configure ocr errors

tmarenko · May 16, 2023 · 4aac275 · 4aac275
1 parent bab7a40
commit 4aac275
Show file tree

Hide file tree

Showing 5 changed files with 16 additions and 8 deletions.
diff --git a/settings.cfg b/settings.cfg
@@ -1,6 +1,8 @@
 [config]
 # available: chi_sim, chi_tra, eng, kor, jpn, spa, fra, rus, tha, vie, deu, ind, por
 language=eng
+# how many wrong letters is allowed for Paimon's name. lesser value == more strict text recognition; default = 1 letter
+ocr_max_errors=1
 
 [genshin_window_name]
 genshin_chi_sim=原神

diff --git a/source/config.cpp b/source/config.cpp
@@ -3,9 +3,10 @@
 
 std::map<std::string, std::string> ParseConfig(const std::string &fileName) {
     std::map<std::string, std::string> configMap = {
-            {"language",    "eng"},
-            {"genshin_eng", "Genshin Impact"},
-            {"paimon_eng",  "Paimon"}
+            {"language",       "eng"},
+            {"genshin_eng",    "Genshin Impact"},
+            {"paimon_eng",     "Paimon"},
+            {"ocr_max_errors", "1"}
     };
     std::ifstream file(fileName);
     if (file.is_open()) {

diff --git a/source/main.cpp b/source/main.cpp
@@ -14,6 +14,7 @@
 typedef struct GenshinWindowInfo {
     std::wstring windowName = L"Genshin Impact";
     std::wstring windowClass = L"UnityWndClass";
+    int maxOcrErrors = 1;
     HWND hwnd = nullptr;
     int width = 0;
     int height = 0;
@@ -144,7 +145,8 @@ bool IsPaimonSpeaking(const std::string &paimonName) {
                                 (int) (OUT_DIALOGUE_NAME_POS.val[3] * frame.rows) - (int) (OUT_DIALOGUE_NAME_POS.val[1] * frame.rows));
     std::string defaultDialogue = GetTextFromImageByRect(frame, cropDefault);
     std::string outDialogue = GetTextFromImageByRect(frame, cropOut);
-    return IsStringsSimilar(defaultDialogue, paimonName) || IsStringsSimilar(outDialogue, paimonName);
+    return IsStringsSimilar(defaultDialogue, paimonName, gwi.maxOcrErrors) ||
+           IsStringsSimilar(outDialogue, paimonName, gwi.maxOcrErrors);
 }
 
 
@@ -223,6 +225,7 @@ int PaimonShutUp() {
     if (InitTesseract(nullptr, configMap["language"].c_str()))
         return 1;
 
+    gwi.maxOcrErrors = std::stoi(configMap["ocr_max_errors"]);
     std::string paimonName = configMap["paimon_" + configMap["language"]];
     std::cout << "Waiting for GenshinImpact.exe process." << std::endl;
     bool paimonWasHere = false;

diff --git a/source/tesseract.cpp b/source/tesseract.cpp
@@ -77,11 +77,13 @@ unsigned int LevenshteinDistance(const std::string &s1, const std::string &s2) {
 }
 
 
-bool IsStringsSimilar(std::string s1, std::string s2, double overlap) {
+bool IsStringsSimilar(std::string s1, std::string s2, int maxDifference) {
+    if (s1.length() == 0) {
+        return false;
+    }
     std::transform(s1.begin(), s1.end(), s1.begin(), ::toupper);
     std::transform(s2.begin(), s2.end(), s2.begin(), ::toupper);
-    double nonSimilarity = s1.length() > 0 ? (double) LevenshteinDistance(s1, s2) / s1.length() : 1;
-    return nonSimilarity <= overlap;
+    return LevenshteinDistance(s1, s2) <= maxDifference;
 }
 
 

diff --git a/source/tesseract.h b/source/tesseract.h
@@ -24,7 +24,7 @@ std::string GetTextFromImage(const cv::Mat &image);
 
 unsigned int LevenshteinDistance(const std::string &s1, const std::string &s2);
 
-bool IsStringsSimilar(std::string s1, std::string s2, double overlap = 0.4);
+bool IsStringsSimilar(std::string s1, std::string s2, int maxDifference);
 
 void DownloadTessdataFileIfNecessary(const std::string &language);