ctbjrj/成套报价软件/扒图/UmiOcrClient.cs

// ============================================================
//  UmiOcrClient:Umi-OCR 本地 OCR 引擎(hiroi-sora/Umi-OCR,离线免费)
//  · 前提:用户在本机运行 Umi-OCR 主程序,全局设置里开启 HTTP 服务(默认开启,端口 1224)
//  · POST {BaseUrl}/api/ocr  application/json:
//      { "base64": "<纯base64无前缀>", "options": { "data.format": "dict" } }
//  · 响应: code=100 成功;data=[{ text, score, box:[[x,y]×4 左上/右上/右下/左下], end }]
//          code=101 无文本(data 为原因字符串);其余 code 失败
//  · 地址存 baidu_ocr.json 的 umi_url(OCR设置窗口写,默认 http://127.0.0.1:1224)
//  · 本地引擎无需 API Key,IsConfigured 恒 true
// ============================================================
using System;
using System.Collections.Generic;
using System.Drawing;
using System.Drawing.Imaging;
using System.IO;
using System.Net;
using System.Text;
using System.Threading.Tasks;
using System.Web.Script.Serialization;

namespace 成套报价软件
{
    /// <summary>Umi-OCR 本地引擎客户端(单实例)。</summary>
    internal sealed class UmiOcrClient
    {
        private readonly JavaScriptSerializer _json = new JavaScriptSerializer();
        private string _baseUrl = "http://127.0.0.1:1224";

        /// <summary>状态提示,接 OcrClient.StatusChanged。</summary>
        public event Action<string> StatusChanged;

        public bool IsConfigured
        {
            get { return true; }   // 本地引擎无需 Key
        }

        /// <summary>服务地址由 OcrClient 从 baidu_ocr.json 的 umi_url 注入(热更新)。</summary>
        public void SetBaseUrl(string url)
        {
            url = (url == null ? "" : url.Trim());
            if (url.Length == 0) url = "http://127.0.0.1:1224";
            if (!url.StartsWith("http")) url = "http://" + url;
            _baseUrl = url.TrimEnd('/');
        }

        /// <summary>识别一块位图 → 文本行列表(带位置)。JSON POST 到本机 Umi-OCR。</summary>
        public Task<List<OcrItem>> RecognizeAsync(Bitmap crop, int timeoutMs)
        {
            if (crop == null || crop.Width <= 0 || crop.Height <= 0)
                return Task.FromResult(new List<OcrItem>());

            return Task.Factory.StartNew(() =>
            {
                var sw = System.Diagnostics.Stopwatch.StartNew();
                var result = RequestAndParse(crop, timeoutMs);
                Notify("Umi-OCR·" + result.Count + "条 " + sw.ElapsedMilliseconds + "ms");
                return result;
            });
        }

        private List<OcrItem> RequestAndParse(Bitmap crop, int timeoutMs)
        {
            // PNG base64(无 data: 前缀);PNG 保坐标精度,本地服务无 4MB 限制
            string b64;
            using (var ms = new MemoryStream())
            {
                crop.Save(ms, ImageFormat.Png);
                b64 = Convert.ToBase64String(ms.ToArray());
            }

            string req = _json.Serialize(new Dictionary<string, object>
            {
                { "base64", b64 },
                // ★ ocr.cls=true:纠正文本方向(PaddleOCR 引擎方向分类,即截图OCR界面的"纠正文本方向";
                //   界面设置不会带给 HTTP 调用,必须显式传——否则旋转90°的横排文字识别必乱)
                { "options", new Dictionary<string, object> { { "data.format", "dict" }, { "ocr.cls", true } } }
            });

            string body;
            using (var wc = new WebClient())
            {
                wc.Encoding = Encoding.UTF8;
                wc.Headers["Content-Type"] = "application/json";
                wc.Proxy = null;   // 本地地址不走系统代理(配了代理会把 127.0.0.1 也代理走)
                body = wc.UploadString(_baseUrl + "/api/ocr", "POST", req);
            }

            var resp = _json.Deserialize<Dictionary<string, object>>(body);
            if (resp == null || !resp.ContainsKey("code"))
                throw new Exception("Umi-OCR 返回解析失败");

            int code;
            int.TryParse(Convert.ToString(resp["code"]), out code);
            if (code == 101)
                return new List<OcrItem>();   // 无文本,不算错误
            if (code != 100)
                throw new Exception("Umi-OCR 错误 code=" + code + ": " +
                    Convert.ToString(resp.ContainsKey("data") ? resp["data"] : ""));

            var items = new List<OcrItem>();
            var arr = resp.ContainsKey("data") ? resp["data"] as System.Collections.ArrayList : null;
            if (arr != null)
            {
                foreach (Dictionary<string, object> w in arr)
                {
                    if (w == null) continue;
                    var boxArr = w.ContainsKey("box") ? w["box"] as System.Collections.ArrayList : null;
                    if (boxArr == null || boxArr.Count < 4) continue;
                    var pts = new PointF[4];
                    for (int i = 0; i < 4; i++)
                    {
                        var xy = boxArr[i] as System.Collections.ArrayList;
                        float x, y;
                        float.TryParse(Convert.ToString(xy != null && xy.Count > 0 ? xy[0] : "0"), out x);
                        float.TryParse(Convert.ToString(xy != null && xy.Count > 1 ? xy[1] : "0"), out y);
                        pts[i] = new PointF(x, y);
                    }
                    float score = 0.95f;
                    if (w.ContainsKey("score"))
                    {
                        float sv;
                        if (float.TryParse(Convert.ToString(w["score"]),
                            System.Globalization.NumberStyles.Float,
                            System.Globalization.CultureInfo.InvariantCulture, out sv)) score = sv;
                    }
                    items.Add(new OcrItem
                    {
                        Text = Convert.ToString(w.ContainsKey("text") ? w["text"] : ""),
                        Box = pts,          // box 即发送图像素坐标(顺时针四点),与 OcrItem 语义一致
                        Score = score
                    });
                }
            }
            return items;
        }

        private void Notify(string msg)
        {
            var h = StatusChanged;
            if (h != null) h(msg);
        }
    }
}