发票分割总结

最新推荐文章于 2024-10-25 14:26:05 发布

点灯师傅

最新推荐文章于 2024-10-25 14:26:05 发布

阅读量3.7k

点赞数

文章标签：计算机视觉机器学习 python

本文链接：https://blog.csdn.net/weixin_43811282/article/details/121703474

版权

这篇博客主要探讨了发票分割的过程，包括预处理步骤，利用最大连通区域进行图像分割，以及如何去除边界上的小凸起。此外，还介绍了如何通过拟合直线来处理发票的四条边。

摘要由CSDN通过智能技术生成

预处理

import argparse
import os
import json
from glob import glob

from PIL import Image
import numpy as np

def get_args():
    parser = argparse.ArgumentParser()
    parser.add_argument('--input', default='/Users/yi/Desktop/vehicle_train_mask/')
    parser.add_argument('--out_dir', default='/Users/yi/Desktop/annotations_vehicle_train_mask/')
    return parser.parse_args()


def imgs(jpg_files):
    if not os.path.isdir(get_args().out_dir):
        os.makedirs(get_args().out_dir)

    for jpg_name in jpg_files:
        imagePath = get_args().input + jpg_name + ".png"
        outPath = get_args().out_dir + jpg_name + ".png"
        img = Image.open(imagePath)
        img_array = np.array(img)  # 把图像转成数组格式img = np.asarray(image)
        shape = img_array.shape
        print(img_array)
        print(img_array.shape)
        height = shape[0]
        width = shape[1]
        dst = np.zeros((height, width, 3))
        for h in range(0, height):
            for w in range(0, width):
                r = img_array[h, w]
                # print(r)
                if r == 0:
                    dst[h, w] = (0, 0, 0)
                else:
                    dst[h, w] = (1, 1, 1)  # 替换为label
        #         (r, g, b) = img_array[h, w]
        #         if (r, g, b) == (255, 0, 0):
        #             img_array[h, w] = (1, 1, 1)  # 替换为label
        #         dst[h, w] = img_array[h, w]

        img2 = Image.fromarray(np.uint8(dst))
        img2.save(outPath)


def main(args):
    files = glob(args.input+ "*.png")
    print(files)
    files = [i.split("/")[-1].split(".png")[0] for i in files]
    test_files = files
    imgs(test_files)


if __name__ == '__main__':
    main(get_args())

最大连通区域

import cv2
import os
import numpy as np

imaPath = '/Users/yi/Desktop/test_20210903_out/'
output = '/Users/yi/Desktop/test_20210903_out1/'
#
# import cv2
imaList = os.listdir(imaPath)
for files in imaList:
    img = os.path.join(imaPath, files)
    path_processed = os.path.join(output, files)

    img = cv2.imread(img)

    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    # cv2.imshow("img", gray)
    # cv2.waitKey(1000)
    ret, binary = cv2.threshold(gray, 156, 255, cv2.THRESH_BINARY)
    # cv2.imshow("img", binary)
    # cv2.waitKey(1000)
    contours, hierarchy = cv2.findContours(binary, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE)
    # 找到最大区域并填充
    area = []

    for j in range(len(contours)):
        area.append(cv2.contourArea(contours[j]))

    max_idx = np.argmax(area)
    max_area = cv2.contourArea(contours[max_idx])
    # print(max_area)
    for k in range(len(contours)):
        if k != max_idx:
            print(k, 1)
            cv2.fillPoly(img, [contours[k]], (198,215,20))

    cv2.drawContours(img, contours[max_idx], -1, (0, 0, 255), 3)

    cv2.imwrite(path_processed, img)

删除边界小凸起

import argparse
from glob import glob
import cv2
import os
import numpy as np
from PIL import Image
def get_args():
    parser = argparse.ArgumentParser()
    parser.add_argument('--input', default='/Users/yi/Desktop/tu/')
    parser.add_argument('--mask_dir', default='/Users/yi/Desktop/tu_out/')
    parser.add_argument('--out_dir', default='/Users/yi/Desktop/tu_out_wx_result/')
    return parser.parse_args()


def imgs(jpg_files):
    print(jpg_files)
    if not os.path.isdir(get_args().mask_dir)