LiDAR-based detection, tracking, and property estimation: A contemporary review

Abstract

Object detection, Person tracking, and Person property estimation (PPE) are identical innovation areas trying to improve their accuracy in different parameters to fit various real applications. For many years, so much research has been done in these fields. Many scientists also used many more techniques and algorithms. But most of the innovations were deeply based on image-based analysis, where cameras were the critical components of data acquisition. Over the years, new technologies arrived, and different types of research are happening. Rather than cameras, some other sensors, like infrared, depth cameras, and very recently LiDAR sensors, are used to estimate person properties, track them, as well as to detect them. Especially, height, age, gender, region, etc., parameters can be measured as person property. Eventually, 3D object detection by LiDAR will be a state-of-the-art research field with the advent of autonomous driving initiations. We studied many articles and found enthusiastic outcomes with these sensor setups to understand contemporary technology and its efficacy. We categorized these research articles into video camera-based studies and other sensor-based studies. So many surveys have been done on video-based analysis, even with deep learning techniques. Another sensor-based research is very recent, and we do not get enough study on it. We thought to summarize these studies in a survey article, especially LiDAR-based analysis. This article covered most of the recent possible sensor-based studies of detection, person tracking and property estimation except cameras (all, RGB, RGB-D, etc.) based learning.