https://sputnikglobe.com/20211025/scientists-offer-new-way-to-protect-pedestrians-1090189897.html
Scientists Offer New Way to Protect Pedestrians
Scientists Offer New Way to Protect Pedestrians
Sputnik International
A new way to detect unusual situations on pedestrian paths has been proposed by the scientists of South Ural State University (SUSU) as part of an... 25.10.2021, Sputnik International
2021-10-25T11:28+0000
2021-10-25T11:28+0000
2022-08-06T13:32+0000
russia
safety
science & tech
https://cdn1.img.sputnikglobe.com/img/07e5/0a/19/1090189600_0:309:3091:2048_1920x0_80_0_0_ba9d76da12168a18e6f639215f18c1fa.jpg
According to the researchers, the method would significantly improve the safety of pedestrians. The results of the study were published in the Safety Science journal.Every year, more than 270,000 pedestrians around the world are reported to be killed in road accidents. Today’s models developed from computer analysis of pedestrian movements, designed to ensure pedestrian safety, are based on a set of predefined attributes. According to the SUSU scientists, such a method is "unreliable" due to the problem of picture scaling.The researchers have developed a new method for predicting pedestrian collisions with other road users, such as car drivers, skateboarders, cyclists, and others on pedestrian paths. The new method is based on a high-accuracy neural network that analyses images received from video cameras in different city locations.According to Sachin Kumar, pre-processing allows the removal of redundant information, which improves the quality of images. As a result, it has become possible to classify an image correctly and to find the needed object in the image in the final stage.In the research process, the MRCNN neural network, integrated with DenseNet networks (a dense high-accuracy network for classifying images) was used by the scientists.The high efficiency of the new method has already been confirmed by various simulators. The researchers noted that road safety services can use their computer programs.The study was conducted by specialists from Russia, Saudi Arabia, and India.
Sputnik International
feedback@sputniknews.com
+74956456601
MIA „Rossiya Segodnya“
2021
Sputnik International
feedback@sputniknews.com
+74956456601
MIA „Rossiya Segodnya“
News
en_EN
Sputnik International
feedback@sputniknews.com
+74956456601
MIA „Rossiya Segodnya“
Sputnik International
feedback@sputniknews.com
+74956456601
MIA „Rossiya Segodnya“
safety, science & tech
Scientists Offer New Way to Protect Pedestrians
11:28 GMT 25.10.2021 (Updated: 13:32 GMT 06.08.2022) A new way to detect unusual situations on pedestrian paths has been proposed by the scientists of South Ural State University (SUSU) as part of an international team.
According to the researchers, the method would significantly improve the safety of pedestrians. The results of the study were published in
the Safety Science journal.
Every year, more than 270,000 pedestrians around the world are reported to be killed in road accidents. Today’s models developed from computer analysis of pedestrian movements, designed to ensure pedestrian safety, are based on a set of predefined attributes. According to the SUSU scientists, such a method is "unreliable" due to the problem of picture scaling.
The researchers have developed a new method for predicting pedestrian collisions with other road users, such as car drivers, skateboarders, cyclists, and others on pedestrian paths. The new method is based on a high-accuracy neural network that analyses images received from video cameras in different city locations.
"A CNN model was developed to extract important features and detect anomalies or undesirable objects (i.e. bicyclists) from images captured by surveillance cameras at pedestrian's pathways. Before applying the CNN model for the analysis of images, pre-processing must be done to enhance the quality of images and to remove redundant information in order the improve the analysis time and accuracy of the object detection", senior researcher at the Department of Data Mining and Virtualisation at SUSU Sachin Kumar explained.
According to Sachin Kumar, pre-processing allows the removal of redundant information, which improves the quality of images. As a result, it has become possible to classify an image correctly and to find the needed object in the image in the final stage.
In the research process, the MRCNN neural network, integrated with DenseNet networks (a dense high-accuracy network for classifying images) was used by the scientists.
The high efficiency of the new method has already been confirmed by various simulators. The researchers noted that road safety services can use their computer programs.
The study was conducted by specialists from Russia, Saudi Arabia, and India.
