如何计算穿过线路后的物体？

Question

我已经检测到视频帧中的对象并对其进行了标记，并且还计算了帧中的对象总数，但我的问题是如何在通过如图所示的线后对对象进行计数。以及对象类别。

这是我的代码

在图像中，我已经计算了框架中的总对象，但我想在它们越线时计算它们。

import cv2
import numpy as np
net = cv2.dnn.readNet('yolov3.weights','yolov3.cfg')
classes = []
with open('coco.names','r') as f:
    classes = f.read().splitlines()
# printing the data which is loaded from the names file
#print(classes)
cap = cv2.VideoCapture('video.mp4')

while True:
    _,img = cap.read()
    height, width, _ = img.shape
    blob = cv2.dnn.blobFromImage(img, 1/255,  (416 , 416), (0,0,0) ,swapRB=True,crop=False)
    net.setInput(blob)
    output_layer_names = net.getUnconnectedOutLayersNames()
    layerOutput = net.forward(output_layer_names)
    boxes = []
    person =0
    truck =0
    car = 0
    confidences = []
    class_ids =[]
    for output in layerOutput:
        for detection in output:
            scores = detection[5:]
            class_id = np.argmax(scores)
            confidence = scores[class_id]
            if confidence > 0.5:
                center_x = int(detection[0]*width)
                center_y = int(detection[1]*height)
                w = int(detection[2]*width)
                h = int(detection[3]*height)


                x = int(center_x - w/2)
                y = int(center_y - h/2)

                boxes.append([x,y,w,h])
                confidences.append((float(confidence)))
                class_ids.append(class_id)
    indexes = cv2.dnn.NMSBoxes(boxes,confidences,0.5,0.4)
    font = cv2.QT_FONT_NORMAL
    colors = np.random.uniform(0,255,size=(len(boxes),3))
    for i in indexes.flatten():
        labelsss = str(classes[class_ids[i]])
        if(labelsss == 'person'):
            person+=1
        if(labelsss == 'car'):
            car+=1
        if(labelsss == 'truck'):
            truck+=1

    for i in indexes.flatten():
        x,y,w,h = boxes[i]
        label =str(classes[class_ids[i]])
        confidence = str(round(confidences[i],1))
        color = colors[i]
        cv2.rectangle(img,(x,y),(x+w , y+h), color, 2)
        cv2.line(img,(1000,250),(5,250),(0,0,0),2)
              
        cv2.putText(img, label + " ", (x, y+20), font, 0.5, (255,255,255),2)
        cv2.putText(img, 'Car'+ ":" + str(car), (20, 20), font, 0.8, (0,0,0),2)
        cv2.putText(img, 'Person'+ ":" + str(person), (20, 50), font, 0.8, (0,0,0),2)
        cv2.putText(img, 'Truck'+ ":" + str(truck), (20, 80), font, 0.8, (0,0,0),2)
        cv2.imshow('Image',img)
        key = cv2.waitKey(1)
        if key == 10:
            break
        
cap.release()
cv2.destroyAllWindows()

Answer 1

我实习期间就做了一个这样的项目。您可以在此处查看代码：https://github.com/sarimmehdi/nanonets_object_tracking/blob/master/test_on_video.py

简而言之：您应该绘制一个矩形（窄），并在穿过矩形时计算跟踪的 ID。如果矩形足够窄，也可以避免重新识别的问题。