From 83a07290b1aaa5840821cd236392e44184400239 Mon Sep 17 00:00:00 2001
From: tian <11429339@qq.com>
Date: Wed, 15 Apr 2026 14:49:30 +0800
Subject: [PATCH] fix: restrict face track association to persons

---
 configs/full_pipeline_1080p_test_alarm.json  |  8 ++++----
 plugins/ai_face_recog/ai_face_recog_node.cpp |  7 +++++++
 tests/test_face_track_association.cpp        | 17 +++++++++++++++--
 3 files changed, 26 insertions(+), 6 deletions(-)

diff --git a/configs/full_pipeline_1080p_test_alarm.json b/configs/full_pipeline_1080p_test_alarm.json
index 9f0d32a..f4d832c 100644
--- a/configs/full_pipeline_1080p_test_alarm.json
+++ b/configs/full_pipeline_1080p_test_alarm.json
@@ -442,10 +442,6 @@
         ],
         [
           "face_det",
-          "face_recog"
-        ],
-        [
-          "face_recog",
           "person_det"
         ],
         [
@@ -454,6 +450,10 @@
         ],
         [
           "person_trk",
+          "face_recog"
+        ],
+        [
+          "face_recog",
           "shoe_det",
           {
             "queue": {
diff --git a/plugins/ai_face_recog/ai_face_recog_node.cpp b/plugins/ai_face_recog/ai_face_recog_node.cpp
index c3eea98..757061b 100644
--- a/plugins/ai_face_recog/ai_face_recog_node.cpp
+++ b/plugins/ai_face_recog/ai_face_recog_node.cpp
@@ -44,6 +44,8 @@ bool ReadFileToString(const std::string& path, std::string& out) {
     return ifs.good();
 }
 
+constexpr int kPersonClassId = 0;
+
 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);
@@ -59,6 +61,10 @@ 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);
 }
 
+inline bool IsPersonDetection(const Detection& det) {
+    return det.cls_id == kPersonClassId;
+}
+
 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;
@@ -66,6 +72,7 @@ int AssociateFaceToPersonTrack(const Rect& face_bbox, const std::vector<Detectio
     int best_track_id = -1;
     float best_overlap = -1.0f;
     for (const auto& det : dets) {
+        if (!IsPersonDetection(det)) continue;
         if (det.track_id < 0) continue;
         if (!ContainsPoint(det.bbox, center_x, center_y)) continue;
 
diff --git a/tests/test_face_track_association.cpp b/tests/test_face_track_association.cpp
index ffe0b8c..14ad681 100644
--- a/tests/test_face_track_association.cpp
+++ b/tests/test_face_track_association.cpp
@@ -10,15 +10,19 @@
 namespace rk3588 {
 namespace {
 
-Detection MakePersonDetection(int track_id, float x, float y, float w, float h) {
+Detection MakeDetection(int cls_id, int track_id, float x, float y, float w, float h) {
     Detection det;
-    det.cls_id = 0;
+    det.cls_id = cls_id;
     det.track_id = track_id;
     det.score = 0.95f;
     det.bbox = Rect{x, y, w, h};
     return det;
 }
 
+Detection MakePersonDetection(int track_id, float x, float y, float w, float h) {
+    return MakeDetection(0, track_id, x, y, w, h);
+}
+
 TEST(FaceTrackAssociationTest, PicksContainingTrackedPerson) {
     const Rect face{50.0f, 50.0f, 20.0f, 20.0f};
     const std::vector<Detection> dets = {
@@ -40,6 +44,15 @@ TEST(FaceTrackAssociationTest, RejectsWhenNoTrackedPersonMatches) {
     EXPECT_EQ(AssociateFaceToPersonTrack(face, dets), -1);
 }
 
+TEST(FaceTrackAssociationTest, RejectsNonPersonDetections) {
+    const Rect face{50.0f, 50.0f, 20.0f, 20.0f};
+    const std::vector<Detection> dets = {
+        MakeDetection(1, 909, 40.0f, 40.0f, 40.0f, 40.0f),
+    };
+
+    EXPECT_EQ(AssociateFaceToPersonTrack(face, dets), -1);
+}
+
 TEST(FaceTrackAssociationTest, DebugLineIncludesPersonTrackId) {
     FaceRecogItem item;
     item.person_track_id = 88;