Embedded Files
2021 - SISCAM, Software Tool to Automatically Detect and Count Motorcycles using Computer Vision
SISCAM allows detecting and automatically counting motorbikes in an urban road in day-light time using computer vision. This software tool was designed to store all information generated in a SQL database. In addition, the software tool has a GUI to ease the operation by users; the software tool has a configuration GUI where users can perform tasks such as login, introducing the location of data acquisition, selecting the source of data (IP cameras, video or a set of consecutive images), selecting the region of interest, and defining the traffic flow direction; also, the software tool has a GUI to observe the process of detection and motorbike counting; finally, the software tools allows generating a report using the data stored in the database according with a search criteria.
It is worth noting that motorbikes have dark colors which are very similar to most paved roads. In addition, motorbikes are relatively small in comparison with other vehicles, causing in this case that motorbikes are occluded by cars or bigger vehicles. In this work, the motorbike detection was performed using YOLOV5 architecture achieving a mAP@50 of 99 %. Afterwards, the detected motorbike must be counted, in this work a novel proposal for counting vehicles was implemented. Results suggest a mean accuracy of 91% when counting motorbikes in unstructured environments and in day-light time. Actually, SISCAM is a registered software legally recognized by the Interior Ministry of Colombia. This document can be downloaded here.
When SISCAM starts, a user identification is needed. It will ask the following information:
User
Password.
In case the user is not registered, the user registration will be performed fulfilling the following data:
Full name
Email.
Password.
Then, the configuration module will ask about the following location information where the measurement will be performed:
State
City
Address
Also, SISCAM will ask about the source of data:
IP camera: here, the IP number of the camera will be asked, also the username and password.
Files into the HD. Here, the full path where the files are placed must be selected in a dialog.
In addition, the following calibration parameters should be introduced:
Principal point.
X focus
Y focus
Distortion constants.
Afterwards, SISCAM will need to load the Tensor Flow Lite file where the deep neuronal network is. This neuronal network must be trained in advance, and export the result into a FLT file. Then, SISCAM will perform the feed-forward process to detect and identify the motorcycles.Depending of the camera point of view, the ROI must be selected using the following procedure:
A pre-visualization is shown to manually select the ROI.
The user will put 4 points into the image to define the ROI
Using the rectangle draw by the user, a line is also draw by the user to define the counting threshold.
Within the same rectangle of the ROI, the direction of the traffic flow is also defined with two points clicked by the user.
In case of error, the ROI, the transversal line and the direction of the traffic flow can be deleted and start again.
Also a confirmation button is present to validate the information.
Once the configuration parameters are set, SISCAM can be operated using the source of information selected. To do so, the user can perform the following actions:
Change and edit the name of the session.
Press the play button to start the image processing.
Press the pause button to pause the image processing.
Enable or disable the ROI
Enable or disable the vehicle detection information.
Enable or disable the tracking information.
Press the stop button to stop the image processing.
Finally, once the session is over. Users can generate a report. In this module, users can perform the following actions:
Selecting the state where the session took place.
Selecting the city where the session took place.
Selecting the date of interest.
Selecting the start and final hour of interest.
Selecting the time unit of the traffic flow.
Changing the report title.
Changing the report description.
Selecting if the document will be saved into the database or in a hard drive file.
This software was developed for industrial service and academic use. If you like to perform a field test, please contact me:
Prof. Bladimir Bacca Cortes Ph.D.
Address: Cra. 100, Street 13, Universidad del Valle, Melendez, Building 354, Office 2006.
Tel: +5723212100 Ext. 7656
Page updated
Report abuse