Identifying and tracking objects for Law Enforcement

Find and identify objects in security footage

Although facial recognition is one of the most well-known AI analyzers on the market, due to featuring frequently in the news, TV shows or movies, object recognition can be as valuable as facial biometry and also has no limitations in its use as of yet, due to privacy protection laws, as object cannot identify a person uniquely.

Intelion can integrate almost any object recognition analyzer thanks to its agnostic AI Bus, and processes the results in a visual manner to provide a valuable investigative tool to police officers or intelligence employees, seeking specific objects in video footage from security or traffic cameras.

Object detection and recognition

Modern AI Analyzers, which can be integrated into Intelion to be part of the investigative workflow, include object detectors. The amount of objects between they can differentiate varies greatly depending on the vendor and the AI model used, but figures around 500 are usually reasonable. This means that the contents of any image of video can be analyzed and yield a list of all the visible objects in frame that the model knows.

For Law Enforcement this is an important help, as platforms like Intelion can sift through hundreds or thousands of videos or video streams, locating specific objects, like a backpack, a handbag, mobile phone, or even a watch or sunglasses. The precision of the detection depends greatly on the video resolution, but in today’s day and age, Full HD video is quite common, which means that the AI analyzers have a decent amount of image pixels to work with and recognize objects.

The list of detected objects also includes how many have been spotted in the videos or images and the timeline helps seeing the specific points in the video.

Not only objects

The thumbnail summary screen shows all the object detections at once (in this case backpacks).

In fact, the moniker “object detection” is a bit misleading, as the object detection module does way more, than just detect what can be called “objects”, as in “inanimate objects”. The truth is that a form of object detection is in many other AI analyzers. The face recognition, for example, first needs to know if there is a person in the frame. And that is work for the object detection. Then, it needs to know if the face is visible and, again, that task is for the object detection. Only when both criteria have been met, the actual face biometry can start working on the image to find the face features.

The same is true for ALPR systems (Automatic License Plate Reader). If a car is detected in the frame, the next step is to locate a number plate. If both conditions can be satisfied, the actual ALPR (which is a specialized OCR) can start working to extract the numbers and letters of the plate.

Privacy preservation and difficult camera angles

Object detection is not just a versatile tool for Law Enforcement investigations, but can also help in difficult situations to keep track of someone with a distinctive object (like a backpack), without infringing local privacy laws that may restrict the use of facial biometrics. Intelion can track objects through multiple camera recordings or feeds, thus following the progression of a specific individual through the city, for example.

And even if face recognition is allowed by law, or a corresponding court order has been secured, in some recordings, a due to difficult camera angles or positions, the face recognition AI model may not work or have enough data to guarantee a reliable identification. In these cases object detection can help out, providing additional means for tracking or identifying someone, until face detection is available again.

The detection list is s quick way to go through all the detections and watch the corresponding video segment.

Watch Intelion in action

See how Intelion locates and detects objects in videos and images.

In this short video Intelion shows how to locate different objects in security camera recordings and track them. With this it is possible to avoid biometric identification and, nonetheless, be able to follow a suspect through the recordings.

Conclusion

Object detection is used in many other AI analyzers as a first stage to discriminate the objects appearing in the image, before starting subsequent analytics that may look for faces, number plates, etc. It is also an additional means for tracking suspects if they have a distinctive object with them (like a bag). There are multiple object detection models on the market, but most of them are able to identify more than 500 different objects, and some can go into the thousands for specific cases. However, models are trained and evolved constantly, so these figures are bound to change.