{
"word": "80.35",
"pos": [{
"x": 1033,
"y": 1130
}, {
"x": 1089,
"y": 1130
}, {
"x": 1089,
"y": 1146
}, {
"x": 1033,
"y": 1146
}]
}, {
"word": "哈哈",
"pos": [{
"x": 1133,
"y": 1132
}, {
"x": 1177,
"y": 1132
}, {
"x": 1177,
"y": 1151
}, {
"x": 1133,
"y": 1151
}]
}
元数据为这样的json结构，四个点的坐标和对应的字，我们现在要通过这四个坐标进行图片切割
# 切割图片 输入是[[x1,y1,x2,y2,x3,y3,x4,y4],]
def crop_small_images(img,polygens):
logger.debug("图像：%r" , img.shape)
cropped_images = []
for pts in polygens:
# crop_img = img[y:y+h, x:x+w]
logger.debug("子图坐标：%r",pts)
pts_np = np.array(pts)
pts_np = pts_np.reshape(4,2)
# print(pts_np)
min_xy = np.min(pts_np,axis=0)
max_xy = np.max(pts_np,axis=0)
# print(min_xy[0],min_xy[1],max_xy[0],max_xy[1])
crop_img = img[min_xy[1]:max_xy[1],min_xy[0]:max_xy[0]]
cropped_images.append(crop_img)
return cropped_images
运行bin中的sh文件，参数是data中的子文件夹名，如要切割的数据是 /split/data/20190505。那么执行命令为
sh start.sh 20190505
一张图片的执行结果为 一个txt文件和多个图片文件，txt记录了哪个汉字对应的图片名是什么