bobo作业21

最新推荐文章于 2025-03-27 22:30:12 发布

壮壮的小仙女呀

最新推荐文章于 2025-03-27 22:30:12 发布

阅读量272

点赞数 1

CC 4.0 BY-SA版权

本文链接：https://blog.youkuaiyun.com/zzdxxn/article/details/89648150

本文介绍了一种基于视觉的目标跟踪方法，通过使用多种跟踪算法（如BOOSTING、MIL、KCF等），在视频序列中对目标进行实时跟踪。在跟踪过程中，每帧生成200个正样本和500个负样本，用于后续的模型训练。通过对目标位置的预测和更新，实现了对目标的稳定跟踪，并保存了样本以供进一步分析。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

作业21：

在跟踪的过程中，按照重叠率在每一帧生成200正，500负样本，并保存。

主函数：

import cv2
import sys
import os
import random
import numpy as np
from baby import read_directory, generate_randombox, solve_coincide

#(major_ver, minor_ver, subminor_ver) = (cv2.__version__).split('.')

if __name__ == '__main__':
    img_list= read_directory('./data/DragonBaby')
    pass
    # Set up tracker.
    # Instead of MIL, you can also use

    tracker_types = ['BOOSTING', 'MIL', 'KCF', 'TLD', 'MEDIANFLOW', 'GOTURN', 'MOSSE', 'CSRT']
    tracker_type = tracker_types[7]

    #if int(minor_ver) < 3:
        #tracker = cv2.Tracker_create(tracker_type)
    #else:
    if tracker_type == 'BOOSTING':
        tracker = cv2.TrackerBoosting_create()
    if tracker_type == 'MIL':
        tracker = cv2.TrackerMIL_create()
    if tracker_type == 'KCF':
        tracker = cv2.TrackerKCF_create()
    if tracker_type == 'TLD':
        tracker = cv2.TrackerTLD_create()
    if tracker_type == 'MEDIANFLOW':
        tracker = cv2.TrackerMedianFlow_create()
    if tracker_type == 'GOTURN':
        tracker = cv2.TrackerGOTURN_create()
    if tracker_type == 'MOSSE':
        tracker = cv2.TrackerMOSSE_create()
    if tracker_type == "CSRT":
        tracker = cv2.TrackerCSRT_create()

    # Initialize tracker with first frame and bounding box
    first_frame = cv2.imread(img_list[0])
    # Uncomment the line below to select a different bounding box
    init_bbox = cv2.selectROI(first_frame, False)
    cv2.destroyAllWindows()
    # 初始化只能在第一帧初始化
    ok = tracker.init(first_frame, init_bbox)

    for i in range(1, len(img_list)):
        img=cv2.imread(img_list[i])

        timer = cv2.getTickCount()

        # Update tracker
        ok, bbox = tracker.update(img)
        p_idx=0
        n_idx=0
        while True:
            new_r = generate_randombox(bbox)
            ratio=solve_coincide(bbox,new_r)
            print(ratio)
            if ratio>=0.5 and p_idx<=200 :
                imCrop = img[new_r[3]:new_r[1], new_r[0]:new_r[2], :]
                res = cv2.resize(imCrop, (32, 32), interpolation=cv2.INTER_CUBIC)
                cv2.imwrite('E:\\pratice\\homework21\\positive\\baby' + str(i) + str(p_idx)+'.jpg', res)
                p_idx+=1
            elif ratio<0.5 and n_idx<=500 :
                imCrop = img[new_r[3]:new_r[1], new_r[0]:new_r[2], :]
                res = cv2.resize(imCrop, (32, 32), interpolation=cv2.INTER_CUBIC)
                cv2.imwrite('E:\\pratice\\homework21\\negative\\baby' + str(i) +str(n_idx)+ '.jpg', res)
                n_idx+=1
            if p_idx==200 and n_idx==500:
                break


        # Calculate Frames per second (FPS)
        fps = cv2.getTickFrequency() / (cv2.getTickCount() - timer);

        # Draw bounding box
        if ok:
            # Tracking success
            p1 = (int(bbox[0]), int(bbox[1]))
            p2 = (int(bbox[0] + bbox[2]), int(bbox[1] + bbox[3]))
            cv2.rectangle(img, p1, p2, (0, 0, 255), 2, 1)
            L_x = int(bbox[0])
            L_y = int(bbox[1])
            R_x = int(bbox[0] + bbox[2])
            R_y = int(bbox[1] + bbox[3])
            w = np.array([L_x, L_y, R_x, R_y])
            # Display tracker type on frame
            cv2.putText(img, tracker_type + " Tracker", (100, 20), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2);
            # Display FPS on frame
            cv2.putText(img, "FPS : " + str(int(fps)), (100, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2);
            # Display result
            new_r = generate_randombox(bbox)
            imCrop = img[new_r[3]:new_r[1], new_r[0]:new_r[2], :]
            cv2.rectangle(img, (new_r[0], new_r[1]), (new_r[2], new_r[3]), (55, 255, 155), 2)
            res = cv2.resize(imCrop, (32, 32), interpolation=cv2.INTER_CUBIC)
            cv2.imshow("Tracking", img)
            cv2.imwrite('pigy' + str(i) + '.jpg', res)
            cv2.imshow("Tracking", img)
        else:
            # Tracking failure
            cv2.putText(img, "Tracking failure detected", (100, 80), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (0, 0, 255), 2)

            # Display tracker type on frame
            cv2.putText(img, tracker_type + " Tracker", (100, 20), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2);
            # Display FPS on frame
            cv2.putText(img, "FPS : " + str(int(fps)), (100, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.75, (50, 170, 50), 2);

            # Display result

#生成随机

            new_r = generate_randombox(bbox)
            cv2.rectangle(img, (new_r[0], new_r[1]), (new_r[2], new_r[3]), (55, 255, 155), 2)
            res = cv2.resize(imCrop, (32, 32), interpolation=cv2.INTER_CUBIC)
            imCrop = img[new_r[3]:new_r[1], new_r[0]:new_r[2], :]
            cv2.imshow("Tracking", img)
            cv2.imwrite('pigy' + str(i) + '.jpg', res)
#计算重叠率


        # Exit if ESC pressed
        k = cv2.waitKey(1000) & 0xff
        if k == 27:
            break

子函数：

import os
import cv2
import sys
import random
import numpy as np

array_of_img = [] # this if for store all of the image data
# this function is for read image,the input is directory name
def read_directory(directory_name):
    # 定位到存放图片的目录
    img_dir = directory_name + r"/img/"
    # 定位到存放ground_truth的目录
    #gt = directory_name+r'/groundtruth.txt'
    # 生成img_dir目录下的所有图像
    img_list = [img_dir+i for i in os.listdir(img_dir)]####str与list
    #f=open(gt)
    #gt_list = f.readlines()
    return img_list

def mat_inter(box1, box2):
    # 判断两个矩形是否相交
    # box=(xA,yA,xB,yB)
    L_x, L_y, R_x, R_y = box1
    left_x_up, left_y_up, right_x_down, right_y_down = box2

    lx = abs((L_x + R_x) / 2 - (left_x_up + right_x_down) / 2)
    ly = abs((L_y + R_y) / 2 - (left_y_up + right_y_down) / 2)
    sax = abs(L_x - R_x)
    sbx = abs(left_x_up - right_x_down)
    say = abs(L_y - R_y)
    sby = abs(left_y_up - right_y_down)
    if lx <= (sax + sbx) / 2 and ly <= (say + sby) / 2:
        return True
    else:
        return False


def solve_coincide(box1, box2):
    # box=(xA,yA,xB,yB)
    # 计算两个矩形框的重合度
    if mat_inter(box1, box2) == True:
        L_x, L_y, R_x, R_y = box1
        left_x_up, left_y_up, right_x_down, right_y_down = box2
        col = min(R_x, right_x_down) - max(L_x, left_x_up)
        row = min(R_y, right_y_down) - max(L_y, left_y_up)
        intersection = col * row
        area1 = abs(R_x - L_x) * abs(R_y - L_y)
        area2 = abs(right_x_down - left_x_up) * abs(right_y_down - left_y_up)
        coincide = abs(intersection / (area1 + area2 - intersection))
        return coincide
    else:
        return False


def generate_randombox(bbox):
    x = int(bbox[0]) +bbox[2] // 2  # r中的元素一定要分清楚Imcrop里谁是左X谁是右X谁是左Y谁是右，按这个对照imCrop = frame[right_y_down:left_y_up, left_x_up:right_x_down, :]
    y = int(bbox[1]) +bbox[3] // 2  # //为python整除
    left_x_up = x - random.randint(0, 100)
    left_y_up = y + random.randint(0, 100)  # 注意这里的调参
    right_x_down = x + random.randint(5, 100)
    right_y_down = y - random.randint(5, 100)
    new_r = np.array([int(left_x_up), int(left_y_up), int(right_x_down), int(right_y_down)])
    return new_r