├── 16f9a22be2872390d459e84e36465654.png ├── 1fdc34704b25bd80a6f58e395f7d09fb.png ├── 31bb1f6f7b743f5d8c6db81871535537.png ├── 63a27a2c25b623e7d9cbe4401eea9168.png ├── 783e37718cba4b59ba766e834598b1db.png ├── README.md ├── b88796905edac4fda184f8909482e589.png └── threh.py /16f9a22be2872390d459e84e36465654.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/16f9a22be2872390d459e84e36465654.png -------------------------------------------------------------------------------- /1fdc34704b25bd80a6f58e395f7d09fb.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/1fdc34704b25bd80a6f58e395f7d09fb.png -------------------------------------------------------------------------------- /31bb1f6f7b743f5d8c6db81871535537.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/31bb1f6f7b743f5d8c6db81871535537.png -------------------------------------------------------------------------------- /63a27a2c25b623e7d9cbe4401eea9168.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/63a27a2c25b623e7d9cbe4401eea9168.png -------------------------------------------------------------------------------- /783e37718cba4b59ba766e834598b1db.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/783e37718cba4b59ba766e834598b1db.png -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 |  2 | # **1.视频演示：** 3 | 4 | [[项目分享]Python基于OpenCV高空抛物检测系统[完整源码＆部署教程]_哔哩哔哩_bilibili](https://www.bilibili.com/video/BV16Y4y1u7yL/?vd_source=bc9aec86d164b67a7004b996143742dc) 5 | 6 | 7 | # **2.图片演示：** 8 | ![1.png](1fdc34704b25bd80a6f58e395f7d09fb.png) 9 | 10 | ![2.png](16f9a22be2872390d459e84e36465654.png) 11 | 12 | ![3.png](b88796905edac4fda184f8909482e589.png) 13 | 14 | ![4.png](31bb1f6f7b743f5d8c6db81871535537.png) 15 | 16 | # **3.算法原理：** 17 | [参考该博客提出的移动侦测法](https://afdian.net/item?plan_id=34c97ad25f2511ed9b8e52540025c377)，即是根据视频每帧或者几帧之间像素的差异，对差异值设置阈值，筛选大于阈值的像素点，做掩模图即可选出视频中存在变化的桢。帧差法较为简单的视频中物体移动侦测，帧差法分为：单帧差、两桢差、和三桢差。随着帧数的增加是防止检测结果的重影。 18 | #### [帧间差分法(Temporal Difference)](https://mbd.pub/o/bread/Y5WTm5Zx) 19 | 由于场景中的目标在运动，目标的影像在不同图像帧中的位置不同。该类算法对时间上连续的两帧或三帧图像进行差分运算，不同帧对应的像素点相减，判断灰度差的绝对值，当绝对值超过一定阈值时，即可判断为运动目标，从而实现目标的检测功能。 20 | 21 | # **4.算法流程图：** 22 | ![b.png](63a27a2c25b623e7d9cbe4401eea9168.png) 23 | 24 | # **5.代码实现：** 25 | ``` 26 | def threh(video,save_video,thres1,area_threh): 27 | cam = cv2.VideoCapture(video)#打开一个视频 28 | input_fps = cam.get(cv2.CAP_PROP_FPS) 29 | ret_val, input_image = cam.read() 30 | index=[] 31 | images=[] 32 | images.append(input_image) 33 | video_length = int(cam.get(cv2.CAP_PROP_FRAME_COUNT)) 34 | input_image=cv2.resize(input_image,(512,512)) 35 | ending_frame = video_length 36 | fourcc = cv2.VideoWriter_fourcc(*'XVID') 37 | out = cv2.VideoWriter(save_video,fourcc, input_fps, (512, 512)) 38 | gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY) 39 | gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0) 40 | background=gray_lwpCV 41 | 42 | # es = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (9, 4)) 43 | 44 | i = 0 # default is 0 45 | outt=[] 46 | while(cam.isOpened()) and ret_val == True and i <2999: 47 | ## if i % 2==1: 48 | ret_val, input_image = cam.read() 49 | input_image=cv2.resize(input_image,(512,512)) 50 | gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY) 51 | gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0) 52 | diff = cv2.absdiff(background, gray_lwpCV) 53 | outt.append(diff) 54 | #跟着图像变换背景 55 | tem_diff=diff.flatten() 56 | tem_ds=pd.Series(tem_diff) 57 | tem_per=1-len(tem_ds[tem_ds==0])/len(tem_ds) 58 | if (tem_per <0.2 )| (tem_per>0.75): 59 | background=gray_lwpCV 60 | else: 61 | diff = cv2.threshold(diff, thres1, 255, cv2.THRESH_BINARY)[1] 62 | ret,thresh = cv2.threshold(diff.copy(),150,255,0) 63 | contours, hierarchy = cv2.findContours(thresh,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE) 64 | # contours, hierarchy = cv2.findContours(diff.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) 65 | for c in contours: 66 | if (cv2.contourArea(c) < area_threh) | (cv2.contourArea(c) >int(512*512*0.3) ) : # 对于矩形区域，只显示大于给定阈值的轮廓（去除微小的变化等噪点） 67 | continue 68 | (x, y, w, h) = cv2.boundingRect(c) # 该函数计算矩形的边界框 69 | cv2.rectangle(input_image, (x, y), (x+w, y+h), (0, 255, 0), 2) 70 | index.append(i) 71 | # cv2.imshow('contours', input_image) 72 | # cv2.imshow('dis', diff) 73 | out.write(input_image) 74 | images.append(input_image) 75 | i = i+1 76 | out.release() 77 | cam.release() 78 | return outt,index,images``` 79 | ##调取函数 80 | outt=threh('new_video.mp4','test6.mp4',25,3000) 81 | ``` 82 | # **6.系统整合：** 83 | 下图[完整源码&环境部署视频教程&自定义UI界面](https://s.xiaocichang.com/s/259db8) 84 | ![在这里插入图片描述](783e37718cba4b59ba766e834598b1db.png#pic_center) 85 | 参考博客[《Python基于OpenCV高空抛物检测系统\[完整源码＆部署教程\]》](https://mbd.pub/o/qunma/work) 86 | 87 | 88 | 89 | 90 | # **7.参考文献：** 91 | *** 92 | * *[1]*[论AI技术在高空抛物中的应用及解决方案](https://kns.cnki.net/kcms/detail/detail.aspx?filename=AQFF202104006&dbcode=CJFD&dbname=CJFD2021&v=qdTGf1uARVbBpDj-ia2sd2zLSa3Qe6c1QcoVNovK6WKKA9YNpAAghNt30x-4ZdK-)[J]. 谢程晨,许言. 中国安全防范技术与应用. 2021(04) 93 | * *[2]*[一种端边云结合的高空抛物检测解决方案](https://kns.cnki.net/kcms/detail/detail.aspx?filename=AQFF202104007&dbcode=CJFD&dbname=CJFD2021&v=qdTGf1uARVbyOnFMsaBnZwDXJA2sMldzMQSWh6GDIiWLS_KMkh6-sNrV_OtmjvJ5)[J]. 王志芳,代翔. 中国安全防范技术与应用. 2021(04) 94 | * *[3]*[高空抛物监测系统关键技术及应用](https://kns.cnki.net/kcms/detail/detail.aspx?filename=AQFF202104008&dbcode=CJFD&dbname=CJFD2021&v=qdTGf1uARVZnWKpIwyRSkndUr2XHg3L2fw7oDoPpSB87XWBwM73OjAj-ElMa45qf)[J]. 王作文. 中国安全防范技术与应用. 2021(04) 95 | * *[4]*[高空抛物致人损害立法研究及启示](https://kns.cnki.net/kcms/detail/detail.aspx?filename=FBZX202202047&dbcode=CJFD&dbname=CJFD2022&v=Ez-82ugnJwirZfpeB3SSrNQRX8yXBOmrdQl9rkqfIJrlrOfhGLPZXfuF3JAgDDNr)[J]. 左中原. 法制博览. 2022(02) 96 | * *[5]*[《关于依法妥善审理高空抛物、坠物案件的意见》的理解与适用](https://kns.cnki.net/kcms/detail/detail.aspx?filename=RMSF202016006&dbcode=CJFD&dbname=CJFD2020&v=KuqujOYnG2iKkyTR6rAycl9N-d105k6zoF9JyiC4GCKtn2hatkIj2dvgLudrVxQk)[J]. 姜启波,郭锋,陈龙业. 人民司法. 2020(16) 97 | * *[6]*[高空抛物现象产生的原因及防御对策](https://kns.cnki.net/kcms/detail/detail.aspx?filename=FBZX202106079&dbcode=CJFD&dbname=CJFD2021&v=8ZjdVpFHgPTaaWOWGBxxLSOeZuV3abDbduffN9AMoPGHYCtukGZvD29ASIWqGP-6)[J]. 任玉梅. 法制博览. 2021(06) 98 | * *[7]*[合力保护“头顶上的安全”——浅谈高空抛物坠物的反思](https://kns.cnki.net/kcms/detail/detail.aspx?filename=FBZX202107073&dbcode=CJFD&dbname=CJFD2021&v=8ZjdVpFHgPQbJDrqiWF1s-NPIQceZm_rPGsikaQDz8H5jgGsqBQ_N4SS2aOOUWaz)[J]. 李洛锋,叶政琴,谢玉婷. 法制博览. 2021(07) 99 | * *[8]*[高空抛物治理不妨多管齐下](https://kns.cnki.net/kcms/detail/detail.aspx?filename=ZYAQ202103004&dbcode=CJFD&dbname=CJFD2021&v=tSGaXLz0BXTPDdMAB_4zbiNwXO_xLVjtJCDtunxXD46PdLXG0FZ57-GAJSzDI2bS)[J]. 刘言. 现代职业安全. 2021(03) 100 | * *[9]*[法律中高空抛物坠物行为的责任承担](https://kns.cnki.net/kcms/detail/detail.aspx?filename=FBZX202113054&dbcode=CJFD&dbname=CJFD2021&v=8ZjdVpFHgPSBGcmZ8eF-KgLrmjpXgq7V-PnX25hFO2pmPBbtiGA1GkDRJQp_6siJ)[J]. 刘丽霞. 法制博览. 2021(13) 101 | * *[10]*[高空抛物的多维度理解](https://kns.cnki.net/kcms/detail/detail.aspx?filename=FBZX202115038&dbcode=CJFD&dbname=CJFD2021&v=8ZjdVpFHgPS7EkHJUkoR2vmcuI_Sz3Qsi8aiZljcrPlLtmx230PmJzPpr4vFSnxl)[J]. 刘春松. 法制博览. 2021(15) 102 | 103 | 104 | 105 | 106 | --- 107 | #### 如果您需要更详细的【源码和环境部署教程】，除了通过【系统整合】小节的链接获取之外，还可以通过邮箱以下途径获取: 108 | #### 1.请先在GitHub上为该项目点赞（Star），编辑一封邮件，附上点赞的截图、项目的中文描述概述（About）以及您的用途需求，发送到我们的邮箱 109 | #### sharecode@yeah.net 110 | #### 2.我们收到邮件后会定期根据邮件的接收顺序将【完整源码和环境部署教程】发送到您的邮箱。 111 | #### 【免责声明】本文来源于用户投稿，如果侵犯任何第三方的合法权益，可通过邮箱联系删除。 -------------------------------------------------------------------------------- /b88796905edac4fda184f8909482e589.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qunshansj/high-altitude-parabola-detection-opencv-python/58378e50fb5635decfa32d60b064cca01fb2d7d0/b88796905edac4fda184f8909482e589.png -------------------------------------------------------------------------------- /threh.py: -------------------------------------------------------------------------------- 1 | 2 | def threh(video,save_video,thres1,area_threh): 3 | cam = cv2.VideoCapture(video)#打开一个视频 4 | input_fps = cam.get(cv2.CAP_PROP_FPS) 5 | ret_val, input_image = cam.read() 6 | index=[] 7 | images=[] 8 | images.append(input_image) 9 | video_length = int(cam.get(cv2.CAP_PROP_FRAME_COUNT)) 10 | input_image=cv2.resize(input_image,(512,512)) 11 | ending_frame = video_length 12 | fourcc = cv2.VideoWriter_fourcc(*'XVID') 13 | out = cv2.VideoWriter(save_video,fourcc, input_fps, (512, 512)) 14 | gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY) 15 | gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0) 16 | background=gray_lwpCV 17 | 18 | # es = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (9, 4)) 19 | 20 | i = 0 # default is 0 21 | outt=[] 22 | while(cam.isOpened()) and ret_val == True and i <2999: 23 | ## if i % 2==1: 24 | ret_val, input_image = cam.read() 25 | input_image=cv2.resize(input_image,(512,512)) 26 | gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY) 27 | gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0) 28 | diff = cv2.absdiff(background, gray_lwpCV) 29 | outt.append(diff) 30 | #跟着图像变换背景 31 | tem_diff=diff.flatten() 32 | tem_ds=pd.Series(tem_diff) 33 | tem_per=1-len(tem_ds[tem_ds==0])/len(tem_ds) 34 | if (tem_per <0.2 )| (tem_per>0.75): 35 | background=gray_lwpCV 36 | else: 37 | diff = cv2.threshold(diff, thres1, 255, cv2.THRESH_BINARY)[1] 38 | ret,thresh = cv2.threshold(diff.copy(),150,255,0) 39 | contours, hierarchy = cv2.findContours(thresh,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE) 40 | # contours, hierarchy = cv2.findContours(diff.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) 41 | for c in contours: 42 | if (cv2.contourArea(c) < area_threh) | (cv2.contourArea(c) >int(512*512*0.3) ) : # 对于矩形区域，只显示大于给定阈值的轮廓（去除微小的变化等噪点） 43 | continue 44 | (x, y, w, h) = cv2.boundingRect(c) # 该函数计算矩形的边界框 45 | cv2.rectangle(input_image, (x, y), (x+w, y+h), (0, 255, 0), 2) 46 | index.append(i) 47 | # cv2.imshow('contours', input_image) 48 | # cv2.imshow('dis', diff) 49 | out.write(input_image) 50 | images.append(input_image) 51 | i = i+1 52 | out.release() 53 | cam.release() 54 | return outt,index,images --------------------------------------------------------------------------------