Moving object detection based on frame difference and W4

Sengar, Sandeep Singh; Mukhopadhyay, Susanta

doi:10.1007/s11760-017-1093-8

Cited by 63 publications

(21 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…For moving object detection, different approaches based on the difference in color distribution or pixel intensity have been proposed by researchers, e.g., [ 1 , 2 , 3 , 4 , 6 , 7 , 8 , 9 , 10 , 23 ], to eliminate the background in video frames. A widely used algorithm with low computational complexity is frame differencing [ 6 , 7 ], which utilizes the gray level difference between two or three adjacent video frames. However, frame differencing is vulnerable to various interferences caused by local motions and complex scenes.…”

Section: Related Workmentioning

confidence: 99%

“…Existing methods for moving object detection include background subtraction [ 1 , 2 , 3 , 4 , 5 ], frame differencing [ 6 , 7 ], optical flow [ 8 , 9 , 10 ], ViBe [ 11 , 12 ] and deep learning [ 13 , 14 , 15 , 16 , 17 , 18 ]. Accuracies are adversely affected when using these methods to achieve real-time detection due to high image resolution and environmental complexities.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Real-Time Moving Object Detection in High-Resolution Video Sensing

Zhu

Wei

et al. 2020

Sensors

View full text Add to dashboard Cite

This paper addresses real-time moving object detection with high accuracy in high-resolution video frames. A previously developed framework for moving object detection is modified to enable real-time processing of high-resolution images. First, a computationally efficient method is employed, which detects moving regions on a resized image while maintaining moving regions on the original image with mapping coordinates. Second, a light backbone deep neural network in place of a more complex one is utilized. Third, the focal loss function is employed to alleviate the imbalance between positive and negative samples. The results of the extensive experimentations conducted indicate that the modified framework developed in this paper achieves a processing rate of 21 frames per second with 86.15% accuracy on the dataset SimitMovingDataset, which contains high-resolution images of the size 1920 × 1080.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Real-Time Moving Object Detection in High-Resolution Video Sensing

Zhu

Wei

et al. 2020

Sensors

View full text Add to dashboard Cite

show abstract

“…Metode deterministik biasanya melakukan deteksi dan pelacakan wajah berdasarkan kemiripan citra referensi dengan citra yang dideteksi. Beberapa metode yang termasuk dalam metode deterministk adalah metode background subtraction [1][2][3], frame difference [4][5], optical flow [6], maupun ekstraksi warna kulit [7].…”

Section: Pendahuluanunclassified

“…Contoh hasil pendeteksian citra wajah ditampilkan pada Gambar Proses terakhir dari satu siklus partikel filter adalah proses estimasi. Proses estimasi posisi wajah yang dilacak dilakukan dengan (4). Dari sini, kita dapat mengestimasi posisi wajah yang dilacak dengan merata-ratakan posisi tiap partikel hasil resampling.…”

Section: A Deteksi Wajahunclassified

Deteksi dan Pelacakan Wajah Berdasarkan Warna Kulit Menggunakan Partikel Filter

Sugandi¹

2018

JRE

View full text Add to dashboard Cite

Deteksi dan tracking wajah merupakan salah satu bidang yang terus berkembang dalam berbagai aplikasi seperti sistem keamanan, biometric, maupun aplikasi lainnya. Akan tetapi deteksi dan pelacakan wajah dapat menjadi masalah yang kompleks ketika dilakukan secara real time. Beberapa masalah yang muncul di antaranya kebergantungan hasil deteksi terhadap cahaya dan keberagamaan citra latar belakang. Dalam artikel ini, kami hadirkan metode baru untuk mengatasi masalah tersebut dengan menggunakan filter warna HSL (Hue, Saturation, Luminance) dikombinasikan dengan partikel filter. Filter warna HSL akan mendeteksi wajah berdasarkan warna kulit dengan cara mengatur thresshold pada tiap komponen warna H, S, dan L. Proses selanjutnya adalah melacak wajah yang sudah terdeteksi dengan menggunakan partikel filter. Sebagai inisialisasi, partikel akan disebar secara random di sekitar wajah yang terdeteksi untuk kemudian dihitung bobot tiap partikel berdasarkan jarak tiap partikel terhadap posisi wajah. Bobot tiap partikel akan di-resampling untuk mendapatkan sebaran partikel baru yang memungkinkan partikel dengan bobot yang besar terpilih beberapa kali. Sementara partikel dengan bobot yang relatif rendah akan hilang dengan sendirinya. Sebaran partikel baru akan berada di sekitar wajah yang sedang dilacak. Proses terakhir dari satu siklus partikel filter adalah proses estimasi posisi wajah yang dilacak yang dilakukan dengan merata-ratakan posisi tiap partikel hasil resampling. Hasil pengujian menunjukkan ratarata kesalahan pelacakan posisi wajah menggunakan jumlah partikel filter 100, 200, dan 500 pada sumbu X dan Y berturut-turut adalah (1.25; 1.02), (0.86; 0.81), dan (0.59; 0.57).

show abstract

“…Early research in [1] detects animal faces using Haar-like features and the Adaboost classifier, while tracking the animals was done using the Kanade-Lucas-Tomasi method. Researchers have investigated different approaches to detect animals or humans: detection of humans in motion using background subtraction (BG) [2], using frame differences with the W4 algorithm [20], using background frame differences based on Gaussian functions [12], and the combination of BG and three-frame differencing [13].…”

Section: Introductionmentioning

confidence: 99%

Detection and Recognition of Badgers Using Deep Learning

Okafor

Berendsen²,

Schomaker

et al. 2018

Artificial Neural Networks and Machine Learning – ICANN 2018

View full text Add to dashboard Cite

This paper describes the use of two different deep-learning algorithms for object detection to recognize different badgers. We use recordings of four different badgers under varying background illuminations. In total four different object detection algorithms based on deep neural networks are compared: The single shot multi-box detector (SSD) with the Inception-V2 or MobileNet as a backbone, and the faster region-based convolutional neural network (Faster R-CNN) combined with Inception-V2 or residual networks. Furthermore, two different activation functions are compared to compute probabilities that some badger is in the detected region: the softmax and sigmoid functions. The results of all eight models show that SSD obtains higher recognition accuracies (97.8%-98.6%) than Faster R-CNN (84.8%-91.7%). However, the training time of Faster R-CNN is much shorter than that of SSD. The use of different output activation functions seems not to matter much.

show abstract

Moving object detection based on frame difference and W4

Cited by 63 publications

References 27 publications

Real-Time Moving Object Detection in High-Resolution Video Sensing

Real-Time Moving Object Detection in High-Resolution Video Sensing

Deteksi dan Pelacakan Wajah Berdasarkan Warna Kulit Menggunakan Partikel Filter

Detection and Recognition of Badgers Using Deep Learning

Contact Info

Product

Resources

About