feat: draw fixed rectangle for intrusion detection
This commit is contained in:
parent
0d7bba41a1
commit
31d8740bee
|
|
@ -5,7 +5,7 @@ cmake_minimum_required(VERSION 3.4.1)
|
|||
project(IntrusionDetectionService CXX)
|
||||
|
||||
# 设置 C++ 标准和编译选项
|
||||
set(CMAKE_CXX_STANDARD 14)
|
||||
set(CMAKE_CXX_STANDARD 17)
|
||||
set(CMAKE_CXX_STANDARD_REQUIRED ON)
|
||||
set(CMAKE_CXX_FLAGS "-pthread")
|
||||
|
||||
|
|
|
|||
|
|
@ -2,14 +2,30 @@
|
|||
#define RKYOLOV5S_H
|
||||
|
||||
#include "rknn_api.h"
|
||||
|
||||
#include "opencv2/core/core.hpp"
|
||||
#include <map> // 用于存储跟踪目标
|
||||
#include <mutex> // 确保线程安全
|
||||
#include <string> // 使用 std::string
|
||||
#include <vector> // 使用 std::vector
|
||||
#include "postprocess.h"
|
||||
|
||||
// 前置声明
|
||||
static void dump_tensor_attr(rknn_tensor_attr *attr);
|
||||
static unsigned char *load_data(FILE *fp, size_t ofst, size_t sz);
|
||||
static unsigned char *load_model(const char *filename, int *model_size);
|
||||
static int saveFloat(const char *file_name, float *output, int element_size);
|
||||
|
||||
// 用于跟踪单个目标的结构体
|
||||
struct TrackedPerson
|
||||
{
|
||||
int id; // 唯一ID
|
||||
cv::Rect box; // 当前位置
|
||||
double entry_time; // 进入入侵区域的时间戳 (秒)
|
||||
bool is_in_zone; // 是否在区域内
|
||||
bool alarm_triggered; // 是否已触发报警
|
||||
int frames_unseen; // 消失的帧数
|
||||
};
|
||||
|
||||
class rkYolov5s
|
||||
{
|
||||
private:
|
||||
|
|
@ -29,12 +45,24 @@ private:
|
|||
|
||||
float nms_threshold, box_conf_threshold;
|
||||
|
||||
// 入侵检测和跟踪相关成员
|
||||
cv::Rect intrusion_zone; // 入侵区域
|
||||
std::map<int, TrackedPerson> tracked_persons; // 存储所有被跟踪的人
|
||||
int next_track_id; // 用于分配新的唯一ID
|
||||
double intrusion_time_threshold; // 入侵时间阈值 (秒)
|
||||
|
||||
// 跟踪逻辑的私有方法
|
||||
void update_tracker(detect_result_group_t &detect_result_group);
|
||||
|
||||
public:
|
||||
rkYolov5s(const std::string &model_path);
|
||||
int init(rknn_context *ctx_in, bool isChild);
|
||||
rknn_context *get_pctx();
|
||||
cv::Mat infer(cv::Mat &ori_img);
|
||||
~rkYolov5s();
|
||||
|
||||
// 用于从外部设置入侵区域的公共方法
|
||||
void set_intrusion_zone(const cv::Rect& zone);
|
||||
};
|
||||
|
||||
#endif
|
||||
#endif // RKYOLOV5S_H
|
||||
|
|
@ -1,16 +1,33 @@
|
|||
#include <stdio.h>
|
||||
#include <mutex>
|
||||
#include "rknn_api.h"
|
||||
#include <chrono> // 用于计时
|
||||
#include <string> // 使用 std::string
|
||||
#include <vector> // 使用 std::vector
|
||||
#include <algorithm> // 使用 std::min/max
|
||||
|
||||
#include "postprocess.h"
|
||||
#include "preprocess.h"
|
||||
#include "rkYolov5s.hpp"
|
||||
#include "coreNum.hpp"
|
||||
|
||||
#include "opencv2/core/core.hpp"
|
||||
#include "opencv2/highgui/highgui.hpp"
|
||||
#include "opencv2/imgproc/imgproc.hpp"
|
||||
#include "rknn_api.h"
|
||||
|
||||
#include "coreNum.hpp"
|
||||
#include "rkYolov5s.hpp"
|
||||
// 报警接口函数 (目前只打印信息)
|
||||
void trigger_alarm(int person_id, const cv::Rect& box) {
|
||||
printf("[ALARM] Intrusion detected! Person ID: %d at location (%d, %d, %d, %d)\n",
|
||||
person_id, box.x, box.y, box.width, box.height);
|
||||
// TODO: 在这里实现真正的报警逻辑，例如发送网络消息、写入数据库等。
|
||||
}
|
||||
|
||||
// 获取当前时间的函数 (返回秒)
|
||||
double get_current_time_seconds() {
|
||||
return std::chrono::duration_cast<std::chrono::duration<double>>(
|
||||
std::chrono::high_resolution_clock::now().time_since_epoch()
|
||||
).count();
|
||||
}
|
||||
|
||||
static void dump_tensor_attr(rknn_tensor_attr *attr)
|
||||
{
|
||||
|
|
@ -19,13 +36,6 @@ static void dump_tensor_attr(rknn_tensor_attr *attr)
|
|||
{
|
||||
shape_str += ", " + std::to_string(attr->dims[i]);
|
||||
}
|
||||
|
||||
// printf(" index=%d, name=%s, n_dims=%d, dims=[%s], n_elems=%d, size=%d, w_stride = %d, size_with_stride=%d, fmt=%s, "
|
||||
// "type=%s, qnt_type=%s, "
|
||||
// "zp=%d, scale=%f\n",
|
||||
// attr->index, attr->name, attr->n_dims, shape_str.c_str(), attr->n_elems, attr->size, attr->w_stride,
|
||||
// attr->size_with_stride, get_format_string(attr->fmt), get_type_string(attr->type),
|
||||
// get_qnt_type_string(attr->qnt_type), attr->zp, attr->scale);
|
||||
}
|
||||
|
||||
static unsigned char *load_data(FILE *fp, size_t ofst, size_t sz)
|
||||
|
|
@ -95,8 +105,19 @@ static int saveFloat(const char *file_name, float *output, int element_size)
|
|||
rkYolov5s::rkYolov5s(const std::string &model_path)
|
||||
{
|
||||
this->model_path = model_path;
|
||||
nms_threshold = NMS_THRESH; // 默认的NMS阈值
|
||||
box_conf_threshold = BOX_THRESH; // 默认的置信度阈值
|
||||
nms_threshold = NMS_THRESH;
|
||||
box_conf_threshold = BOX_THRESH;
|
||||
|
||||
// 初始化跟踪器和入侵检测参数
|
||||
next_track_id = 1;
|
||||
intrusion_time_threshold = 3.0; // 报警时间阈值：3秒
|
||||
// 默认设置一个无效的入侵区域，将在第一帧时根据图像大小初始化
|
||||
intrusion_zone = cv::Rect(0, 0, 0, 0);
|
||||
}
|
||||
|
||||
void rkYolov5s::set_intrusion_zone(const cv::Rect& zone) {
|
||||
std::lock_guard<std::mutex> lock(mtx);
|
||||
this->intrusion_zone = zone;
|
||||
}
|
||||
|
||||
int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
||||
|
|
@ -104,7 +125,6 @@ int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
|||
printf("Loading model...\n");
|
||||
int model_data_size = 0;
|
||||
model_data = load_model(model_path.c_str(), &model_data_size);
|
||||
// 模型参数复用/Model parameter reuse
|
||||
if (share_weight == true)
|
||||
ret = rknn_dup_context(ctx_in, &ctx);
|
||||
else
|
||||
|
|
@ -115,7 +135,6 @@ int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
|||
return -1;
|
||||
}
|
||||
|
||||
// 设置模型绑定的核心/Set the core of the model that needs to be bound
|
||||
rknn_core_mask core_mask;
|
||||
switch (get_core_num())
|
||||
{
|
||||
|
|
@ -145,7 +164,6 @@ int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
|||
}
|
||||
printf("sdk version: %s driver version: %s\n", version.api_version, version.drv_version);
|
||||
|
||||
// 获取模型输入输出参数/Obtain the input and output parameters of the model
|
||||
ret = rknn_query(ctx, RKNN_QUERY_IN_OUT_NUM, &io_num, sizeof(io_num));
|
||||
if (ret < 0)
|
||||
{
|
||||
|
|
@ -154,7 +172,6 @@ int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
|||
}
|
||||
printf("model input num: %d, output num: %d\n", io_num.n_input, io_num.n_output);
|
||||
|
||||
// 设置输入参数/Set the input parameters
|
||||
input_attrs = (rknn_tensor_attr *)calloc(io_num.n_input, sizeof(rknn_tensor_attr));
|
||||
for (int i = 0; i < io_num.n_input; i++)
|
||||
{
|
||||
|
|
@ -168,7 +185,6 @@ int rkYolov5s::init(rknn_context *ctx_in, bool share_weight)
|
|||
dump_tensor_attr(&(input_attrs[i]));
|
||||
}
|
||||
|
||||
// 设置输出参数/Set the output parameters
|
||||
output_attrs = (rknn_tensor_attr *)calloc(io_num.n_output, sizeof(rknn_tensor_attr));
|
||||
for (int i = 0; i < io_num.n_output; i++)
|
||||
{
|
||||
|
|
@ -208,6 +224,94 @@ rknn_context *rkYolov5s::get_pctx()
|
|||
return &ctx;
|
||||
}
|
||||
|
||||
void rkYolov5s::update_tracker(detect_result_group_t &detect_result_group)
|
||||
{
|
||||
std::vector<cv::Rect> current_detections;
|
||||
for (int i = 0; i < detect_result_group.count; i++) {
|
||||
detect_result_t *det_result = &(detect_result_group.results[i]);
|
||||
if (strcmp(det_result->name, "person") == 0) {
|
||||
current_detections.push_back(cv::Rect(
|
||||
det_result->box.left, det_result->box.top,
|
||||
det_result->box.right - det_result->box.left,
|
||||
det_result->box.bottom - det_result->box.top));
|
||||
}
|
||||
}
|
||||
|
||||
// 1. 对于已有的跟踪目标，增加其未见帧数
|
||||
for (auto it = tracked_persons.begin(); it != tracked_persons.end(); ++it) {
|
||||
it->second.frames_unseen++;
|
||||
}
|
||||
|
||||
// 2. 将当前帧的检测结果与已有的跟踪目标进行匹配
|
||||
for (const auto& det_box : current_detections) {
|
||||
bool is_matched = false;
|
||||
int best_match_id = -1;
|
||||
double max_iou = 0.3; // IoU阈值，用于判断是否为同一目标
|
||||
|
||||
for (auto const& [id, person] : tracked_persons) {
|
||||
// 计算交并比 (Intersection over Union)
|
||||
double iou = (double)(det_box & person.box).area() / (double)(det_box | person.box).area();
|
||||
if (iou > max_iou) {
|
||||
max_iou = iou;
|
||||
best_match_id = id;
|
||||
}
|
||||
}
|
||||
|
||||
if (best_match_id != -1) {
|
||||
// 匹配成功，更新目标信息
|
||||
tracked_persons[best_match_id].box = det_box;
|
||||
tracked_persons[best_match_id].frames_unseen = 0;
|
||||
is_matched = true;
|
||||
} else {
|
||||
// 匹配失败，创建新的跟踪目标
|
||||
TrackedPerson new_person;
|
||||
new_person.id = next_track_id++;
|
||||
new_person.box = det_box;
|
||||
new_person.entry_time = 0;
|
||||
new_person.is_in_zone = false;
|
||||
new_person.alarm_triggered = false;
|
||||
new_person.frames_unseen = 0;
|
||||
tracked_persons[new_person.id] = new_person;
|
||||
}
|
||||
}
|
||||
|
||||
// 3. 处理和更新每个目标的状态
|
||||
double current_time = get_current_time_seconds();
|
||||
for (auto it = tracked_persons.begin(); it != tracked_persons.end(); ++it) {
|
||||
TrackedPerson& person = it->second;
|
||||
// 判断人员包围盒是否与入侵区域有交集
|
||||
bool currently_in_zone = (intrusion_zone & person.box).area() > 0;
|
||||
|
||||
if (currently_in_zone) {
|
||||
if (!person.is_in_zone) {
|
||||
// 刚进入区域
|
||||
person.is_in_zone = true;
|
||||
person.entry_time = current_time;
|
||||
} else {
|
||||
// 已在区域内，检查是否超时
|
||||
if (!person.alarm_triggered && (current_time - person.entry_time) > intrusion_time_threshold) {
|
||||
person.alarm_triggered = true;
|
||||
trigger_alarm(person.id, person.box);
|
||||
}
|
||||
}
|
||||
} else {
|
||||
// 不在区域内，重置状态
|
||||
person.is_in_zone = false;
|
||||
person.entry_time = 0;
|
||||
person.alarm_triggered = false;
|
||||
}
|
||||
}
|
||||
|
||||
// 4. 移除消失太久的目标
|
||||
for (auto it = tracked_persons.begin(); it != tracked_persons.end(); ) {
|
||||
if (it->second.frames_unseen > 20) { // 超过20帧未见则移除
|
||||
it = tracked_persons.erase(it);
|
||||
} else {
|
||||
++it;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
cv::Mat rkYolov5s::infer(cv::Mat &orig_img)
|
||||
{
|
||||
std::lock_guard<std::mutex> lock(mtx);
|
||||
|
|
@ -220,14 +324,12 @@ cv::Mat rkYolov5s::infer(cv::Mat &orig_img)
|
|||
memset(&pads, 0, sizeof(BOX_RECT));
|
||||
cv::Size target_size(width, height);
|
||||
cv::Mat resized_img(target_size.height, target_size.width, CV_8UC3);
|
||||
// 计算缩放比例/Calculate the scaling ratio
|
||||
|
||||
float scale_w = (float)target_size.width / img.cols;
|
||||
float scale_h = (float)target_size.height / img.rows;
|
||||
|
||||
// 图像缩放/Image scaling
|
||||
if (img_width != width || img_height != height)
|
||||
{
|
||||
// rga
|
||||
rga_buffer_t src;
|
||||
rga_buffer_t dst;
|
||||
memset(&src, 0, sizeof(src));
|
||||
|
|
@ -237,13 +339,6 @@ cv::Mat rkYolov5s::infer(cv::Mat &orig_img)
|
|||
{
|
||||
fprintf(stderr, "resize with rga error\n");
|
||||
}
|
||||
/*********
|
||||
// opencv
|
||||
float min_scale = std::min(scale_w, scale_h);
|
||||
scale_w = min_scale;
|
||||
scale_h = min_scale;
|
||||
letterbox(img, resized_img, pads, min_scale, target_size);
|
||||
*********/
|
||||
inputs[0].buf = resized_img.data;
|
||||
}
|
||||
else
|
||||
|
|
@ -260,11 +355,9 @@ cv::Mat rkYolov5s::infer(cv::Mat &orig_img)
|
|||
outputs[i].want_float = 0;
|
||||
}
|
||||
|
||||
// 模型推理/Model inference
|
||||
ret = rknn_run(ctx, NULL);
|
||||
ret = rknn_outputs_get(ctx, io_num.n_output, outputs, NULL);
|
||||
|
||||
// 后处理/Post-processing
|
||||
detect_result_group_t detect_result_group;
|
||||
std::vector<float> out_scales;
|
||||
std::vector<int32_t> out_zps;
|
||||
|
|
@ -276,32 +369,32 @@ cv::Mat rkYolov5s::infer(cv::Mat &orig_img)
|
|||
post_process((int8_t *)outputs[0].buf, (int8_t *)outputs[1].buf, (int8_t *)outputs[2].buf, height, width,
|
||||
box_conf_threshold, nms_threshold, pads, scale_w, scale_h, out_zps, out_scales, &detect_result_group);
|
||||
|
||||
// 绘制框体/Draw the box
|
||||
char text[256];
|
||||
for (int i = 0; i < detect_result_group.count; i++)
|
||||
{
|
||||
detect_result_t *det_result = &(detect_result_group.results[i]);
|
||||
// 更新跟踪器状态
|
||||
// 首次运行时，根据图像尺寸初始化入侵区域 (设定在画面中央)
|
||||
if (intrusion_zone.width == 0 || intrusion_zone.height == 0) {
|
||||
intrusion_zone = cv::Rect(orig_img.cols / 4, orig_img.rows / 4, orig_img.cols / 2, orig_img.rows / 2);
|
||||
}
|
||||
update_tracker(detect_result_group);
|
||||
|
||||
// ========================== 修改开始 ==========================
|
||||
// 增加一个判断，只处理标签名称为 "person" 的结果
|
||||
if (strcmp(det_result->name, "person") == 0)
|
||||
{
|
||||
sprintf(text, "%s %.1f%%", det_result->name, det_result->prop * 100);
|
||||
// 打印预测物体的信息/Prints information about the predicted object
|
||||
// printf("%s @ (%d %d %d %d) %f\n", det_result->name, det_result->box.left, det_result->box.top,
|
||||
// det_result->box.right, det_result->box.bottom, det_result->prop);
|
||||
int x1 = det_result->box.left;
|
||||
int y1 = det_result->box.top;
|
||||
int x2 = det_result->box.right;
|
||||
int y2 = det_result->box.bottom;
|
||||
rectangle(orig_img, cv::Point(x1, y1), cv::Point(x2, y2), cv::Scalar(256, 0, 0, 256), 3);
|
||||
putText(orig_img, text, cv::Point(x1, y1 + 12), cv::FONT_HERSHEY_SIMPLEX, 0.4, cv::Scalar(255, 255, 255));
|
||||
// 绘制入侵区域
|
||||
cv::rectangle(orig_img, intrusion_zone, cv::Scalar(255, 255, 0), 2); // 黄色
|
||||
|
||||
// 绘制框体和报警状态
|
||||
for (auto const& [id, person] : tracked_persons) {
|
||||
// 根据是否触发报警决定颜色 (BGR: 红色 vs 绿色)
|
||||
cv::Scalar box_color = person.alarm_triggered ? cv::Scalar(0, 0, 255) : cv::Scalar(0, 255, 0);
|
||||
int line_thickness = person.alarm_triggered ? 3 : 2;
|
||||
|
||||
cv::rectangle(orig_img, person.box, box_color, line_thickness);
|
||||
std::string label = "Person " + std::to_string(id);
|
||||
if (person.is_in_zone) {
|
||||
label += " (In Zone)";
|
||||
}
|
||||
// ========================== 修改结束 ==========================
|
||||
cv::putText(orig_img, label, cv::Point(person.box.x, person.box.y - 10),
|
||||
cv::FONT_HERSHEY_SIMPLEX, 0.5, box_color, 2);
|
||||
}
|
||||
|
||||
ret = rknn_outputs_release(ctx, io_num.n_output, outputs);
|
||||
|
||||
return orig_img;
|
||||
}
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue