菜鸟看论文——Free Space Compute reviews

参考：

16.Identifying Good Training Data for Self-Supervised Free Space Estimation

12.Recovering free space of indoor scenes from a single image

02. Real time obstacle  detection in stereovision on non flat road geometry  through” v-disparity

15. Ground Segmentation and Occupancy Grid Generation Using  Probability Fields

In a typical traffic scenario, the following challenges and problems are faced in the analysis of the 3D environments: 1) detecting road surfaces and lane areas; 2) classification of potential obstacles above road surfaces and obtaining their 3D information; 3) classification of the roadside structures such as guardrails, lamp poles, and traffic signs.

The problem of estimating free space in structured and  static environments is usually solved by exploiting properties  of certain well defined structures. Two examples of  free space estimation solutions are that of Hedau et al. [2]  and Labayrade et al. [3]. While the first exploits the box  like geometry of furniture to estimate free space in indoor  scenes from a single camera image, the second uses the planar  geometry of a road and identifiable lane markings to estimate  the free space in urban road scenes. In unstructured  or unknown environments such as forest areas, the lack of  structure of the scene causes methods relying on static scene  properties to fail.

The slope of the road surface shown in a road image is  coupled with a lateral slope and a longitudinal slope in  terms of the road direction. Compared with the  longitudinal slope, the change in the lateral slope is small.  Accordingly, most algorithms that estimate the slope of  the road surface have typically estimated the longitudinal  slope [1,7]. This paper also deals with the estimation of

LoPORS. The estimation of the lateral profile of the road  surface remains a problem needing to be solved.

To account for the ever changing properties of free space  in unstructured scenes, it is natural to resort to learning  based systems, which usually require a training phase in  which training data representing free space is used as an  input to the learning algorithm.

Recent free space estimation approaches tackle this  problem through self-supervision, where one classifier directly  supervises input to a second classifier. The first classifier  uses data it is confident about to label parts of the environment  as free space; this data is then provided as input  to the second classifier that extends the labeling over  the whole environment. The proposed system in this paper  lies within this framework, allowing long range fully  autonomous free space estimation without relying on any  rigid assumptions such as a planar ground or bootstrapping  methods.

A variety of sensors and sensor combinations have previously  been employed for free space estimation.Apart from the being expensive and power hungry, radars offer a narrow field of view and low accuracy in lateral  direction [4]. LiDARS are also relatively expensive and big  consumers of energy. Range measurement can alternatively  be acquired using a stereo camera rig. The advantage of such  system is its relatively low cost and the added bandwidth of  information that a picture offers; however it suffers from  drawbacks such as the requirement for calibration between  the two cameras and the degradation of depth information as  the depth increase.Due to their high resolution and precision, the use of vision sensors as the primary sense to environment perception became popular over the last decade. In contrast with other active sensors such as lasers or radars, vision sensors are passive one and provide the richest information. Although being sensitive to weathe