top of page

Human Dectect

Human Detect System For Automatic RTG Cranes

HDS is advanced intelligent system for human detecting which will recognize human automatically in watching area of a variety of people's behavior and even very poor environment. 

* H/W Requirements;

  1. Two Camera at engine side

  2. Two Camera at e-rom side

  3. Two Camera at trolley

  4. One Industrial Switch(8Port)

  5. One HDS Processor

It learns mainly focus on upper body of a human
The system will be streng by repeat deep learning for following items

  • Various colored and various shaped clothes(White, black )

  •  Wearing Hats, helmets or nothing wear.

  • Various environmental conditions such as night work, rainy day,  cloudy day, sunset 

  • Various location and background 

  • Various human’s behavior

Advantages of vision system by Deep Learning based 

  • Can be detected just like humans even variety object and complex patterns 

  • Lesser impact on environmental factors (light, Location, size)

  • Auto correction error by re-learning for judgment error 

  • Getting robust system over the time 

Location of Cameras (Over whole) 

▲ Location of Cameras (Top View)  

The six cameras install at above location. 

※ Far distance as possible is good

▲ Location of Cameras (Engine side View)

Two cameras install at engine side

※ Far distance as possible is good.

Two cameras install at e-room side

※ Far distance as possible is good.

▲ Location of Cameras (E-Room side View) 


  1. Watching area is in black which defined by slot based (slot wide *2 alpha)

  2. Scanned area in blue

  3. Once detect human, HDS will report to CIU with in 3 sec as following format.
    Current slot number, Current Row number, HX Location, HY_Location

ROC & Distribution Chart 

Shows the normalized distribution of the deviation for bad and good images. Ideally, theses two curves should not be crossing. The green line represents the threshold below which a area will be considered as EMPTY. On the opposite the red line represents the threshold above which a area will be considered as HUMAN. In between those lines is a grey zone, in which the area will be considered as intermediate

ROC Chart

(Receiver Operating Characteristic) 

Indicate on this chart for all possible thresholds from output of image controller. Right side is positive true position(Human) and left side is negative true position(Empty) 

The AUC (Area Under Curve) value is summary of this curve and1.00 means that perfect detect human.

∅ processing time 73.4 ± 7.8 msec

Managing Watching Area 

Processed Image (ACU Value 96%) 

Processed Image (ACU Value 61%) 

Processed Image (ACU Value 04%) 

Processed Image (ACU Value 90%) 

Processed Image (ACU Value 15%) 

Processed Image (ACU Value 55%) 

Processed Image (ACU Value 98%) 

Processed Image (ACU Value 11%) 

Processed Image (ACU Value 99%) 

bottom of page