用Tesseract训练验证码遇到的问题

1.准备验证码图片

import os
from urllib.request import urlretrieve

urlPath='http://www.189.cn/portal/captcha/simple.do?date=1503125232339'
localPath='G:python_workchapter11captcha'

# 根据文件名创建文件
def createFileByFileName(localPath,fileName):
    if not os.path.exists(localPath):
        os.makedirs(localPath)
    totalPath=localPath+'\'+fileName
    if not os.path.exists(totalPath):
        file=open(totalPath, 'a+')
        file.close()
        return totalPath

# 根据图片的地址, 下载图片并保存在本地
def getAndSaveImg(imgUrl, filename):
    if(len(imgUrl)!=0):
        file=filename+'.jpg'
        urlretrieve(imgUrl, createFileByFileName(localPath, file))

# 下载图片
i=0
while i<100:
    getAndSaveImg(urlPath, str(i))
    i=i+1

2.将验证码图片转为灰度图 , 扩展名改为.tif

import os
from PIL import Image

def getFileList(path):
    # 返回目录中所有jpg图像的文件名列表
    return [os.path.join(path, f) for f in os.listdir(path) if f.endswith('.jpg')]

# 批量将图片转为灰度图
def patchConvert(srcPath, destPath):
    fileList = getFileList(srcPath)
    for file in fileList:
        img = Image.open(file).convert('L')
        name_ext = getFileNameAndExt(file)
        destFile = os.path.join(destPath, name_ext[0]+'.tif')
        img.save(destFile)

# 获取文件名和扩展名
def getFileNameAndExt(filename):
    # 取目录与文件名
    (filepath, tempfilename) = os.path.split(filename)
    # 取文件名(不带扩展名)和扩展名
    (shotname, extension) = os.path.splitext(tempfilename)
    return shotname, extension

patchConvert('captcha', 'captcha_gray')

3. 后续步骤参考原先的博客:

http://www.cnblogs.com/CoolJayson/p/7395824.html

问题: 生成box文件后, 用jTessBoxEditor修改box文件时, 里面的矩形画的误差较大, 字符个数识别不准确 , 应该如何对矩形进行修改???