当前位置：首页 > backend >正文

Python文字识别OCR

backend 2025/9/1 23:36:54

一.引言

文字识别，也称为光学字符识别（Optical Character Recognition, OCR），是一种将不同形式的文档（如扫描的纸质文档、PDF文件或数字相机拍摄的图片）中的文字转换成可编辑和可搜索的数据的技术。随着技术的发展，文字识别技术已经成为信息管理、自动化办公和智能系统的关键组成部分。

二.简介

为了易于集成和使用，我们将文字识别OCR封装为DLL（动态链接库）。这种封装方式不仅保留了算法的性能优势，还提供了跨平台和跨语言的兼容性，目前支持编程语言如下：

C++
Python
易语言

1.C++头文件

 #ifndef __SN_OCR__H__
#define __SN_OCR__H__#include "windows.h"//返回参数
typedef struct SN_STATU {int code;			//错误码，如果为 0 表示成功，否则表示错误号char message[4096];	//错误信息，如果为 "OK" 表示成功，否则返回错误信息}SN_STATU;/*启动OCR文字识别服务
*
* 参数：
*	[in]  szOnnxFilePath：	设置 onnx 模型文件路径，如果设置为 NULL，默认和 DLL文件同级目录
* 	[out] pResult：			返回错误信息，参数pResult->code(错误码)如果为 0 表示成功，否则表示错误号;
*
* 返回值：成功返回0，失败返回错误号，详细错误信息请参考 pResult
*
*/
int WINAPI apiSNInitOCRServer(char* szOnnxFilePath, SN_STATU* pStatu);/*创建OCR文字识别句柄
*
* 参数：
*	[in]  szKey：		卡密(购买卡密：https://shop.4yuns.com/links/7C9F16B7)
* 	[in]  pOnnxFilePath：设置 onnx 模型文件路径，如果设置为 NULL，默认和 DLL文件同级目录
* 	[out] pResult：		返回错误信息，参数pResult->code(错误码)如果为 0 表示成功，否则表示错误号;
*
* 返回值：成功返回句柄，失败返回NULL
*
*/
HANDLE WINAPI apiSNCreateOCRHandle(char* szKey, char* szOnnxFilePath, SN_STATU* pStatu);/*获取OCR文字识别卡密到期时间
*
* 参数：
*	[in]  handle：		句柄（通过调用apiSNCreateOCRHandle得到）
* 	[out] pResult：		返回错误信息，参数pResult->code(错误码)如果为 0 表示成功，否则表示错误号;
*
* 返回值：返回卡密到期时间，失败返回NULL，错误信息请查看参数 pResult->message
*
*/
char* WINAPI apiSNGetKeyExpiresTime(HANDLE handle, SN_STATU* pResult);/*获取OCR文字识别结果（以json字符串形式返回）
*
* 参数：
*	[in]  handle：			句柄（通过调用apiSNCreateOCRHandle得到）
*  	[in]  szImageFilePath：	图片路径
* 	[out] pResult：			返回错误信息，参数pResult->code(错误码)如果为 0 表示成功，否则表示错误号;
*
* 返回值：返回OCR文字识别结果（以json字符串形式返回），失败返回NULL，错误信息请查看参数 pResult->message
*
*/
char* WINAPI apiSNGetOCRFromImage(HANDLE handle, char* szImageFilePath, SN_STATU* pStatu);/*释放OCR文字识别句柄（释放内存）
*
* 参数：
*	[in] handle：		句柄（通过调用apiSNCreateOCRHandle得到）
*
* 返回值：返回 0 表示成功，其他值表示错误号;
*
*/
int WINAPI apiSNDestroyOCRHandle(HANDLE handle);#endif

2.Python调用dll接口

from ctypes import cdll, c_char_p, Structure, byref
import ctypes# 定义SN_STATU结构体
class SN_STATU(Structure):_fields_ = [("code", ctypes.c_int),("message", c_char_p * 4096)]# 加载DLL
lib = cdll.LoadLibrary('D://SNOCR.dll')# 设置函数参数类型
lib.apiSNInitOCRServer.argtypes = [c_char_p, ctypes.POINTER(SN_STATU)]
lib.apiSNInitOCRServer.restype = ctypes.c_intlib.apiSNCreateOCRHandle.argtypes = [c_char_p, c_char_p, ctypes.POINTER(SN_STATU)]
lib.apiSNCreateOCRHandle.restype = ctypes.c_void_plib.apiSNGetKeyExpiresTime.argtypes = [ctypes.c_void_p, ctypes.POINTER(SN_STATU)]
lib.apiSNGetKeyExpiresTime.restype = c_char_plib.apiSNGetOCRFromImage.argtypes = [ctypes.c_void_p, c_char_p, ctypes.POINTER(SN_STATU)]
lib.apiSNGetOCRFromImage.restype = c_char_plib.apiSNDestroyOCRHandle.argtypes = [ctypes.c_void_p]
lib.apiSNDestroyOCRHandle.restype = ctypes.c_int# 初始化变量
statu = SN_STATU()
key = b"SNKJe9xffLhdFY7r3TcffXq44ThDVcE3BQFQFfVA9VG4"
onnx_path = b"D://SNOCR.onnx"
image_path = b"D://7.jpg"# 1. 启动OCR服务
ret = lib.apiSNInitOCRServer(onnx_path, byref(statu))
if ret < 0:print(f"Error:{statu.message.decode('utf-8')}")exit()# 2. 创建OCR句柄
handle = lib.apiSNCreateOCRHandle(key, onnx_path, byref(statu))
if not handle:print(f"Error:{statu.message.decode('utf-8')}")exit()# 3. 获取卡密到期时间
expires_time = lib.apiSNGetKeyExpiresTime(handle, byref(statu))
if not expires_time:print(f"Error:{statu.message.decode('utf-8')}")exit()
print(f"Expires Time: {expires_time.decode('utf-8')}")# 4. 识别OCR，返回Json字符串
ocr_result = lib.apiSNGetOCRFromImage(handle, image_path, byref(statu))
if not ocr_result:print(f"Error:{statu.message.decode('utf-8')}")exit()
try:print(f"OCR Result: {ocr_result.decode('utf-8')}")
except UnicodeDecodeError:print(f"OCR Result: {ocr_result.decode('GBK')}")# 5. 释放内存
lib.apiSNDestroyOCRHandle(handle)# 等待输入，防止程序直接退出
input("Press Enter to exit...")

三.效果演示

1.图片1

识别效果：

{"type":	0,"task_id":	1,"err_code":	0,"ocr_result":	{"single_result":	[{"single_rate":	0.939104,"left":	102.208336,"top":	41.812500,"right":	329.854156,"bottom":	67.829170,"single_str_utf8":	"中国建设银行"}, {"single_rate":	0.966887,"left":	104.431534,"top":	68.423492,"right":	309.992828,"bottom":	84.602386,"single_str_utf8":	"China Construction Bank"}, {"single_rate":	0.968900,"left":	102.672920,"top":	96.168755,"right":	403.258331,"bottom":	111.964584,"single_str_utf8":	"龙卡通(储蓄卡)LONG CARD(DEBIT CARD)"}, {"single_rate":	0.975151,"left":	41.781921,"top":	137.955643,"right":	410.251556,"bottom":	164.107880,"single_str_utf8":	"6227 0033 2069 0222 205"}, {"single_rate":	0.935433,"left":	20.770407,"top":	210.668716,"right":	77.230583,"bottom":	230.122101,"single_str_utf8":	"ATM"}, {"single_rate":	0.960131,"left":	103.137505,"top":	185.368759,"right":	192.337509,"bottom":	207.204163,"single_str_utf8":	"CCB GZ"}, {"single_rate":	0.929293,"left":	338.376495,"top":	201.118103,"right":	417.111450,"bottom":	224.273529,"single_str_utf8":	"UnionPa"}, {"single_rate":	0.917808,"left":	367.485413,"top":	220.677078,"right":	413.479156,"bottom":	239.260422,"single_str_utf8":	"银联"}],"unknown_1":	446,"unknown_2":	280}
}

2.图片2

识别效果：

{"type":	0,"task_id":	1,"err_code":	0,"ocr_result":	{"single_result":	[{"single_rate":	0.919637,"left":	622.061157,"top":	123.251556,"right":	1046.638920,"bottom":	190.015121,"single_str_utf8":	"马托13610000670"}, {"single_rate":	0.996936,"left":	40.618664,"top":	324.310150,"right":	541.513184,"bottom":	371.843231,"single_str_utf8":	"广州利驰服装有限公司"}, {"single_rate":	0.997014,"left":	624.066650,"top":	218.300000,"right":	1040.933350,"bottom":	242.966675,"single_str_utf8":	"地址:广州市海珠区赤岗路173号"}, {"single_rate":	0.964238,"left":	624.066650,"top":	251.600000,"right":	844.833374,"bottom":	276.266663,"single_str_utf8":	"金丰大厦503室"}, {"single_rate":	0.979285,"left":	625.300000,"top":	286.133331,"right":	1044.633300,"bottom":	309.566681,"single_str_utf8":	"厂址:广州市海珠区赤岗西路232"}, {"single_rate":	0.996300,"left":	624.066650,"top":	318.200000,"right":	759.733337,"bottom":	341.633331,"single_str_utf8":	"号-234号"}, {"single_rate":	0.991057,"left":	624.066650,"top":	351.500000,"right":	925,"bottom":	374.933350,"single_str_utf8":	"热线:400-688-7260"}, {"single_rate":	0.964125,"left":	625.239319,"top":	381.016510,"right":	922.600220,"bottom":	404.499695,"single_str_utf8":	"电话:020-84022958"}, {"single_rate":	0.993601,"left":	624.066650,"top":	408.233337,"right":	923.766663,"bottom":	432.900000,"single_str_utf8":	"传真:020-84022572"}, {"single_rate":	0.950434,"left":	625.203430,"top":	438.767609,"right":	1107.616580,"bottom":	464.666626,"single_str_utf8":	"邮箱:kunhemwl@yahoo.com.cn"}, {"single_rate":	0.962023,"left":	624.066650,"top":	471.133331,"right":	1001.466670,"bottom":	494.566681,"single_str_utf8":	"网址:www.hxkunhe.com"}],"unknown_1":	1184,"unknown_2":	614}
}