How Drones are Used to Enhance Physical Security in 2021
In 2020, the drone market size was estimated to be around $22.5 billion and is supposed to grow at 13.8% CAGR. Let us discuss some of the challenges faced by this industry to get a glimpse of what the future holds.
Advantages of using Artificial Intelligence in Video Surveillance
According to a recent projection, the AI video analytics market is all set to grow into a $24,341.5 million industry by 2027. Let's have a look at ways how different AI technologies could be leveraged for physical threat detection in real-world scenarios.
Photos of the person are required to launch the AI-powered Person Search System. The more diverse, the better. On average we’ll need 4+ photos where the person is visible from different angles. The photos can be uploaded to the system from phones or other sources (as long as the person of interest is currently wearing the same clothes and accessories). Alternatively, the photos can be taken from live streams of any camera that is in the system, or forensically from recordings.
No, the system takes into account the overall appearance of the person of interest.
Scylla’s Person Search technology looks for the person of interest in all cameras all over the field of view. As such, the same person can appear of different height and/or proportion when shots are taken from different distances and angles. Thus the height cannot be taken into account. Same applies to the weight and/or shape of the person.
Yes, absolutely. The Scylla Face Recognition technology can be implemented to aid the Scylla Person Search System and to cover the cases where the first one is more effective. For more details please refer to the relevant question below.
The obvious advantage is the ability to search for the person not based on facial biometrics. This means that the Person Search System will detect the person of interest where the person is not facing the camera, the face is under a mask or hidden under a hoodie, etc. Moreover, as an input Scylla Person Search System can take low quality photos of the person of interest where the face is not visible at all.
Scylla Person Search System can be implemented on most of the cameras with versatile placement location and view angle. Opposed to that, face recognition based search can operate only on cameras that are placed in a certain location and angle - to ensure people pass in front of the camera looking at it.
Since the Scylla Person Search System looks for the person of interest based on similarity of appearance, there might be more false guesses compared to the more “choosy” face recognition based search. In other words, there might be quite a few people in the view of the camera that wear very similar clothes. The verification feedback of the operating user should help to minimize errors during the search.
The search depends on the appearance (i.e. clothing). Thus it cannot be implemented in cases where the clothing was changed. For example, it cannot be used to search for the same person on different days. Similarly, it cannot be used in cases where the input photos were taken outdoors (i.e. with a coat on) and the search is engaged indoors (i.e. the coat is off).
While the Scylla Person Search System can search for multiple people simultaneously, the total number is limited by the hardware capabilities that run the software. The optimal maximum number of people that can be simultaneously searched for is 10.
Yes, the system can search for several people on all the cameras simultaneously. The search instances can be initiated and halted independently. The maximum number of people simultaneously sought for is limited by the hardware that runs the software. Typical optimal maximum is 10 people of interest.
There is. When Scylla Person Search System finds a person similar to the one requested it sends suggestions to the dashboard for the user to review and confirm. In case the user rejects the suggestion this information is sent back to the search engine as an “anti-input” that narrows down the search criteria and helps to eliminate similar mistakes henceforward. Similarly, if the user confirms the guess, this information is added to the system to aid the search in real-time.
Scylla Person Search System can operate on HD cameras. However, especially for the cases where people are captured at a distance, Person Search technology will perform better if the cameras are Full HD or higher.
The maximum distance at which Scylla Person Search System will reliably operate depends on a number of camera parameters such as resolution and view angle. These will define the size of the people in pixels and affect the performance of the system. Naturally, the more pixels per person height there are, the more details reach the system and the more accurate the matching guesses will be. The minimum recommended height of people in pixels is ~120. This typically corresponds to 25 meters for Full HD cameras and 16 meters for Full HD.