Traffic Sign Detection And Recognition In Ethiopia Street Scene

Traffic sign detection is one of the most studied subjects in the Computer Vision field.Given an arbitrary image or video frame,the goal of traffic sign detection is to determine whether there is any traffic sign in the image and,if present,return the image location and the extent of each traffic sign.The high degree of variability and traffic sign existence in the uncontrolled outdoor environment makes it an object very difficult to detect,mainly in complex environments.Recently,deep learning approaches started to be applied for Computer Vision tasks with great results.They opened new research possibilities in different applications which including traffic sign detection.Traffic sign detection and recognition is one of the important applications in computer vision.Traffic signs provide drivers important information and help them to drive more safely and more easily by guiding and warning them and thus regulating their actions.In automatic traffic sign maintenance and in a visual driver assistance system,traffic sign detection and recognition are two of the most important functions.For human and other living animals,visual perception is one of the most important senses to detect,recognize and interact with an object within its environment.Despite the fact that,there are many challenges that make hardening the detection and reorganization of the object in the environment.Some of them are various weather conditions(e.g.light intensity,shadow,distance,etc.)and viewpoint.However,in the recent decades,the advancements in the technology like Artificial Intelligence have made machines to interact with human beings and it becomes an integral part of our everyday lives by getting man-like abilities,sometimes more in a sophisticated way to detect and recognize objects that a human would recognize.Whereas,in the real world it is very hard to identify which pieces go together as parts of the same object since some parts of an object can be hidden behind other objects often to see the whole object and the major challenge to separate objects one from others.Those challenges are the main reasons that enforce the invention of computer vision that works with the images.While reproducing a part of human visual perception abilities to processing the image content by combining with several different algorithms is a major challenge in computer vision.To address the problems,traffic signs with distinct shapes such as circles,triangles,rectangles,and octagons as well as composed of basic colors were designed with highly saturated properties and reflective attributes in the background that helps to detect and recognize target in a varied weather condition.Indeed,the researchers conduct many experimental types of research that combine several different algorithms with system run in the real time to reproduce a part of human visual perception abilities and prevent different traffic accidents and damage.Even though the current situation in the case of Ethiopia is far from these realities,that is why many more traffic accidents and damage occur.To the bests of our knowledge,no works done to address the problems using computer visions by using traffic sign dataset.These and other problems impel needs to study deep learning and practiced the basic network model of deep learning to build the local dataset traffic sign detection and recognition in Ethiopia,train the SSD model to get the detection result.To this end,training dataset was collected from Addis Ababa,Ethiopia on a wide variety of roads and in a diverse set of lighting and weather conditions.The collected dataset was classified into six classes of traffic signs such as crossing sign(496),stop sign(322),stopping sign(660),U-turn sign(157),Waiting sign(254),and parking sign(505).A total of 2394 traffic sign images was collected and 90%of the traffic sign images used for training and reaming 10%of traffic sign image were used for testing using the SSD model.One of deep learning model among others choice because SSD network can carry out the end to end object detection directly,the given image and its label can be trained,and the calculation is small,the result involves multi-scale,the detection efficiency is good and the speed is fast.The experimental result with an accuracy of 78.4%shows that the validity of the SSD model in testing the traffic sign detection in the context of Ethiopian road traffic sign.The experimental results reveal Ethiopia traffic sign data set constructed can meet the needs of the current model training test,but it still needs to be expanded.Even though,the traffic sign detection model based on the deep learning SSD solves the problem of traffic sign detection in Ethiopia to a certain extent.But there is still room for further improvement.The model can be replaced by a more accurate and efficient model.Since the deep learning model is based on data-driven,and the data amount is too little,it will be considered in my future work to expand this dataset.Overall,the study has confirmed that SSD model can be used to detect traffic signs.Besides,all the experiments gave valid results and can be used for traffic sign detection.