CogletESP-camera-version/main/face_tracker.cc

// [T05/T06] 人脸追踪任务
// 只有 ESP32-S3 + CONFIG_XIAOZHI_ENABLE_FACE_TRACKING=y 才编译完整实现
// 其他情况编译 3 个空函数，保证链接通过

#include "face_tracker.h"
#include "sdkconfig.h"

#if defined(CONFIG_XIAOZHI_ENABLE_FACE_TRACKING) && defined(CONFIG_IDF_TARGET_ESP32S3)

#include "human_face_detect.hpp"
#include "dl_image_define.hpp"
#include "dl_detect_define.hpp"
#include "board.h"
#include "esp32_camera.h"

#include <esp_heap_caps.h>
#include <esp_log.h>
#include <esp_timer.h>
#include <freertos/FreeRTOS.h>
#include <freertos/task.h>
#include <list>
#include <new>

static const char* TAG = "FaceTracker";
static TaskHandle_t s_handle = nullptr;
static volatile bool s_stop = false;
static float s_last_fps = 0.0f;

// T06: uart_send_face 由 T07 在 uart_component.{h,cc} 中提供
// 此处用前向声明 + 弱符号，让 T07 完成前 face_tracker.cc 仍能通过编译
// T07 完成后该弱符号被真实实现覆盖，无需改动本文件
extern "C" __attribute__((weak)) void uart_send_face(int x_offset, int y_offset);

static void face_tracker_task(void* arg) {
    (void)arg;
    // 等待摄像头 ISP 预热 + 视频流启动稳定
    vTaskDelay(pdMS_TO_TICKS(500));

    ESP_LOGI(TAG, "face_tracker task started on core %d", xPortGetCoreID());

    // 构造检测器：默认 model_type 由 CONFIG_DEFAULT_HUMAN_FACE_DETECT_MODEL 决定
    // lazy_load=true（默认）以减少启动期内存瞬时占用
    auto* detector = new(std::nothrow) HumanFaceDetect();
    if (!detector) {
        ESP_LOGE(TAG, "HumanFaceDetect 构造失败（PSRAM 不足？）");
        multi_heap_info_t info;
        heap_caps_get_info(&info, MALLOC_CAP_SPIRAM);
        ESP_LOGE(TAG, "PSRAM free=%u total_allocated=%u",
                 (unsigned)info.total_free_bytes,
                 (unsigned)info.total_allocated_bytes);
        s_handle = nullptr;
        vTaskDelete(NULL);
        return;
    }

    // 一次性打印启动时 PSRAM 占用供诊断（RESEARCH R2 风险跟踪）
    {
        multi_heap_info_t info;
        heap_caps_get_info(&info, MALLOC_CAP_SPIRAM);
        ESP_LOGI(TAG, "PSRAM after detector init: free=%u allocated=%u",
                 (unsigned)info.total_free_bytes,
                 (unsigned)info.total_allocated_bytes);
    }

    // 按 Kconfig 配置的 FPS 计算节拍
    const TickType_t period = pdMS_TO_TICKS(1000 / CONFIG_XIAOZHI_FACE_TRACKING_FPS);
    TickType_t last_wake = xTaskGetTickCount();
    int hit = 0, miss = 0;
    int64_t last_report_us = esp_timer_get_time();

    while (!s_stop) {
        vTaskDelayUntil(&last_wake, period);

        auto* cam = dynamic_cast<Esp32Camera*>(Board::GetInstance().GetCamera());
        if (!cam) {
            continue;
        }

        Esp32Camera::FrameRef f;
        if (!cam->CaptureForDetection(&f)) {
            // [T04 策略] 拿不到 mutex（MCP 拍照中）或 DQBUF 失败 → 正常跳帧
            continue;
        }

        // 组装 esp-dl 图像描述符
        // RESEARCH Pitfall A1：先假定 YUYV；若首轮 score 低于 0.5 可改 RGB565LE（决策点 D-B）
        dl::image::img_t img{};
        img.data = (void*)f.data;
        img.width = f.width;
        img.height = f.height;
        img.pix_type = dl::image::DL_IMAGE_PIX_TYPE_YUYV;

        int64_t t0 = esp_timer_get_time();
        auto& results = detector->run(img);
        int64_t t1 = esp_timer_get_time();

        // 立即归还 V4L2 缓冲，避免 face_track 占用时间长
        cam->ReleaseDetectionFrame(f);

        if (results.empty()) {
            miss++;
        } else {
            hit++;
            // PLAN 未明确排序策略，esp-dl 内部 nms 后 list 顺序不稳定
            // 为健壮性，挑 score 最高的那个（避免多脸时摇摆）
            const dl::detect::result_t* best = nullptr;
            for (const auto& r : results) {
                if (best == nullptr || r.score > best->score) {
                    best = &r;
                }
            }
            // box: [left_up_x, left_up_y, right_down_x, right_down_y]
            int cx = (best->box[0] + best->box[2]) / 2;
            int cy = (best->box[1] + best->box[3]) / 2;
            // 坐标映射（RESEARCH Pitfall 7）：严格保持 cx * 224 / width - 112
            // 对齐 RP2040 端 deadzone=20 / x_adj_factor=10 的基准
            int x_offset = (f.width  > 0) ? (cx * 224 / f.width  - 112) : 0;
            int y_offset = (f.height > 0) ? (cy * 224 / f.height - 112) : 0;

            // T07 完成后，uart_send_face 弱符号会被真实实现覆盖
            if (uart_send_face != nullptr) {
                uart_send_face(x_offset, y_offset);
            }
            ESP_LOGD(TAG, "face score=%.2f offset=(%d,%d) infer=%lldus",
                     best->score, x_offset, y_offset, (long long)(t1 - t0));
        }

        // 每 10 秒汇报一次统计（加保底避免除零）
        int64_t now = esp_timer_get_time();
        if (now - last_report_us > 10000000LL) {
            float elapsed_s = (now - last_report_us) / 1e6f;
            if (elapsed_s > 0.1f) {
                s_last_fps = (hit + miss) / elapsed_s;
                ESP_LOGI(TAG, "face stats: hit=%d miss=%d fps=%.1f",
                         hit, miss, s_last_fps);
            }
            hit = miss = 0;
            last_report_us = now;
        }
    }

    delete detector;
    ESP_LOGI(TAG, "face_tracker task exiting");
    s_handle = nullptr;
    vTaskDelete(NULL);
}

extern "C" void face_tracker_start(void) {
    if (s_handle != nullptr) {
        ESP_LOGW(TAG, "face_tracker already running, ignore start");
        return;
    }
    s_stop = false;
    // Core 0 + 优先级 2：低于 LVGL / 音频，避免抢占主路径
    // 栈 8KB：给 esp-dl 推理留充足空间
    BaseType_t ok = xTaskCreatePinnedToCore(
        face_tracker_task, "face_track",
        8 * 1024, nullptr, 2, &s_handle, 0);
    if (ok != pdPASS) {
        ESP_LOGE(TAG, "xTaskCreatePinnedToCore failed");
        s_handle = nullptr;
    }
}

extern "C" void face_tracker_stop(void) {
    s_stop = true;
}

extern "C" float face_tracker_get_fps(void) {
    return s_last_fps;
}

#else  // 非 S3 或功能未启用：提供空壳，保证链接通过

extern "C" void face_tracker_start(void) {}
extern "C" void face_tracker_stop(void) {}
extern "C" float face_tracker_get_fps(void) { return 0.0f; }

#endif  // CONFIG_XIAOZHI_ENABLE_FACE_TRACKING && CONFIG_IDF_TARGET_ESP32S3