offlinemot.objects_classes.TrafficObj

class offlinemot.objects_classes.TrafficObj(frame, frame_id, box, track_id, config=<config.configs object>, tracker=<built-in function TrackerKCF_create>, class_id=-1, detection_way=1, detect_prob=0.0)

A class used to represent any traffic entity, with its full track across frame until it gets lost.

…

Variables

• ~TrafficObj.box (list) – a box is list of [x,y,w,h] where x,y is the top-left point coordinates and w,h are the width and height of the bounding box unrotated.

• ~TrafficObj.boxes (list) – a list of box lists for every time step saved in time_steps

• ~TrafficObj.true_wh_max (tuple) – the width and height when the object is moving horizontally, vertically or the nearst to these directions

• ~TrafficObj.tracking_state (list) – the list of the tracking success variable

• ~TrafficObj.time_steps (list) – list of frames where object is tested for detection or tracking.

• ~TrafficObj.trust_level (list) – list of three values rows where each value refer to a boolean, indicating the sucess of detection, tracking and background subtraction respectively

• ~TrafficObj.tracker_class (Tracker function) – The function that will build the tracker object (default: TrackerKCF_create)

• ~TrafficObj.tracker (Tracker object) – the tracker object that will perform the tracking

• ~TrafficObj.class_id (int) – an integer representing the type of object whether it is -1: unknown (default), 1 pedistrains, 2 cyclist, 3 cars.

• ~TrafficObj.colors_map (list) – what color to draw each traffic object

• ~TrafficObj.track_id (int) – a unique id assigned to each traffic object

• ~TrafficObj.last_detect_prob (float) – the probabilty of the last detection. it is needed to solve overlaping conflict (default:0)

• ~TrafficObj.img_wh (tuple) – the width and height of the frame

• ~TrafficObj.angels (list) – a placeholder list to calculate the angle of direction at the end for the object.

• ~TrafficObj.centers (list) – a placeholder list to calculate the centers of the boxes at the end for the object.

find_center()

Calculate the current center for the box. the output is integers

track(new_frame, frame_id)

Track the object in the new frame, save the result, and update the true size

re_init_tracker(frame)

start a new tracker object in the frame provided with the current box

draw(new_frame=numpy array)

Draw the current box position with color code and track ID on the frame

update()

Test the object if still need to be tracked, or it is lost so if it needs to be deleted

filter_by_detections_dist(detections, check=False)

It assigns the object to one of the detections in the frame if a minimum distance and detection probability is found.

filter_by_bg_objs(bg_objs)

check if the object is within a minimum distance to any moving area.

set_detection(ok)

save the result of the current detection.

set_bg_substract(ok)

save the result of the current background subtraction.

set_track_id(id_)

set track id for the object once it has a detected class

find_true_size(new_box)

Test evey box of the current position whether it moves in the horizontal or vertical direction or the closest to that and save the current size if so.

get_detection_format()

Get the bounding box in (top-left point, bottom-right) point format and add class_id and dummy probabilty.