08-09-2014, 10:03 AM
An Intelligent Automated Door Control System Based on a Smart Camera
Door Control System.pdf (Size: 429.93 KB / Downloads: 21)
Abstract
This paper presents an innovative access control system, based on human detection and path analysis, to
reduce false automatic door system actions while increasing the added values for security applications. The
proposed system can first identify a person from the scene, and track his trajectory to predict his intention for
accessing the entrance, and finally activate the door accordingly. The experimental results show that the
proposed system has the advantages of high precision, safety, reliability, and can be responsive to demands,
while preserving the benefits of being low cost and high added value.
Introduction
Automatic entrance/exit door control is widely used in public places such as grocery stores, businesses,
transportation stations, airports, and wholesale department stores to eliminate the need of manually opening
and closing actions. Contemporary sensor-based automatic door control technologies include infrared,
ultrasonic/radio, or other wireless sensing methods. The first can be further divided into active and passive
approaches. The active process emits infrared signals from the controller and captures the reflected signals to
determine if there is any object close to the door. This approach is accurate and capable of identifying the
position and the speed of the object but its high cost has made it less popular. The passive approach detects the
infrared signals radiated by people and is the most widely used for being simple, effective, and low cost. The
ultrasonic/radio approach, on the other hand, emits ultrasonic or radio waves to scan the environment and
analyzes the returned signals for door access control.
Although these techniques are all successful in detecting objects, they are not capable of understanding the
type and the intention of the objects. For instance, a puppy or a passing pedestrian may accidentally trigger
the door and cause a false opening action. Frequent false action is not only annoying, and results in air
conditioning energy waste, but also reduces equipment lifetime. This calls for the need of an automatic door
control system based on the detection and intention analysis of people
Design Concept and Principal Theory
The proposed innovative door control system is mainly based on human detection and intention analysis. The
first part involves face detection or contour detection that identifies whether the detected object is a person,
while the latter includes trajectory tracking and statistical analysis for intention estimation. A flowchart of the
control procedures is shown in Figure 1.
Figure 1.
System flowchart.
It is known that face detection with reasonable detection rates had been well developed in the literature and
has the closest relevance to human characteristics, thus is good for people identification. In short, once a
person is detected in the region of interest (ROI), the trajectory of his face can be tracked and then analyzed
by a statistical analyzer to calculate the corresponding cumulative probability, and will be used as the
estimation of the intention.
On the other hand, in the door-closing procedure of Figure 1, the door is closed if no object is detected in the
passage on both sides of the door. Overall, processes of the state transition, human detection, intention
analysis, and the theoretical performance evaluation are described in the following subsections.
Human Detection
In recent years, human detection techniques, especially those implemented by face detection strategies, have
been successfully applied to many consumer products such as digital cameras, smart phones, or surveillance
systems for detecting people [1,2]. In the proposed system, although both face detection and contour
detection are adopted for human identification, the face detection is used as the primary solution and the
contour detection based on head and shoulder shape, similar to Reference [3], is left as the optional solution
for the application which has special privacy considerations. Thus, human detection function is represented
Experiment Results and Discussion
The proposed system can determine the user's intention and provide access to a door by detecting his position
and tracking his trajectories of movement in the image sequence. To implement the system, a DSP based
multimedia module, TI DM368, is selected as the hardware platform. The algorithms developed were
programmed in C and executed under the Linux platform. The specifications of hardware and software are as
follows: an ARM9 CPU with a speed of 900 MHz with embedded processing algorithm is used. The system
prototype and the DSP platform board are shown in Figure 5(a,b). The image sensor resolution is 500 M and
the test image size is 320 × 240. The other installation parameters are described as follows: the camera is
installed on a wall at 2.2 m high and 30° tilt. The size of ROI is 3.11 m (width) and 4.11 m (distance to door)
respectively, as shown in Figure 5©. The width of door is 1.6 m. Thus, according to Equation 2(a,b), σ of s
and λ of s can be calculated as σ = 0.26 and λ = 4.11, respectively. The infrared signal scanning in ROI is
indicated by the red area, as shown in Figure 5(d). The scanning width and distance are 3.11 m and 1.28 m
(red line to door), respectively.
Conclusions and Future Works
In this paper, an automatic door control system is implemented on a DSP platform. The system can first
identify a target as person by face detection, and then analyze the path trajectory to determine whether the
person has intention to access the door or not, thus to control the door accordingly. It is noted that the system
has advantages of low false rate (near 0%), high correct activating rate (99.6%), and short response time
(within 2 s) from detecting the target, confirming his intention, to activate the door opening. Moreover, via
statistical analysis on detected face in consecutive time sequence, case of passing persons with missing face
can still be confirmed within 4 s thus to activate the door correctly, as shown in Table 1.
To sum up, the proposed method builds up the statistic model of moving trajectories in ROI first, and the
corresponding probability of a face at certain location can be obtained by lookup table. If the average
probability of face trajectory is greater than th within T period, the person is said to have intention to go
through the entrance. That is, region within dotted line as shown in Figure 4(d), whose pdf is greater than
th , is viewed as key region while determining one's intention. If the face trajectory locates mostly within
the key region, the person is said to have intention to go through the entrance. According to Equation (6), if
th is set to be 0.8 (or 0.2) and T is set by 2 s, the FR will become 0.18 (or 6.5 × 10 ). The response time
to confirm one's intention will increase when th increase, while lower th value will result more false
opening actions. It is observed, by our experiments, that if th is set to be the average of ROI, said μ , the
FR is about 6.3 × 10 and is suitable for most application cases.