feat: 实验性增加 OCR 字符等价类支持

src/MaaCore/Vision/OcrImageAnalyzer.cpp

@@ -15,6 +15,11 @@ bool asst::OcrImageAnalyzer::analyze()
 
     std::vector<TextRectProc> preds_vec;
 
+    preds_vec.emplace_back([](TextRect& tr) -> bool {
+        tr.text = equivalent_class_preprocess(tr.text);
+        return true;
+    });
+
     if (!m_replace.empty()) {
         TextRectProc text_replace = [&](TextRect& tr) -> bool {
             for (const auto& [regex, new_str] : m_replace) {
@@ -93,19 +98,25 @@ void asst::OcrImageAnalyzer::set_use_cache(bool is_use) noexcept
 
 void asst::OcrImageAnalyzer::set_required(std::vector<std::string> required) noexcept
 {
+    ranges::for_each(required, [](std::string& str) { str = equivalent_class_preprocess(str); });
     m_required = std::move(required);
 }
 
-void asst::OcrImageAnalyzer::set_replace(std::unordered_map<std::string, std::string> replace) noexcept
+void asst::OcrImageAnalyzer::set_replace(const std::unordered_map<std::string, std::string>& replace) noexcept
 {
-    m_replace = std::move(replace);
+    m_replace = {};
+    for (auto&& [key, val] : replace) {
+        auto new_key = equivalent_class_preprocess(key);
+        auto new_val = equivalent_class_preprocess(val);
+        m_replace[new_key] = new_val;
+    }
 }
 
 void asst::OcrImageAnalyzer::set_task_info(OcrTaskInfo task_info) noexcept
 {
-    m_required = std::move(task_info.text);
+    set_required(std::move(task_info.text));
     m_full_match = task_info.full_match;
-    m_replace = std::move(task_info.replace_map);
+    set_replace(task_info.replace_map);
     m_use_cache = task_info.cache;
     m_use_char_model = task_info.is_ascii;
 
@@ -124,6 +135,44 @@ std::vector<asst::TextRect>& asst::OcrImageAnalyzer::get_result() noexcept
     return m_ocr_result;
 }
 
+std::string asst::OcrImageAnalyzer::equivalent_class_preprocess(const std::string& in)
+{
+    using equivalent_class = std::vector<std::string>;
+    static const std::vector<equivalent_class> classes {
+            // hiragana
+            { "あ", "ぁ" },
+            { "い", "ぃ" },
+            { "う", "ぅ" },
+            { "え", "ぇ" },
+            { "お", "ぉ" },
+            { "つ", "っ" },
+            { "や", "ゃ" },
+            { "ゆ", "ゅ" },
+            { "よ", "ょ" },
+            { "わ", "ゎ" },
+            // katakana
+            { "ア", "ァ" },
+            { "イ", "ィ" },
+            { "ウ", "ゥ" },
+            { "エ", "ェ" },
+            { "オ", "ォ" },
+            { "ツ", "ッ" },
+            { "ヤ", "ャ" },
+            { "ユ", "ュ" },
+            { "ヨ", "ョ" },
+            { "ワ", "ヮ" },
+    }; // TODO: store this into a file
+
+    auto result = in;
+    for (const auto& cls : classes) {
+        // replace the elements in group into the first one
+        ranges::for_each(cls.begin() + 1, cls.end(),
+                      [&](const std::string& elem) { utils::string_replace_all_in_place(result, elem, cls.front()); });
+    }
+
+    return result;
+}
+
 void asst::OcrImageAnalyzer::set_task_info(std::shared_ptr<TaskInfo> task_ptr)
 {
     set_task_info(*std::dynamic_pointer_cast<OcrTaskInfo>(task_ptr));

src/MaaCore/Vision/OcrImageAnalyzer.h

@@ -25,7 +25,7 @@ namespace asst
         virtual void sort_result_by_required(); // 按传入的需求数组排序，传入的在前面结果接在前面
 
         void set_required(std::vector<std::string> required) noexcept;
-        void set_replace(std::unordered_map<std::string, std::string> replace) noexcept;
+        void set_replace(const std::unordered_map<std::string, std::string>& replace) noexcept;
 
         virtual void set_task_info(std::shared_ptr<TaskInfo> task_ptr);
         virtual void set_task_info(const std::string& task_name);
@@ -38,6 +38,9 @@ namespace asst
         virtual const std::vector<TextRect>& get_result() const noexcept;
         virtual std::vector<TextRect>& get_result() noexcept;
 
+        // normalize similar characters into the same one
+        static std::string equivalent_class_preprocess(const std::string& in);
+
     protected:
         virtual void set_task_info(OcrTaskInfo task_info) noexcept;