CAPTCHA Recognition

多引擎验证码识别：ddddocr 本地模型（OCR 文字识别 / 目标检测 / 滑块缺口匹配）与超级鹰打码平台（通用图片识别 / 报错返分 / 查询题分）。

/api/ddddocr/

Service Description

验证码识别服务按「线路」聚合多种识别引擎：当前接入 ddddocr（本地模型）与 超级鹰（第三方打码平台），在上方线路 Tab 之间切换即可查看各自的接口与调试面板。

ddddocr 线路（/api/ddddocr/）：纯本地模型，无需第三方账号、无额外费用。识别图片中的文字、检测图片内的目标位置，以及匹配滑块验证码的缺口位置（两种算法）。识别实例的字符范围可配置（预定义编号或自定义字符集），识别特定站点的验证码前先设置字符集通常能明显提升准确率。

超级鹰线路（/api/chaojiying/）：第三方人工 + 智能打码平台，通用图片识别，账号由本站统一配置、调用方无需自备。按题分计费，上传即扣费（无论最终识别对错）；官方标准价 1 元 = 1000 题分（充值有赠送，实际单价低于此）。调用前请用 codetype 精确匹配图片内容 —— 类型选得越准，识别越快越准，花费也越低。各识别类型的官方单价与折合人民币如下：

识别类型	说明	题分	折合人民币（约）
`1902`	4~6位英文数字	10/12/15	0.01~0.015 元
`1004`	1~4位英文数字	10	0.01 元
`1005`	1~5位英文数字	12	0.012 元
`1006`	1~6位英文数字	15	0.015 元
`1007`	1~7位英文数字	17.5	0.0175 元
`1008`	1~8位英文数字	20	0.02 元
`1009`	1~9位英文数字	22.5	0.0225 元
`1010`	1~10位英文数字	25	0.025 元
`1012`	1~12位英文数字	30	0.03 元
`1020`	1~20位英文数字	50	0.05 元
`2001`	1位纯汉字	10	0.01 元
`2002`	1~2位纯汉字	20	0.02 元
`2003`	1~3位纯汉字	30	0.03 元
`2004`	1~4位纯汉字	40	0.04 元
`2005`	1~5位纯汉字	50	0.05 元
`2006`	1~6位纯汉字	60	0.06 元
`2007`	1~7位纯汉字	70	0.07 元
`3004`	1~4位纯英文	10	0.01 元
`3005`	1~5位纯英文	12	0.012 元
`3006`	1~6位纯英文	15	0.015 元
`3007`	1~7位纯英文	17.5	0.0175 元
`3008`	1~8位纯英文	20	0.02 元
`3012`	1~12位纯英文	30	0.03 元
`4004`	1~4位纯数字	10	0.01 元
`4005`	1~5位纯数字	12	0.012 元
`4006`	1~6位纯数字	15	0.015 元
`4007`	1~7位纯数字	17.5	0.0175 元
`4008`	1~8位纯数字	20	0.02 元
`5000`	不定长汉字英文数字	2.5/英文, 10/汉字(基础10)	每英文 0.0025 元、每汉字 0.01 元（基础 0.01 元）
`5108`	8位英文数字(含特殊字符)	22	0.022 元
`5201`	拼音首字母 / 计算题 / 成语混合	首字母20, 计算20, 成语40	首字母 0.02 元 / 计算 0.02 元 / 成语 0.04 元
`6001`	计算题(返回计算结果)	15	0.015 元
`6004`	问答题 / 智能回答题	15	0.015 元
`6900`	中英数混合(配合8f指令取特定色彩)	基础10	每英文 0.0025 元、每汉字 0.01 元（基础 0.01 元）
`9602`	水平拼图的两个坐标	36	0.036 元
`9801`	按指定字符定位(返回指定字坐标)	9/定位, 最低28	每定位 0.009 元（最低 0.028 元）
`9800`	文字定位(返回所有字及坐标)	6/定位, 最低28	每定位 0.006 元（最低 0.028 元）
`9901`	定位图上有且仅有的一个图形块中心点坐标	16	0.016 元
`9902`	定位图上有且仅有的两个图形块中心点坐标	24	0.024 元
`9900`	定位各种滑块 / 缺口 / 色块	8/定位+8	每定位 0.008 元 + 基础 0.008 元
`9101`	按操作提示返回固定1个坐标	16	0.016 元
`9102`	按操作提示返回固定2个坐标	24	0.024 元
`9103`	按操作提示返回固定3个坐标	32	0.032 元
`9104`	按操作提示返回固定4个坐标	40	0.04 元
`9004`	按操作提示返回1~4个坐标	40	0.04 元
`9005`	按操作提示返回1~5个坐标	48	0.048 元
`9006`	按操作提示返回1~6个坐标	56	0.056 元
`9008`	按操作提示返回1~8个坐标	72	0.072 元
`9009`	按操作提示返回1~9个坐标	80	0.08 元

超级鹰结果校验：平台会随识别结果返回 md5，本站按官方算法 md5("软件ID,软件KEY,图片ID,图片结果") 逐条核验，校验不通过会以 4xxxx 拒绝返回，避免结果被中途篡改。

超级鹰报错返分：识别结果确实错误时，可凭识别接口返回的 pic_id 调「报错返分」退回题分。官方限制：须在拿到 pic_id 后 3 分钟内、部分类型不支持；仅允许在确认结果错误时调用，对正确结果报错会被评估信用。

This service requires project signature authentication. Accuracy depends on image quality and noise level, so plan for retries or a manual fallback on failures.

Response Format

Every endpoint returns the same JSON envelope; the payload is always inside data:

{
  "code": 10000,
  "msg": "成功",
  "data": { ... }
}

code: business status code. 10000 means success; anything else is an error — see the Error Codes page for the full list.
msg: a human-readable message you can show to end users, but never use it for program logic.
data: the payload; usually null on error. Each endpoint's Response section lists its data fields.

Judge success by code, not by the HTTP status code. View all error codes

ddddocr General Recognition

Python ddddocr (local model) Signature Required

Pure local model recognition with no third-party dependency callbacks. The first call loads the model, so a slower response is normal.

POST /api/ddddocr/ocr Total calls: 0

OCR Text Recognition

Recognize the text content in an image and return the recognized text; it can also return probability and filter by color.

Image File (Upload) image

Upload a local image (page debugging now supports real file upload)

or image URL/base64 image_url

or fill in the image URL / base64 (use when no local file is available)

Return Probability Distribution probability

When true, returns data={text, probability}; when false, data is the text

Fix Transparent PNG png_fix

When true, apply white-background repair to images with a transparent channel

Color Filter (JSON Array) colors

Optional: recognize only text of the specified colors, e.g. ["red","blue","green"]

Custom Color Ranges (JSON Object) custom_color_ranges

Optional: an HSV range dictionary, used together with colors

Beta Model beta

When true, use the Beta OCR model (more accurate for some CAPTCHAs)

This Service requires a project signature; when uploading a file, the request body is multipart/form-data.
Provide at least one of the file field or the URL/base64 field; when both are provided, the file takes precedence.

Response Common envelope

data varies with the parameters: when probability=false it is the recognized text string; when probability=true it is an object {text, probability}, where probability is ddddocr's raw probability distribution.

Example response

"8k3m"

POST /api/ddddocr/set-ranges Total calls: 0

Set OCR Character Range

Set the character range of this recognition instance (a predefined number or a custom character set).

Character Range ranges *

Required: predefined 0=digits/1=lowercase/2=uppercase/3=upper+lowercase/4=lowercase+digits/5=uppercase+digits/6=upper+lowercase+digits/7=default; or a custom string, e.g. "0123456789"

Beta Model beta

Optional: used together with the Beta model

Note: the backend creates a new recognition instance on every call; this Endpoint only applies to the instance of "that call" and does not change the character set of subsequent /ocr requests.

Response Common envelope

This endpoint returns no business data; data is null on success — judge the result from code and msg.

POST /api/ddddocr/detect Total calls: 0

Object Detection

Detect target positions in an image and return the bounding box and confidence of each target.

Image File (Upload) image

Upload a local image (page debugging now supports real file upload)

or image URL/base64 image_url

or fill in the image URL / base64 (use when no local file is available)

Provide at least one of the file field or the URL/base64 field; when both are provided, the file takes precedence.

Response Common envelope

data is an array of detection results; each element is [x1, y1, x2, y2], the top-left and bottom-right coordinates (in pixels) of a target's bounding box.

Example response

[[24, 60, 102, 138]]

POST /api/ddddocr/slide-match Total calls: 0

Slider Matching (Edge Matching Method)

Match the gap position of the slider thumbnail within the background image via edge detection (algorithm 1).

Slider Image (Upload) slide_image

Upload a local image (page debugging now supports real file upload)

or slider image URL/base64 slide_image_url

or fill in the image URL / base64 (use when no local file is available)

Background Image (Upload) bg_image

Upload a local image (page debugging now supports real file upload)

or background image URL/base64 bg_image_url

or fill in the image URL / base64 (use when no local file is available)

Simple Target Mode simple_target

Optional: Default false

Provide at least one from each of the slider (slide_image / slide_image_url) and background (bg_image / bg_image_url); when both are provided, the file takes precedence.

Response Common envelope

Field	Type	Description
target_x	int	X coordinate of the gap's top-left corner
target_y	int	Y coordinate of the gap's top-left corner

Example response

{"target_x": 128, "target_y": 45}

POST /api/ddddocr/slide-comparison Total calls: 0

Slider Matching (Difference Comparison Method)

Locate the gap by directly comparing the difference between the slider image and the background image (algorithm 2).

Slider Image (Upload) slide_image

Upload a local image (page debugging now supports real file upload)

or slider image URL/base64 slide_image_url

or fill in the image URL / base64 (use when no local file is available)

Background Image (Upload) bg_image

Upload a local image (page debugging now supports real file upload)

or background image URL/base64 bg_image_url

or fill in the image URL / base64 (use when no local file is available)

Provide at least one from each of the slider (slide_image / slide_image_url) and background (bg_image / bg_image_url); when both are provided, the file takes precedence.

Response Common envelope

Field	Type	Description
target_x	int	X coordinate of the gap's top-left corner
target_y	int	Y coordinate of the gap's top-left corner

Example response

{"target_x": 128, "target_y": 45}

超级鹰打码平台

超级鹰（chaojiying.com） Signature Required

第三方打码平台，按题分计费（上传即扣费）；账号由本站统一配置，调用方无需自备。各识别类型的题分单价与折合人民币见本页「服务说明」。

POST /api/chaojiying/ocr Total calls: 0

图片识别

上传验证码图片，返回识别结果与图片标识号（成功即扣题分）。

Image File (Upload) file

上传本地图片（不超过 2MB，推荐 bmp / jpg / png）

或图片 base64 / 图片地址 image

或填图片 base64 字符串、http(s) 图片地址（与 file 二选一，file 优先）

识别类型(codetype) codetype *

必填：按图片内容精确匹配类型（如 1902 = 4~6 位英文数字），可提高识别速度与正确率。各类型的题分与折合人民币见本页「服务说明」的价格表。

附加信息(str_debug) str_debug

选填：定位类类型（如 9801）的附加指令，普通类型留空

图片来源三选一：上传 file 文件、填 image(base64)、填 image(http 图片地址)；图片不超过 2MB，推荐 bmp / jpg / png，分辨率比例要正常。
Platform business errors (e.g. wrong account/password, insufficient credits) return 4xxxx, with the platform's Chinese description in msg.
识别结果的大小写以平台为准（英文数字类通常返回小写），比对时请自行做大小写归一。
各类型的题分单价与折合人民币见本页「服务说明」中的价格表。

Response Common envelope

Field	Type	Description
pic_id	string	Image ID, used when reporting an error for a refund
pic_str	string	Recognized text
md5	string	Result checksum

Example response

{"pic_id": "9160109360600112681", "pic_str": "8k3m", "md5": "0f3d2c1b4a5968778695a4b3c2d1e0f1"}

POST /api/chaojiying/report-error Total calls: 0

报错返分

识别结果确实错误时退回题分（须在拿到 pic_id 后 3 分钟内调用）。

图片标识号 pic_id *

必填：识别接口返回的 pic_id

平台限制：仅限识别结果确实错误时调用，对正确结果报错会被评估信用；部分识别类型不支持返分；超过 3 分钟失效。

Response Common envelope

This endpoint returns no business data; data is an empty object {} on success — judge the result from code and msg.

GET /api/chaojiying/score Total calls: 0

查询题分余额

查询本站平台账号的题分余额（用于确认可用额度）。

At the official standard rate, 1000 credits ≈ 1 CNY.

Response Common envelope

Field	Type	Description
tifen	int	Credit balance (tifen)
tifen_lock	int	Locked credits (tifen)

Example response

{"tifen": 1000, "tifen_lock": 0}