Our project called ChanPol (derived from InterPol) smoothens the process of collecting evidences against a crime, given the facial identity of suspect is known. It can collect important intel such as the location tracking, location history and course of suspect's actions to monitor them by aiming at face extraction and detection applied onto the video footage of 2,000+ pre-installed CCTVs spreaded over 114 km2 area to capture various angles and views.

Integration of CCTV Feeds using Various Cameras

• Python
• OpenCV (Haarcascade classifiers, Deepface)
• Flask

• We have 'N' no. of CCTV footages of different locations provided with us.
• Using Face Detection, our system finds all the faces present in any video.
• Then using Face Extraction, cropped screenshots of the faces are extracted from the video.
• A PDF is then auto-generated which consists of the faces of the persons seen in the video along with statistics such as the no. of occurrences of that face in the video & the particular timestamps when that face has occurred in the video.
• Then a particular suspect/criminal image is provided to us.
• We search for that particular image on all of the CCTV footages that we have.
• A PDF is then auto-generated which consists of a report that the suspect's face has been identified on which videos along with the above mentioned statistics.

• Given any suspect image, we can search and conclude from the generated report that where that suspect has been found.
• From the conclusions, we can visualise and generate a possible route of the suspect and probably find the current location of the suspect.
• From the first report that we generated, in which all the identified faces in all the videos were found, we can use that unused data in any future case.