feat: associate face results with person tracks

include/face/face_recog_debug.h

@@ -41,6 +41,7 @@ inline std::string BuildFaceRecogDebugSummaryLine(
         << " id=" << node_id
         << " frame=" << frame_id
         << " status=" << status
+        << " person_track_id=" << item.person_track_id
         << " candidate=" << candidate
         << " candidate_id=" << item.candidate_person_id
         << " best_sim=" << std::fixed << std::setprecision(2) << item.best_sim

include/face/face_result.h

@@ -31,6 +31,7 @@ struct FaceDetResult {
 
 struct FaceRecogItem {
     Rect bbox{};
+    int person_track_id = -1;
 
     int best_person_id = -1;
     std::string best_name;

plugins/ai_face_recog/ai_face_recog_node.cpp

@@ -44,6 +44,41 @@ bool ReadFileToString(const std::string& path, std::string& out) {
     return ifs.good();
 }
 
+inline float IntersectionArea(const Rect& a, const Rect& b) {
+    const float left = std::max(a.x, b.x);
+    const float top = std::max(a.y, b.y);
+    const float right = std::min(a.x + a.w, b.x + b.w);
+    const float bottom = std::min(a.y + a.h, b.y + b.h);
+    const float w = right - left;
+    const float h = bottom - top;
+    if (w <= 0.0f || h <= 0.0f) return 0.0f;
+    return w * h;
+}
+
+inline bool ContainsPoint(const Rect& r, float x, float y) {
+    return x >= r.x && y >= r.y && x <= (r.x + r.w) && y <= (r.y + r.h);
+}
+
+int AssociateFaceToPersonTrack(const Rect& face_bbox, const std::vector<Detection>& dets) {
+    const float center_x = face_bbox.x + face_bbox.w * 0.5f;
+    const float center_y = face_bbox.y + face_bbox.h * 0.5f;
+
+    int best_track_id = -1;
+    float best_overlap = -1.0f;
+    for (const auto& det : dets) {
+        if (det.track_id < 0) continue;
+        if (!ContainsPoint(det.bbox, center_x, center_y)) continue;
+
+        const float overlap = IntersectionArea(face_bbox, det.bbox);
+        if (overlap > best_overlap) {
+            best_overlap = overlap;
+            best_track_id = det.track_id;
+        }
+    }
+
+    return best_track_id;
+}
+
 struct GalleryEntry {
     int person_id = -1;
     std::string name;
@@ -958,6 +993,7 @@ private:
 
             FaceRecogItem item;
             item.bbox = face.bbox;
+            item.person_track_id = frame->det ? AssociateFaceToPersonTrack(face.bbox, frame->det->items) : -1;
             item.has_landmarks = face.has_landmarks;
             item.landmarks = face.landmarks;
 

tests/CMakeLists.txt

@@ -48,6 +48,7 @@ add_executable(rk3588_gtests
     test_alarm_behavior_events.cpp
     test_log_action.cpp
     test_face_recog_debug.cpp
+    test_face_track_association.cpp
     test_infer_backend.cpp
     test_image_processor.cpp
     test_codec_backend.cpp

tests/test_face_track_association.cpp

@@ -0,0 +1,59 @@
+#include <gtest/gtest.h>
+
+#include <string>
+#include <vector>
+
+#include "face/face_recog_debug.h"
+
+#include "../plugins/ai_face_recog/ai_face_recog_node.cpp"
+
+namespace rk3588 {
+namespace {
+
+Detection MakePersonDetection(int track_id, float x, float y, float w, float h) {
+    Detection det;
+    det.cls_id = 0;
+    det.track_id = track_id;
+    det.score = 0.95f;
+    det.bbox = Rect{x, y, w, h};
+    return det;
+}
+
+TEST(FaceTrackAssociationTest, PicksContainingTrackedPerson) {
+    const Rect face{50.0f, 50.0f, 20.0f, 20.0f};
+    const std::vector<Detection> dets = {
+        MakePersonDetection(101, 45.0f, 45.0f, 20.0f, 20.0f),
+        MakePersonDetection(202, 40.0f, 40.0f, 40.0f, 40.0f),
+        MakePersonDetection(303, 10.0f, 10.0f, 15.0f, 15.0f),
+    };
+
+    EXPECT_EQ(AssociateFaceToPersonTrack(face, dets), 202);
+}
+
+TEST(FaceTrackAssociationTest, RejectsWhenNoTrackedPersonMatches) {
+    const Rect face{50.0f, 50.0f, 20.0f, 20.0f};
+    const std::vector<Detection> dets = {
+        MakePersonDetection(-1, 45.0f, 45.0f, 20.0f, 20.0f),
+        MakePersonDetection(404, 200.0f, 200.0f, 20.0f, 20.0f),
+    };
+
+    EXPECT_EQ(AssociateFaceToPersonTrack(face, dets), -1);
+}
+
+TEST(FaceTrackAssociationTest, DebugLineIncludesPersonTrackId) {
+    FaceRecogItem item;
+    item.person_track_id = 88;
+    item.candidate_person_id = 7;
+    item.candidate_name = "alice";
+    item.best_sim = 0.89f;
+    item.second_sim = 0.74f;
+    item.unknown = false;
+
+    const std::string line = BuildFaceRecogDebugSummaryLine("face_recog", 42, item);
+
+    EXPECT_NE(line.find("person_track_id=88"), std::string::npos);
+    EXPECT_NE(line.find("candidate=alice"), std::string::npos);
+}
+
+}  // namespace
+}  // namespace rk3588