yolov8 +双目实现区域多类别计数功能

1. 双目测距基础

本项目是在YOLOV8 + 双目测距的基础上实现的，测距原理可见上边文章

2. 计数功能

2.1 计数模块

在指定区域内计数模块

region_points = [(20, 20), (20, 600), (1200, 600), (1200, 20)]
count_class1 = 0
count_class2 = 0
for i, box in enumerate(boxes):
   x1, y1, x2, y2 = box[:4]
   x_center = (x1 + x2) / 2
   y_center = (y1 + y2) / 2
   center_point = (int(x_center), int(y_center))
   if is_inside_region(center_point, region_points):
       if box[-1] == 0:  # 类别1的标签.人
           count_class1 += 1
       elif  box[-1] == 2:  # 类别2的标签，车
           count_class2 += 1

2.2 判断模块

def is_inside_region(point, region_points):
    """
    判断点是否在指定区域内
    """
    return cv2.pointPolygonTest(np.array(region_points), point, False) >= 0

3. 初始代码

import time
import cv2
import numpy as np

from ultralytics import YOLO
from stereo import stereoconfig_040_2
from stereo.stereo import stereo_40
from stereo.stereo import stereo_threading, MyThread
from stereo.dianyuntu_yolo import preprocess, undistortion, getRectifyTransform, draw_line, rectifyImage, \
    stereoMatchSGBM

def is_inside_region(point, region_points):
    """
    判断点是否在指定区域内
    """
    return cv2.pointPolygonTest(np.array(region_points), point, False) >= 0

def count_objects_in_region(boxes, region_points):
    """
    在指定区域内计数物体
    """
    count = 0
    for box in boxes:
        x_center, y_center, width, height = box
        center_point = (int(x_center), int(y_center))
        if is_inside_region(center_point, region_points):
            count += 1
    return count

def detect():

    model = YOLO("yolov8n.pt")
    cv2.namedWindow('region couter', cv2.WINDOW_NORMAL)
    cv2.resizeWindow('region couter', 1280, 360)  # 设置宽高
    cap = cv2.VideoCapture('ultralytics/assets/a3.mp4')
    out_video = cv2.VideoWriter('output.avi', cv2.VideoWriter_fourcc(*'XVID'), 30, (2560, 720))

    region_points = [(20, 20), (20, 600), (1200, 600), (1200, 20)]
    region_points_np = np.array(region_points)

    assert cap.isOpened(), "Error reading video file"
    while cap.isOpened():
        success, im0 = cap.read()
        if not success:
            print("Video frame is empty or video processing has been successfully completed.")
            break

        config = stereoconfig_040_2.stereoCamera()
        map1x, map1y, map2x, map2y, Q = getRectifyTransform(720, 1280, config)
        thread = MyThread(stereo_threading, args=(config, im0, map1x, map1y, map2x, map2y, Q))
        thread.start()
        tracks = model.track(im0, persist=True)
        annotated_frame = tracks[0].plot()
        boxes = tracks[0].boxes.data
        count_class1 = 0
        count_class2 = 0

        for i, box in enumerate(boxes):
            x1, y1, x2, y2 = box[:4]
            x_center = (x1 + x2) / 2
            y_center = (y1 + y2) / 2
            center_point = (int(x_center), int(y_center))
            if is_inside_region(center_point, region_points):
                if box[-1] == 0:  # 类别1的标签.人
                    count_class1 += 1
                elif box[-1] == 2:  # 类别2的标签，车
                    count_class2 += 1
            if (0 < x2 < 1280):
                thread.join()
                points_3d = thread.get_result()
                a = points_3d[int(y_center), int(x_center), 0] / 1000
                b = points_3d[int(y_center), int(x_center), 1] / 1000
                c = points_3d[int(y_center), int(x_center), 2] / 1000
                distance = ((a ** 2 + b ** 2 + c ** 2) ** 0.5)
                if (distance != 0):
                    text_dis_avg = "dis:%0.2fm" % distance
                    cv2.putText(annotated_frame, text_dis_avg, (int(x2 + 5), int(y1 + 30)), cv2.FONT_ITALIC, 1.0,
                                (0, 255, 255), 2)

        cv2.polylines(annotated_frame, [region_points_np], isClosed=True, color=(255, 0, 0), thickness=2)

        print("Number of objects in class 1:", count_class1)
        print("Number of objects in class 2:", count_class2)
        cv2.imshow("region couter", annotated_frame)
        out_video.write(annotated_frame)
        if cv2.waitKey(1) == ord('q'):
            break
    out_video.release()
    cap.release()
    cv2.destroyAllWindows()
if __name__ == '__main__':
    detect()

4. 实验结果

本实验可以实现区域多类别计数
（1）区域计数实验

（2）如需把区域背景换成填充色，可将上边代码里的

cv2.polylines(annotated_frame, [region_points_np], isClosed=True, color=(255, 0, 0), thickness=2)

替换为

region_mask = np.zeros_like(annotated_frame)
cv2.fillPoly(region_mask, [region_points_np], color=(255, 0, 0))
# 使用透明度将填充后的区域与原始帧混合
alpha = 0.2  # 调整透明度
annotated_frame = cv2.addWeighted(annotated_frame, 1 - alpha, region_mask, alpha, 0)

（3）计数在图上显示
在annotated_frame = cv2.addWeighted(annotated_frame, 1 - alpha, region_mask, alpha, 0)添加

text1 = "count_class1:%d" % count_class1
text2 = "count_class2:%d" % count_class2
cv2.putText(annotated_frame, text1, (20, 30), cv2.FONT_ITALIC, 1.0, (0, 255, 255), 2)
cv2.putText(annotated_frame, text2, (20, 60), cv2.FONT_ITALIC, 1.0, (0, 255, 255), 2)

5. 完整代码

import time
import cv2
import numpy as np

from ultralytics import YOLO
from stereo import stereoconfig_040_2
from stereo.stereo import stereo_40
from stereo.stereo import stereo_threading, MyThread
from stereo.dianyuntu_yolo import preprocess, undistortion, getRectifyTransform, draw_line, rectifyImage, \
    stereoMatchSGBM

def is_inside_region(point, region_points):
    """
    判断点是否在指定区域内
    """
    return cv2.pointPolygonTest(np.array(region_points), point, False) >= 0

def count_objects_in_region(boxes, region_points):
    """
    在指定区域内计数物体
    """
    count = 0
    for box in boxes:
        x_center, y_center, width, height = box
        center_point = (int(x_center), int(y_center))
        if is_inside_region(center_point, region_points):
            count += 1
    return count

def detect():

    model = YOLO("yolov8n.pt")
    cv2.namedWindow('region couter', cv2.WINDOW_NORMAL)
    cv2.resizeWindow('region couter', 1280, 360)  # 设置宽高
    cap = cv2.VideoCapture('ultralytics/assets/a3.mp4')
    out_video = cv2.VideoWriter('output.avi', cv2.VideoWriter_fourcc(*'XVID'), 30, (2560, 720))

    region_points = [(20, 20), (20, 600), (1200, 600), (1200, 20)]
    region_points_np = np.array(region_points)

    assert cap.isOpened(), "Error reading video file"
    while cap.isOpened():
        success, im0 = cap.read()
        if not success:
            print("Video frame is empty or video processing has been successfully completed.")
            break

        config = stereoconfig_040_2.stereoCamera()
        map1x, map1y, map2x, map2y, Q = getRectifyTransform(720, 1280, config)
        thread = MyThread(stereo_threading, args=(config, im0, map1x, map1y, map2x, map2y, Q))
        thread.start()
        tracks = model.track(im0, persist=True)
        annotated_frame = tracks[0].plot()
        boxes = tracks[0].boxes.data
        count_class1 = 0
        count_class2 = 0

        for i, box in enumerate(boxes):
            x1, y1, x2, y2 = box[:4]
            x_center = (x1 + x2) / 2
            y_center = (y1 + y2) / 2
            center_point = (int(x_center), int(y_center))
            if is_inside_region(center_point, region_points):
                if box[-1] == 0:  # 类别1的标签.人
                    count_class1 += 1
                elif box[-1] == 2:  # 类别2的标签，车
                    count_class2 += 1
            if (0 < x2 < 1280):
                thread.join()
                points_3d = thread.get_result()
                a = points_3d[int(y_center), int(x_center), 0] / 1000
                b = points_3d[int(y_center), int(x_center), 1] / 1000
                c = points_3d[int(y_center), int(x_center), 2] / 1000
                distance = ((a ** 2 + b ** 2 + c ** 2) ** 0.5)
                if (distance != 0):
                    text_dis_avg = "dis:%0.2fm" % distance
                    cv2.putText(annotated_frame, text_dis_avg, (int(x2 + 5), int(y1 + 30)), cv2.FONT_ITALIC, 1.0,
                                (0, 255, 255), 2)

        # cv2.polylines(annotated_frame, [region_points_np], isClosed=True, color=(255, 0, 0), thickness=2)
        region_mask = np.zeros_like(annotated_frame)
        cv2.fillPoly(region_mask, [region_points_np], color=(255, 0, 0))
        # 使用透明度将填充后的区域与原始帧混合
        alpha = 0.2  # 调整透明度
        annotated_frame = cv2.addWeighted(annotated_frame, 1 - alpha, region_mask, alpha, 0)
        text1 = "count_class1:%d" % count_class1
        text2 = "count_class2:%d" % count_class2
        cv2.putText(annotated_frame, text1, (20, 30), cv2.FONT_ITALIC, 1.0, (0, 255, 255), 2)
        cv2.putText(annotated_frame, text2, (20, 60), cv2.FONT_ITALIC, 1.0, (0, 255, 255), 2)
        print("Number of objects in class 1:", count_class1)
        print("Number of objects in class 2:", count_class2)
        cv2.imshow("region couter", annotated_frame)
        out_video.write(annotated_frame)
        if cv2.waitKey(1) == ord('q'):
            break
    out_video.release()
    cap.release()
    cv2.destroyAllWindows()
if __name__ == '__main__':
    detect()

6. 源码

可以去 YOLOV8 + 双目测距 下载源码，然后把上边主代码贴进去运行即可