Academia.eduAcademia.edu

Outline

Title
Abstract
FAQs
Figures
Introduction
Background
Experiments and Results
Our Dataset
CALTECH-101 Dataset
Conclusion
References
All Topics
Computer Science
Computer Vision

2d Image Features Detector and Descriptor Selection Expert System

Profile image of Ibon MerinoIbon Merino

2019, 8th International Conference on Natural Language Processing (NLP 2019)

https://doi.org/10.5121/CSIT.2019.91206
visibility

…

description

11 pages

link

1 file

Abstract

Detection and description of keypoints from an image is a well-studied problem in Computer Vision. Some methods like SIFT, SURF or ORB are computationally really efficient. This paper proposes a solution for a particular case study on object recognition of industrial parts based on hierarchical classification. Reducing the number of instances leads to better performance, indeed, that is what the use of the hierarchical classification is looking for. We demonstrate that this method performs better than using just one method like ORB, SIFT or FREAK, despite being fairly slower.

FAQs

sparkles

AI

What key benefits does the proposed expert system provide for industrial object recognition?add

The paper demonstrates that the expert system significantly enhances recognition accuracy, achieving F1 scores of 0.94 compared to 0.845 using traditional methods. This improvement is attributed to hierarchical classification and tailored recognition pipelines for similar object groups.

How does the proposed hierarchical recognition method improve over classical feature-based methods?add

The hierarchical method shows superior performance with an average F1 score increase, achieving 0.94 on the proprietary dataset versus 0.843 on the Caltech-101 dataset. This approach effectively clusters objects by recognition behavior, optimizing pipeline selection.

What evaluation technique was employed to assess the performance of recognition pipelines?add

The study utilized Leave-One-Out Cross-Validation (LOOCV), conducting |X| iterations to generate a confusion matrix for performance assessment. This method allowed for a detailed score calculation using F1 metrics for each instance.

Which industrial recognition methods were compared against the proposed expert system?add

The study compared classical methods such as SIFT, SURF, and ORB within diverse recognition pipelines. Notably, the expert system outperformed ORB, which achieved an F1 score of 0.845, demonstrating superior efficiency in similar industrial contexts.

What impact does varying the number of views per instance have on recognition accuracy?add

The findings indicate that increasing the number of views leads to enhanced performance, particularly benefits observed when comparing from 10 to 50 views per instance. Higher view counts systematically contribute to better F1 scores across tested subsets.

Figures (9)
Computer Science & Information Technology (CS & IT) need to be recognized (test). Different kind of methods could be used to match features, but, mainly, distance based techniques are used. This techniques make use of different distances (L2, hamming.,...) to find the closest feature to the one that needs to be labeled. Those two features (the test feature and the closest to this one) are considered a match. In order to discard ambiguous features, we use the Lowe’s ratio test [14] to define whether two features are a ’good match’. Assuming f; is the feature to be recognized, and f;; and fj2 its two closest features from the model, then (ft, fj1) is a good match if:
Computer Science & Information Technology (CS & IT) need to be recognized (test). Different kind of methods could be used to match features, but, mainly, distance based techniques are used. This techniques make use of different distances (L2, hamming.,...) to find the closest feature to the one that needs to be labeled. Those two features (the test feature and the closest to this one) are considered a match. In order to discard ambiguous features, we use the Lowe’s ratio test [14] to define whether two features are a ’good match’. Assuming f; is the feature to be recognized, and f;; and fj2 its two closest features from the model, then (ft, fj1) is a good match if:
Table 1: Example of a confusion matrix for 3 instances. precision and the recall. The mean of all the F;’s, F, (5) is used for calculating ~)**.
Table 1: Example of a confusion matrix for 3 instances. precision and the recall. The mean of all the F;’s, F, (5) is used for calculating ~)**.
The first step of the hierarchical recognition is to recognize the typology with the ~)**. Given the typology t¢ as the typology predicted, the w/ is used to recognize the instance y of the object. We call the hierarchical recognition Y. The Figure 2 shows an scheme of the hierarchical recognition for clarification.
The first step of the hierarchical recognition is to recognize the typology with the ~)**. Given the typology t¢ as the typology predicted, the w/ is used to recognize the instance y of the object. We call the hierarchical recognition Y. The Figure 2 shows an scheme of the hierarchical recognition for clarification.
Table 2: Pipelines composition.
Table 2: Pipelines composition.
Figure 3: Parts used in our dataset.
Figure 3: Parts used in our dataset.
Table 4: Time in seconds that needs each pipeline in recognize a piece.
Table 4: Time in seconds that needs each pipeline in recognize a piece.
Table 5: F\’s of the ~**’s for each test (Caltech-101). p stands for number of parts and ¢ for number of pictures per part.
Table 5: F\’s of the ~**’s for each test (Caltech-101). p stands for number of parts and ¢ for number of pictures per part.
Figure 5: 6 random examples of images from the Caltech-101 dataset. The classes are: Face, Leopard, Motorbike, Airplane, Accordion and Anchor.
Figure 5: 6 random examples of images from the Caltech-101 dataset. The classes are: Face, Leopard, Motorbike, Airplane, Accordion and Anchor.

Related papers

Hierarchical object recognition system in machine vision
Andrey N Kokoulin

Journal of Physics: Conference Series, 2019

Authors propose an efficient architecture of the hierarchical machine vision system and the multimedia transfer algorithm for a distributed client-server system which can adapt to mobile network speed variations. This algorithm is based on the wavelet transform of an input multimedia source and allows for data exchange using all the available bandwidth of the unstable mobile network. The main principle of the hierarchical recognition system is the distributed processing network utilizing the “coarse-fine” paradigm in computer vision. Each source of a video stream is processed in-place by the tiny SoC computer which acts as an Edge Computing Unit and detects the presence of an object fragment in a video frame and crops the bounds of the ROI. The resulting stream containing ROI is directed to the main server if the object is detected, reducing traffic and resource consumption.

View PDFchevron_right
An efficient object recognition scheme for a prototype component inspection
Thompson Sarkodie-Gyan

Mechatronics, 1997

An object recognition scheme for inspection of high tolerances in manufactured engine components has been developed and tested. A mechano-optical arrangement is used to generate and capture images of the components. A simple edge detection algorithm and an effective feature transformation plus indexing technique are applied to extract and analyse the features of the images. A classification scheme based on fuzzy membership functions is proposed. Some promising results of feature extraction, feature analysis and classification are shown.

View PDFchevron_right
Local Detectors and Descriptors for Object Class Recognition
Mohiy Hadhoud

International Journal of Intelligent Systems and Applications, 2015

Local feature detection and description are widely used for object recognition such as augmented reality applications. There have been a number of evaluations and comparisons between feature detectors and descriptors and between their different implementations. Those evaluations are carried out on random sets of image structures. However, feature detectors and descriptors respond differently depending on the image structure. In this paper, we evaluate the overall performance of the most efficient detectors and descriptors in terms of speed and efficiency. The evaluation is carried out on a set of images of different object classes and structures with different geometric and photometric deformations. This evaluation would be useful for detecting the most suitable detector and descriptor for a particular object recognition application. Moreover, multi-object applications such as digilog books could change the detector and descriptor used based on the current object. From the results, it has been observed that some detectors perform better with certain object classes. Differences in performance of the descriptors vary with different image structures.

View PDFchevron_right
Comparative analysis of feature descriptors and classifiers for real-time object detection
IJRES Team

International Journal of Reconfigurable and Embedded Systems (IJRES), 2025

Detecting objects within complex environments, such as urban settings, holds significant importance across various applications, including driver assistance systems, traffic monitoring, and obstacle detection systems. Particularly crucial for these applications is the accurate differentiation between cars and roads. This study introduces a novel approach that leverages traditional feature descriptors and classifiers for real-time object detection. It conducts an exhaustive comparative analysis of feature descriptors and classifiers to identify the most effective model for real-time object detection. Handcrafted features of images are extracted using algorithms such as scale invariant feature transform (SIFT), oriented fast and brief (ORB), fast retina key-point (FREAK), and local binary pattern (LBP). Seven classifiers are employed, including support vector machine (SVM), K-nearest neighbors (KNN), random forest (RF), decision tree (DT), logistic regression (LR), Naive Bayes, and extreme gradient boosting (XGBoost). The performance of the 28 generated combinations of feature descriptors and classifiers is evaluated based on the parameters of accuracy, precision, F1 score, and recall. The model utilizing LBP and XGBoost achieves the highest accuracy, reaching 83.59%. The system architecture comprises a camera, a high-speed computing unit, a display, and an audio subsystem, with the algorithm implemented on a Raspberry Pi 4B (8 GB).

View PDFchevron_right
HYPER: A New Approach for the Recognition and Positioning of Two-Dimensional Objects
Olivier Faugeras

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000

A new method has been designed to identify and locate objects lying on a flat surface. The merit of the approach is to provide strong robustness to partial occlusions (due for instance to uneven lighting conditions, shadows, highlights, touching and overlapping objects) thanks to a local and compact description of the objects boundaries and to a new fast recognition method involving generation and recursive evaluation of hypotheses named HYPER (HYpotheses Predicted and Evaluated Recursively). The method has been integrated within a vision system coupled to an indutrial robot arm, to provide automatk picking and repositioning of partially overlapping industrial parts.

View PDFchevron_right
Hierarchical object geometric categorization and appearance classification for mobile manipulation
Andreas Holzbach

2010 10th IEEE-RAS International Conference on Humanoid Robots, 2010

In this paper we present a comprehensive object categorization and classification system, of great importance for mobile manipulation applications in indoor environments. In detail, we tackle the problem of recognizing everyday objects that are useful for a personal robotic assistant in fulfilling its tasks, using a hierarchical multi-modal 3D-2D processing and classification system. The acquired 3D data is used to estimate geometric labels (plane, cylinder, edge, rim, sphere) at each voxel cell using the Radius-based Surface Descriptor (RSD). Then, we propose the use of a Global RSD feature (GRSD) to categorize point clusters that are geometrically identical into one of the object categories (bowl, box, plate, mug, pan, etc). Once a geometric category and a 3-DOF position is obtained for each object cluster, we extract the region of interest in the camera image and compute a 2D SURF feature vector for each patch. We obtain the orientation around the up-right axis, and the exact object instance from the appearance. The resultant system provides a hierarchical categorization of objects into basic classes from their geometry and identifies objects and their poses based on their appearance, with near real-time performance. We validate our approach on an extensive database of objects that we acquired using real sensing devices, and show classification of 98.63% for 3D object category recognition and close to 98% for visual feature classification.

View PDFchevron_right
A Comparative Study of Local Descriptors for Object Category Recognition: SIFT vs HMAX
Plinio Moreno

2007

In this paper we evaluate the performance of the two most successful state-of-the-art descriptors, applied to the task of visual object detection and localization in images. In the first experiment we use these descriptors, combined with binary classifiers, to test the presence/absence of object in a target image. In the second experiment, we try to locate faces in images, by using a structural model. The results show that HMAX performs slightly better than SIFT in these tasks.

View PDFchevron_right
A Comparative Study of Local Descriptors for Object Category Recognition: SIFT vs HMAX. (Extended Version)
Plinio Moreno

2007

In this paper we evaluate the performance of the two most successful state-of-the-art descriptors, applied to the task of visual object detection and localization in images. In the first experiment we use these descriptors, combined with binary classifiers, to test the presence/absence of object in a target image. In the second experiment, we try to locate faces in images, by using a structural model. The results show that HMAX performs slightly better than SIFT in these tasks.

View PDFchevron_right
Improved object recognition results using SIFT and ORB feature detector
ANUPAM GARG, Munish Jindal

Object recognition has a wide domain of applications such as content-based image classification, video data mining, video surveillance and more. Object recognition accuracy has been a significant concern. Although deep learning had automated the feature extraction but hand crafted features continue to deliver consistent performance. This paper aims at efficient object recognition using hand crafted features based on Oriented Fast & Rotated BRIEF (Binary Robust Independent Elementary Features) and Scale Invariant Feature Transform features. Scale Invariant Feature Transform (SIFT) are particularly useful for analysis of images in light of different orientation and scale. Locality Preserving Projection (LPP) dimensionality reduction algorithm is explored to reduce the dimensions of obtained image feature vector. The execution of the proposed work is tested by using k-NN, decision tree and random forest classifiers. A dataset of 8000 samples of 100-class objects has been considered for experimental work. A precision rate of 69.8% and 76.9% has been achieved using ORB and SIFT feature descriptors, respectively. A combination of ORB and SIFT feature descriptors is also considered for experimental work. The integrated technique achieved an improved precision rate of 85.6% for the same.

View PDFchevron_right

References (28)

  1. REFERENCES
  2. A. E. Abdel-Hakim and A. A. Farag. Csift: A sift descriptor with color invariant char- acteristics. In 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06), volume 2, pages 1978-1983. Ieee, 2006.
  3. A. Alahi, R. Ortiz, and P. Vandergheynst. FREAK: Fast Retina Keypoint. In 2012 IEEE Conference on Computer Vision and Pattern Recognition, pages 510-517. IEEE, June 2012.
  4. H. Bay, T. Tuytelaars, and L. Van Gool. SURF: Speeded Up Robust Features. In Computer Vision ECCV 2006, pages 404-417. 2006.
  5. M. Calonder, V. Lepetit, C. Strecha, and P. Fua. Brief: Binary robust independent elementary features. In European conference on computer vision, pages 778-792, 2010.
  6. L. Fei-Fei, R. Fergus, and P. Perona. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories. Computer Vision and Image Understanding, 106(1):59 -70, 2007.
  7. C. Goutte and E. Gaussier. A probabilistic interpretation of precision, recall and f-score, with implication for evaluation. In European Conference on Information Retrieval, pages 345-359, 2005.
  8. M. Heikkil, M. Pietikinen, and C. Schmid. Description of interest regions with local binary patterns. Pattern Recognition, 42(3):425-436, March 2009.
  9. P. Indyk and R. Motwani. Approximate nearest neighbors: Towards removing the curse of dimensionality. In Proceedings of the Thirtieth Annual ACM Symposium on Theory of Computing.
  10. Y. Ke and R. Sukthankar. PCA-SIFT: a more distinctive representation for local image descriptors. In Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., volume 2, pages 506-513. IEEE, 2004.
  11. R. Kohavi. A study of cross-validation and bootstrap for accuracy estimation and model selection. In Ijcai, volume 14, pages 1137-1145. Montreal, Canada, 1995.
  12. S. Leutenegger, M. Chli, and R. Y. Siegwart. BRISK: Binary Robust invariant scalable key- points. In 2011 International Conference on Computer Vision, pages 2548-2555, November 2011.
  13. W. Liao. Region Description Using Extended Local Ternary Patterns. In 2010 20th Interna- tional Conference on Pattern Recognition, pages 1003-1006, August 2010.
  14. D. G. Lowe. Object recognition from local scale-invariant features. In Proceedings of the Seventh IEEE International Conference on Computer Vision, volume 2, pages 1150-1157 vol.2, September 1999.
  15. D. G. Lowe. Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, 60(2):91-110, November 2004.
  16. J. MacQueen. Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics, pages 281-297. University of California Press, 1967.
  17. E. Mair, G. D. Hager, D. Burschka, M. Suppa, and G. Hirzinger. Adaptive and generic corner detection based on the accelerated segment test. In European conference on Computer vision, pages 183-196. Springer, 2010.
  18. K. Mikolajczyk and C. Schmid. A performance evaluation of local descriptors. IEEE Trans- actions on Pattern Analysis and Machine Intelligence, 27(10):1615-1630, October 2005.
  19. M. Muja and D. G. Lowe. Fast approximate nearest neighbors with automatic algorithm configuration. VISAPP, 2(331-340):2, 2009.
  20. S. Murala, R. P. Maheshwari, and R. Balasubramanian. Local Tetra Patterns: A New Feature Descriptor for Content-Based Image Retrieval. IEEE Transactions on Image Processing, 21 (5):2874-2886, May 2012.
  21. L. Nanni, S. Brahnam, and A. Lumini. A local approach based on a local binary patterns variant texture descriptor for classifying pain states. Expert Systems with Applications, 37 (12):7888-7894, 2010.
  22. T. Ojala, M. Pietikinen, and D. Harwood. A comparative study of texture measures with classification based on featured distributions. Pattern Recognition, 29(1):51-59, January 1996.
  23. M. Pietikinen, A. Hadid, G. Zhao, and T. Ahonen. Local Binary Patterns for Still Images. In Computer Vision Using Local Binary Patterns, Computational Imaging and Vision, pages 13-47. Springer London, 2011.
  24. John T. Robinson. The k-d-b-tree: A search structure for large multidimensional dynamic in- dexes. In Proceedings of the 1981 ACM SIGMOD International Conference on Management of Data, SIGMOD '81, pages 10-18, 1981.
  25. E. Rosten and T. Drummond. Fusing points and lines for high performance tracking. In Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1, pages 1508-1515 Vol. 2. IEEE, 2005.
  26. E. Rublee, V. Rabaud, K. Konolige, and G. Bradski. ORB: An efficient alternative to SIFT or SURF. In 2011 International Conference on Computer Vision, pages 2564-2571, November 2011.
  27. E. Tola, V. Lepetit, and P. Fua. DAISY: An Efficient Dense Descriptor Applied to Wide- Baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(5): 815-830, May 2010.
  28. L. Wang and D.C. He. Texture classification using texture spectrum. Pattern Recognition, 23(8):905-910, 1990.

Related papers

2D FEATURES-BASED DETECTOR AND DESCRIPTOR SELECTION SYSTEM FOR HIERARCHICAL RECOGNITION OF INDUSTRIAL PARTS
Ibon Merino

2019

Detection and description of keypoints from an image is a well-studied problem in Computer Vision. Some methods like SIFT, SURF or ORB are computationally really efficient. This paper proposes a solution for a particular case study on object recognition of industrial parts based on hierarchical classification. Reducing the number of instances leads to better performance, indeed, that is what the use of the hierarchical classification is looking for. We demonstrate that this method performs better than using just one method like ORB, SIFT or FREAK, despite being fairly slower.

View PDFchevron_right
Histogram-Based Descriptor Subset Selection for Visual Recognition of Industrial Parts
Ibon Merino

Applied Sciences

This article deals with the 2D image-based recognition of industrial parts. Methods based on histograms are well known and widely used, but it is hard to find the best combination of histograms, most distinctive for instance, for each situation and without a high user expertise. We proposed a descriptor subset selection technique that automatically selects the most appropriate descriptor combination, and that outperforms approach involving single descriptors. We have considered both backward and forward mechanisms. Furthermore, to recognize the industrial parts a supervised classification is used with the global descriptors as predictors. Several class approaches are compared. Given our application, the best results are obtained with the Support Vector Machine with a combination of descriptors increasing the F1 by 0.031 with respect to the best descriptor alone.

View PDFchevron_right
Classification of Images Based on a System of Hierarchical Features
Professor Dr.Yousef Daradkeh

Computers, Materials & Continua, 2022

The results of the development of the new fast-speed method of classification images using a structural approach are presented. The method is based on the system of hierarchical features, based on the bitwise data distribution for the set of descriptors of image description. The article also proposes the use of the spatial data processing apparatus, which simplifies and accelerates the classification process. Experiments have shown that the time of calculation of the relevance for two descriptions according to their distributions is about 1000 times less than for the traditional voting procedure, for which the sets of descriptors are compared. The introduction of the system of hierarchical features allows to further reduce the calculation time by 2-3 times while ensuring high efficiency of classification. The noise immunity of the method to additive noise has been experimentally studied. According to the results of the research, the marginal degree of the hierarchy of features for reliable classification with the standard deviation of noise less than 30 is the 8-bit distribution. Computing costs increase proportionally with decreasing bit distribution. The method can be used for application tasks where object identification time is critical.

View PDFchevron_right
Object recognition and inspection in difficult industrial environments
J. Muoz, Carlos Spinola, Enrique Domínguez, Rafael Marcos Luque Baena

2006 IEEE International Conference on Industrial Technology, 2006

The manufacturing environment is characterized by increasing attention is being given to integration of computing rapid change, originating new challenges and problems to the technologies with the manufacturing systems leading to the production and operation manager in the industry. In response to the need for fast and flexible manufacturing, increasing attention deveopmento fible manactuin e is being given to integration of computing technologies with the high performance vision capabilities. manufacturing systems leading to the development of fast and Machine vision systems for component inspection processes flexible manufacturing systems aided with high performance can be regarded as a hierarchy of three basic levels vision capabilities. This paper presents an inspection and incorporating the data acquisition operation, feature extraction recognition system for manufacturing applications in difficult incorporang te c quision operation,fature etrat. In industrial environments. In such difficult environments, where alori and recogniton and decision-making strategies. In objects to be recognized can be dirty and illumination conditions industrial environments where the single production is cannot be sufficiently controlled, the required accuracy and manufactured, there is clear need to fast recognition of objects. rigidity of the system are critical features. The purpose of the The proposed object recognition scheme for inspection in real-time system is to inspect and recognize air-conditioning manufactured air-conditioning components has been developed

View PDFchevron_right
An Object Recognizing System for Industrial Applications
Tania Centeno

2003

On several automatic systems for manufacturing and assembly, specially on Flexible Manufacturing Systems (FMS), identifying mechanical parts produced and detecting their position and orientation are important issues related to the necessity of handling these parts by industrial robots. Currently, the recognizing process is performed by human check, but it may lead to increasing errors and accident probabilities. So the implementation of an effective automatic system, in order to recognize the parts, would not only avoid these risks but it would also improve the process velocity as well as its reliability This work presents an automatic system for free-form objects recognizing which identifies mechanical parts produced by a FMS. This system can recognize objects in a monochromatic image, captured by a charge couple device (CCD) camera. In addition to this, the system can be easily enabled to verify the parts orientation. This work used the concept of behavior vector, from the image indexing techniques, as a solution for the objects representation. Then, during the recognizing process, at least one hypothesis are generated by a backpropagation neural network trained to recognize the pattern vectors (known objects). Finally, the hypotheses are evaluated through a final verification process. As a result, the system offers quick and correct answers and also flexibility to be applied in other applications.

View PDFchevron_right

Related topics

  • Computer Vision
  • Object Recognition (Computer Vis...
  • Fuzzy Expert System
    • 580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved