Method for detection of an object
The present invention relates to a real time method for detection of an object that enables to define, by means of a thermal camera, objects that are in the possession of people. The present invention particularly relates to a method that enables the detection of objects that are in the possession of people, through different deep learning methods that are subbranches of artificial intelligence using thermal camera images, wherein the images obtained via thermal cameras are processed real time and input into two different deep learning models.
1 . A computer-implemented method of detecting objects concealed on a person using thermal camera images, the computer-implemented method comprising:
obtaining a single thermal image and performing pre-processing on the single thermal image by the computer;
processing the pre-processed single thermal image by the computer, using a weapon detection model that automatically detects whether a weapon is present or absent in the pre-processed single thermal image, without user supervision, human detection or motion detection, wherein the weapon detection model is an artificial intelligence VGG-16 model that is a convolutional neural network (CNN) based on a deep learning model, with added layers, and does not include motion detection or human detection;
when the weapon detection model automatically detects that a weapon is present in the pre-processed single thermal image, asking a preference of displaying or not displaying the location of the detected weapon to a user;
when the weapon detection model automatically detects that a weapon is present in the pre-processed single thermal image and the user preference requests that the location of the detected weapon be displayed, performing additional processing on the pre-processed single thermal image with the weapon present by the computer using a location determination model, wherein the location determination model is different and separate from the object detection model and wherein the location determination model is an artificial intelligence Yolo (You look only once) CNN model modified by fine tuning for location detection, and displaying on a screen to the user, the location of the detected weapon in the pre-processed single thermal image processed by the location determination model;
when the weapon detection model automatically detects that a weapon is present in the pre-processed single thermal image and the user preference does not request that the location of the weapon be displayed, displaying a warning on the screen that a weapon is present in the processed image and do not perform processing on the pre-processed single thermal image using the weapon location model;
when the weapon detection model automatically detects that a weapon is present in the pre-processed single thermal image, recording a confidence index value via the pre-processed single thermal image that has been processed, comparing the confidence index value to a threshold value by the computer, generating an alarm by the computer when the confidence index value is higher than the threshold value, and when the confidence index value is lower than the threshold value an image is waited from the thermal camera to process the next image; and
when the weapon detection model does not detect a weapon in the single thermal image obtained by the computer, the computer waits for a new image from the thermal camera.
2 . The method of claim 1 , wherein the confidence index is recorded to a database that resides on the computer.
3 . The method of claim 1 , wherein the confidence index is recorded in a database that resides on a remote server.
4 . The method of claim 1 , wherein the threshold value has been recorded in a database prior to the comparison with the confidence index value.
5 . The method of claim 1 , wherein the computer comprises a processor, a screen, a camera, and a graphic processing unit.
6 . The method of claim 1 , wherein the computer is a mobile device comprising a phone, a tablet, a smart watch, or glasses.