如何使用jTessBoxEditor提高文本识别的准确性？

Question

我一直在尝试从扫描的pdf文档中提取数据。 我已经将pdf文件转换为jpeg文件（我已在下面的图片链接中附加了内容），使用不同的字体裁剪了单词和数字，合并为tiff文件，并使用jTessBoxEditor训练了字体以生成新的语言，并且我在Tesseract-OCR从文件中提取数据。但是我无法提取确切的数据。 tesseract-ocr的文本识别精度非常低。

有人可以提出一种提高准确性的方法吗？

Answer 1

我无法提取确切数据

您使用image_to_string方法吗？

text = pytesseract.image_to_string(image)

部分输出：

Prepared By: Pat Schelbel Inst No:95018369 Date:12/27/1995
Post Office Address: + Barnett Bank of Volusia County Doc Stamp—Mort : 525.00
Intang. Tax : 320. 0G
SYD CROSBY, FLAGLER County
LOANNO, — 288155 By: 1, Dtpaneaad D.C. Time:16243:
[Space Above This Line For Recording Data]
MORTGAGE
THIS MORTGAGE (Seeuri Instrument’) Is givenon DECEMBER 20, 1995 . The mortgagoris
JOYCE M. PETERSON, ASTRUSTEE AND INDIVIDUALLY AND ROGER E. PETERSON, HER HUSBAND
of the Trust Agreement dated January 25, 1995.
{"Borrower").
This Security Instrument is given to Barnett Bank of Volusia County , care of Barnett
Mortgage Company, P.O. Box 40843, Jacksonville, Florida 32203-0843, which Is organized and existing under the laws of
the State of Florida ,and whose address is. 1825 Business Park Blvd, Daytona Beach, FL 32114
Borrower owes Lender the principal sum of One Hundred Fifty Thousand Dollars and no/100
Dollars (US.$ 150,000.00 ). This debtis
evidenced by Borrower's note dated the same date as this Security Instrument ("Note"), which provides for monthly payments, with the full debt, if
not paid earlier, due and payable on January 1, 2026 . This Security Instrument secures to Lender: (a) the repayment of the dabt
evidenced by the Note, with interest, and all renewals, extensions and modifications of the Note; (b) the payment of all other sums, with interest,
advanced under paragraph 7 to protect the security of this Security instrument; and (c) the performance of Borrower's covenants and agreements
under this Security Instrument and the Note. For this purpose, Borrower does hereby mortgage, grant and convey to Lender the following
descrived property locatedin Flagler County, Florida:
.

已更新：使用Inst No提取了pytesseract