CNN-based 3D object classification using Hough space of LiDAR point clouds

Song, Wei; Zhang, Lingfeng; Tian, Yifei; Fong, Simon; Gozho, Amanda

doi:10.1186/s13673-020-00228-8

Cited by 35 publications

(28 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It achieved a substantial improvement over ZFNet, which was the winner in 2013 [65]. In another study, Wei et al proposed a CNN based 3D object classification model and achieved an accuracy of up to 93.3% [66]. The network architecture of GoogLeNet differs from other models because a convolution layer of 1 × 1 kernels with ReLU activation is used in the middle of the model.…”

Section: Googlenet (Inception Model)mentioning

confidence: 99%

Deep learning scheme for character prediction with position-free touch screen-based Braille input method

Shokat

Riaz

Rizvi

et al. 2020

Hum. Cent. Comput. Inf. Sci.

View full text Add to dashboard Cite

Smart devices are effective in helping people with impairments, overcome their disabilities, and improve their living standards. Braille is a popular method used for communication by visually impaired people. Touch screen smart devices can be used to take Braille input and instantaneously convert it into a natural language. Most of these schemes require location-specific input that is difficult for visually impaired users. In this study, a position-free accessible touchscreen-based Braille input algorithm is designed and implemented for visually impaired people. It aims to place the least burden on the user, who is only required to tap those dots that are needed for a specific character. The user has input English Braille Grade 1 data (a–z) using a newly designed application. A total dataset comprised of 1258 images was collected. The classification was performed using deep learning techniques, out of which 70%–30% was used for training and validation purposes. The proposed method was thoroughly evaluated on a dataset collected from visually impaired people using Deep Learning (DL) techniques. The results obtained from deep learning techniques are compared with classical machine learning techniques like Naïve Bayes (NB), Decision Trees (DT), SVM, and KNN. We divided the multi-class into two categories, i.e., Category-A (a–m) and Category-B (n–z). The performance was evaluated using Sensitivity, Specificity, Positive Predicted Value (PPV), Negative Predicted Value (NPV), False Positive Rate (FPV), Total Accuracy (TA), and Area under the Curve (AUC). GoogLeNet Model, followed by the Sequential model, SVM, DT, KNN, and NB achieved the highest performance. The results prove that the proposed Braille input method for touch screen devices is more effective and that the deep learning method can predict the user's input with high accuracy.

show abstract

Section: Googlenet (Inception Model)mentioning

confidence: 99%

Deep learning scheme for character prediction with position-free touch screen-based Braille input method

Shokat

Riaz

Rizvi

et al. 2020

Hum. Cent. Comput. Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…The method developed by Cui et al [20] was used for vehicle classification before detecting the queue length. Song et al [23] developed a CNN-based 3D object classification using Hough space of LiDAR point clouds. Premebida et al [24] developed a Gaussian Mixture Model classifier to distinguish vehicles and pedestrians from the LiDAR data.…”

Section: Related Workmentioning

confidence: 99%

Object Classification with Roadside LiDAR Data Using a Probabilistic Neural Network

Zhang

Ma³

et al. 2021

Electronics

View full text Add to dashboard Cite

Object classification is important information for different transportation areas. This research developed a probabilistic neural network (PNN) classifier for object classification using roadside Light Detection and Ranging (LiDAR). The objective was to classify the road user on the urban road into one of four classes: Pedestrian, bicycle, passenger car, and truck. Five features calculated from the point cloud generated from the roadside LiDAR were selected to represent the difference between different classes. A total of 2736 records (2062 records for training, and 674 records for testing) were manually marked for training and testing the PNN algorithm. The data were collected at three different sites representing different scenarios. The performance of the classification was evaluated by comparing the result of the PNN with those of the support vector machine (SVM) and the random forest (RF). The comparison results showed that the PNN can provide the results of classification with the highest accuracy among the three investigated methods. The overall accuracy of the PNN for object classification was 97.6% using the testing database. The errors in the classification results were also diagnosed. Discussions about the direction of future studies were also provided at the end of this paper.

show abstract

“…PointNet++ (Qi et al, 2017b) can process more fine-grained patterns and complex scenes. (Song et al, 2020) explains how points of segmented objects can be identified using CNN's and Hough transformations. Hough transformations are a commonly used used feature extraction method in the 2D computer vision domain (Cantoni and Mattia, 2013).…”

Section: Related Workmentioning

confidence: 99%

Automatic Modelling of 3d Trees Using Aerial Lidar Point Cloud Data and Deep Learning

Kippers

Moth

Elberink

2021

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. 3D tree objects can be used in various applications, like estimation of physiological equivalent temperature (PET). During this project, a method is designed to extract 3D tree objects from a country-wide point cloud. To apply this method on large scale, the algorithm needs to be efficient. Extraction of trees is done in two steps: point-wise classification using the PointNet deep learning network, and Watershed segmentation to split points into individual trees. After that, 3D tree models are made. The method is evaluated on 3 areas, a park, city center and housing block in the city of Deventer, the Netherlands. This resulted into an average accuracy of 92% and a F1-score of 0.96.

show abstract

CNN-based 3D object classification using Hough space of LiDAR point clouds

Cited by 35 publications

References 24 publications

Deep learning scheme for character prediction with position-free touch screen-based Braille input method

Deep learning scheme for character prediction with position-free touch screen-based Braille input method

Object Classification with Roadside LiDAR Data Using a Probabilistic Neural Network

Automatic Modelling of 3d Trees Using Aerial Lidar Point Cloud Data and Deep Learning

Contact Info

Product

Resources

About