰階轉換, 二值化, 膨脹, 腐蝕, 邊緣檢測等方法, 去除圖檔噪聲, 突出我們想要得到的結果.

代碼

import numpy as np
import cv2
from matplotlib import pyplot as plt
from my_functions import resize


def read_template(image_path, visualize=False):
    """
    讀取模闆
    :param image_path: 圖檔路徑
    :param visualize: 可視化, 預設為False
    :return: 傳回模闆, 二值化後的模闆
    """

    # 讀取模闆
    template = cv2.imread(image_path)

    # 轉換成灰階圖
    template_gray = cv2.cvtColor(template, cv2.COLOR_BGR2GRAY)

    # 二值化
    ret, template_thresh = cv2.threshold(template_gray, 10, 255, cv2.THRESH_BINARY_INV)

    # 如果展示為真
    if visualize:
        """圖檔展示"""

        # 繪制子圖
        f, ax = plt.subplots(3, 1, figsize=(10, 8))
        ax[0].imshow(template)
        ax[1].imshow(template_gray, "gray")
        ax[2].imshow(template_thresh, "gray")

        # 标題
        ax[0].set_title("template")
        ax[1].set_title("template gray")
        ax[2].set_title("template binary inverse")

        plt.show()

    # 傳回
    return template, template_thresh


def read_image(image_path, visualize=False):
    """
    讀取銀行卡圖檔
    :param image_path: 圖檔路徑
    :param visualize: 可視化, 預設為False
    :return: 傳回裁剪後的圖檔, 灰階圖, 處理後的圖
    """

    # 初始化卷積核
    rectKernel = cv2.getStructuringElement(cv2.MORPH_RECT, (9, 3))
    sqKernel = cv2.getStructuringElement(cv2.MORPH_RECT, (5, 5))

    # 讀取圖檔
    image = cv2.imread(image_path)

    # 更改尺寸
    image_resize = resize(image, width=300)

    # 轉換成灰階圖
    image_gray = cv2.cvtColor(image_resize, cv2.COLOR_BGR2GRAY)

    # 禮帽操作, 突出明亮區域
    tophat = cv2.morphologyEx(image_gray, cv2.MORPH_TOPHAT, rectKernel)

    # Sobel邊緣檢測
    edge = cv2.Sobel(tophat, ddepth=cv2.CV_32F, dx=1, dy=0, ksize=-1)
    edge = np.absolute(edge)

    # 标準化
    edge = 255 * cv2.normalize(edge, None, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_32F)
    edge = edge.astype("uint8")

    # 通過閉操作（先膨脹，再腐蝕）将數字連在一起
    edge_close = cv2.morphologyEx(edge, cv2.MORPH_CLOSE, rectKernel)

    # THRESH_OTSU會自動尋找合适的門檻值，适合雙峰，需把門檻值參數設定為0
    ret, thresh = cv2.threshold(edge_close, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)

    # 再來一個閉操作
    thresh_close = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, sqKernel)  # 再來一個閉操作

    # 如果展示為真
    if visualize:
        """圖檔展示"""

        # 繪制子圖
        f, ax = plt.subplots(2, 2, figsize=(12, 8))
        ax[0, 0].imshow(cv2.cvtColor(image_resize, cv2.COLOR_BGR2RGB))
        ax[0, 1].imshow(image_gray, "gray")
        ax[1, 0].imshow(tophat, "gray")
        ax[1, 1].imshow(edge, "gray")

        # 标題
        ax[0, 0].set_title("image resize")
        ax[0, 1].set_title("image gray")
        ax[1, 0].set_title("image tophat")
        ax[1, 1].set_title("image edge")

        plt.show()

        # 繪制子圖
        f, ax = plt.subplots(2, 2, figsize=(12, 8))
        ax[0, 0].imshow(edge, "gray")
        ax[0, 1].imshow(edge_close, "gray")
        ax[1, 0].imshow(thresh, "gray")
        ax[1, 1].imshow(thresh_close, "gray")

        # 标題
        ax[0, 0].set_title("image edge")
        ax[0, 1].set_title("image close")
        ax[1, 0].set_title("image binary")
        ax[1, 1].set_title("image binary close")

        plt.show()

    # 傳回
    return image_resize, image_gray, thresh      

模闆預處理

銀行卡預處理

計算輪廓

代碼

import cv2
from matplotlib import pyplot as plt, gridspec
from my_functions import sort_contours


def template_calculate_contours(template, template_binary, visualize=False):
    """
    計算模闆輪廓
    :param template: 模闆
    :param template_binary: 二值化的模闆
    :return: 輪廓
    """

    # 擷取輪廓
    contours, hierarchy = cv2.findContours(template_binary.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    # 輪廓排序, 從左到右, 從上到下
    contours = sort_contours(contours)

    digits = {}

    # 周遊每一個輪廓
    for (i, c) in enumerate(contours):
        # 計算矩陣
        (x, y, w, h) = cv2.boundingRect(c)

        # 擷取roi
        roi = template_binary[y:y + h, x:x + w]
        roi = cv2.resize(roi, (57, 88))

        # 每一個數字對應每一個模闆
        digits[i] = roi

    if visualize:
        """圖檔展示"""

        # 繪制子圖
        plt.figure(figsize=(12, 6))
        gs = gridspec.GridSpec(2, 10)

        # 軸1
        plt.subplot(gs[0, :10])
        plt.imshow(template, "gray")
        plt.xticks([])
        plt.yticks([])
        plt.title("original")

        # 軸2
        for (number, image) in digits.items():
            plt.subplot(gs[1, number])
            plt.xticks([])
            plt.yticks([])
            plt.imshow(image, "gray")
            plt.title("number: {}".format(number))

        plt.show()

    print(digits)
    return digits


def image_calculate_contours(image, thresh, visualize=False):
    """
    計算輪廓
    :param image: 圖檔
    :param thresh: 處理後的圖檔
    :param visualize: 可視化, 預設為False
    :return: 輪廓
    """

    # 擷取輪廓
    contours, hierarchy = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    # 繪制輪廓
    cur_img = image.copy()
    image_with_contour = cv2.drawContours(cur_img, contours, -1, (0, 0, 255), 3)

    # 位置
    locations = []

    # 周遊輪廓
    for (i, c) in enumerate(contours):

        # 計算矩形
        (x, y, w, h) = cv2.boundingRect(c)
        ar = w / float(h)

        # 選擇合适的區域，根據實際任務來，這裡的基本都是四個數字一組
        if ar > 2.5 and ar < 4.0:

            if (w > 40 and w < 55) and (h > 10 and h < 20):
                # 符合的留下來
                locations.append((x, y, w, h))

    # 将符合的輪廓從左到右排序
    locs = sorted(locations, key=lambda x: x[0])

    if visualize:
        """圖檔展示"""

        # 繪制子圖
        f, ax = plt.subplots(2, 1, figsize=(12, 8))
        ax[0].imshow(cv2.cvtColor(image_with_contour, cv2.COLOR_BGR2RGB))
        for r in locs:
            (x, y, w, h) = r
            rectangle = cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 2)
            ax[1].imshow(cv2.cvtColor(rectangle, cv2.COLOR_BGR2RGB))

        # 标題
        ax[0].set_title("original")
        ax[1].set_title("detect")

        plt.show()

    return locs      

模闆輪廓

銀行卡輪廓

其他程式

import cv2


def img_show(name, img):
    """圖檔展示"""

    cv2.imshow(name, img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()


def sort_contours(contours):
    """
    輪廓排序 (從左到右)
    :param contours: 輪廓
    :return: 傳回排序好的輪廓
    """

    boundingBoxes = [cv2.boundingRect(c) for c in contours]  # 用一個最小的矩形，把找到的形狀包起來x,y,h,w
    (contours, boundingBoxes) = zip(*sorted(zip(contours, boundingBoxes), key=lambda b: b[1][0]))
    return contours


def resize(image, width=None, height=None, inter=cv2.INTER_AREA):
    """
    修改圖檔大小
    :param image: 原圖
    :param width: 寬
    :param height: 高
    :param inter: 模式
    :return: 修改好的圖檔
    """

    dim = None
    (h, w) = image.shape[:2]
    if width is None and height is None:
        return image
    if width is None:
        r = height / float(h)
        dim = (int(w * r), height)
    else:
        r = width / float(w)
        dim = (width, int(h * r))
    resized = cv2.resize(image, dim, interpolation=inter)
    return resized      

主函數

代碼

import numpy as np
import argparse
import cv2
from matplotlib import pyplot as plt
from my_functions import sort_contours
from pre_process import read_template
from pre_process import read_image
from calculate_contours import template_calculate_contours
from calculate_contours import image_calculate_contours


def extract_number(image_gray, locations, digits, visualize=False):
    """
    提取數字
    :param image_gray: 灰階圖
    :param locations: 圖檔輪廓
    :param digits: 模闆輪廓
    :param visualize: 可視化, 預設為False
    :return: 讀取完數字的圖檔
    """

    # 輸出
    output = []

    # 圖檔
    total_img = []

    # 周遊每一個輪廓中的數字
    for (i, (gX, gY, gW, gH)) in enumerate(locations):
        # 組輸出
        groupOutput = []
        group_img = []

        # 根據坐标提取每一個組
        group = image_gray[gY - 5:gY + gH + 5, gX - 5:gX + gW + 5]

        # 預處理
        group_binary = cv2.threshold(group, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]

        # 計算每一組的輪廓
        digitCnts, hierarchy = cv2.findContours(group_binary.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
        digitCnts = sort_contours(digitCnts)

        # 計算每一組中的每一個數值
        for c in digitCnts:
            # 找到目前數值的輪廓，resize成合适的的大小
            (x, y, w, h) = cv2.boundingRect(c)
            roi = group[y:y + h, x:x + w]
            roi = cv2.resize(roi, (57, 88))
            group_img.append(roi)

            # 計算比對得分
            scores = []

            # 在模闆中計算每一個得分
            for (digit, digitROI) in digits.items():
                # 模闆比對
                result = cv2.matchTemplate(roi, digitROI, cv2.TM_CCOEFF)
                (_, score, _, _) = cv2.minMaxLoc(result)
                scores.append(score)

            # 得到最合适的數字
            groupOutput.append(str(np.argmax(scores)))

        # 畫出來
        cv2.rectangle(image, (gX - 5, gY - 5), (gX + gW + 5, gY + gH + 5), (0, 0, 255), 1)
        cv2.putText(image, "".join(groupOutput), (gX, gY - 15), cv2.FONT_HERSHEY_SIMPLEX, 0.65, (0, 0, 255), 2)

        # 得到結果
        output.extend(groupOutput)

        # 添加圖檔
        total_img.append(group_img)

    if visualize:
        """圖檔展示"""

        # 繪制子圖
        f, ax = plt.subplots(4, 4, figsize=(8, 8))
        for i, group_im in enumerate(total_img):
            for j, im in enumerate(group_im):
                ax[i, j].imshow(im, "gray")
                ax[i, j].set_xticks([])
                ax[i, j].set_yticks([])
                ax[i, j].set_title("group: {}".format(i + 1))

        plt.show()

    # 展示最終圖檔
    plt.figure(figsize=(10, 8))
    plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    plt.title("Final Result")

    plt.show()

    return image


def parse_opt():
    """設定參數"""

    parser = argparse.ArgumentParser()

    parser.add_argument("--image_path", type=str, default="images/credit_card_01.png", help="輸入圖檔路徑")
    parser.add_argument("--template_path", type=str, default="template/template.png", help="模闆圖檔路徑")

    args = parser.parse_args()

    return args


if __name__ == "__main__":
    args = parse_opt()

    # 讀取模闆
    template, template_binary = read_template(args.template_path, True)

    # 計算模闆輪廓
    digits = template_calculate_contours(template=template, template_binary=template_binary, visualize=True)

    # 讀取圖檔
    image, image_gray, thresh = read_image(args.image_path, visualize=True)

    # 計算圖檔輪廓
    locations = image_calculate_contours(image, thresh, visualize=True)

    # 提取數字
    result = extract_number(image_gray=image_gray, locations=locations, digits=digits, visualize=True)

    # 儲存最終結果
    cv2.imwrite("Final_result.png", result)      

數字分割

最終結果