Image segmentation is the most important step for any visual scene understanding system. In this paper, we use a semantic approach where each pixel is labeled with a semantic object category. Location of objects inside a tunnel’s road is a crucial task for an automatic tunnel incident detection system. It needs in particular to accurately detect and localize different types of zones, such as road lane, emergency lane, and sidewalk. Unfortunately, the existing methods often fail in providing acceptable image regions due to dynamic environment conditions: change in the lighting conditions, shadow appearance, objects variability, etc. To overcome these difficulties, we proposed to use the semantic tunnel image segmentation approach and a Convolutional Neural Network (CNN) to solve this problem. To evaluate the performance of the proposed approach, we performed a comparison to the state of the art and recent methods on two different datasets collected from two tunnels in France, called the ”T1” and ”T2”. Our extensive study leads to the provide of the best tunnel scene segmentation approach. The proposed method has been deployed by VINCI Autoroutes company in a real-world environment for automatic incident detection system.
Automatic License Plate Recognition (ALPR) is a technology designed to automatically read vehicle license plates. Traditional ALPR systems first detect the License Plate (LP), then apply the Optical Character Recognition (OCR) pipeline, which includes LP image pre-processing, character segmentation, character classification and post-processing. An ALPR system developed with such approaches often fails to provide acceptable results due to numerous challenging situations, which significantly increase the appearance variability of LPs as well as the characters to be classified. Recently, Convolutional Neural Network (CNN) models have proved efficient for ALPR problems. However, many of these CNN-based methods yet exhibit vulnerabilities to properly localize the region of the characters’ sequence and therefore provide an incorrect segmentation. Herein, this paper presents a novel real-time ALPR system that uses the concept of saliency map within the CNN model. The key contribution is at the segmentation step where the characters are located by means of the saliency map, which helps to refine the character classification step. The proposed ALPR pipeline consists of the two modules: 1) LPlocalization-CNN to detect the LP and 2) Saliency-Map-CNN to segment the characters in the LP. Experiments are conducted on a private and two public datasets and the proposed method is compared to the state-of-the-art methods. Results show that it performs well with respect to both accuracy and computation time, and hence clearly demonstrate the usefulness of the proposed system for the real-world ALPR applications.
The segmentation of microscopic images is an important issue in biomedical image processing. Many works can be found in the literature; however, there is not a gold standard method that is able to provide good results for all kinds of microscopic images. Then, authors propose methods for a given kind of microscopic images. This paper deals with new segmentation framework based on evidence theory, called ESA (Evidential Segmentation Algorithm) to segment blood cell images. The proposed algorithm allows solving the segmentation problem of blood cell images. Herein, our goal is to extract the components of a given cell image by using evidence theory, that allows more flexibility to classify the pixels. The obtained results showed the efficiency of the proposed algorithm compared to other competing methods.
In Computer Assisted Orthopaedic Surgery (CAOS), surgeons have to acquire some anatomical landmarks as
inputs to the system. To do so, they use manual pointers that are localized in the Operating Room (OR) space
using an infrared camera. When the needed landmark is not reachable through an opening, it is palpated directly
on skin and there is a loss of precision that can vary from several millimeters to centimeters depending on the
thickness of soft tissues. In this paper, we propose a new framework based on three main steps to register the
bone surface and extract automatically anatomical landmarks with an ultrasound probe. This framework is based
on an oriented gradient calculation, a simulated-compound and a contour closure using a graph representation.
The oriented gradient allows extracting a set of pixels that probably belong to the bone surface. The simulatedcompound
step allows using ultrasound images properties to define a set of small segments which may belong
to the bone surface, and the graph representation allows eliminating false positive detection among remaining
segments. The proposed method has been validated on a database of 230 ultrasound images of anterior femoral
condyles (on the knee). The average computation time is 0.11 sec per image, and average errors are: 0.54 mm
for the bone surface extraction, 0.31 mm for the condylar line, and 1.4 mm for the trochlea middle.
In this paper, we present a new image thresholding algorithm based on fractional filter (FF). Our experiments showed
that a good segmentation result corresponds to an optimal order of the filter. Then, we propose to use geometric
moments to find the optimal order. The proposed algorithm, called FLM, allows including contextual information such
as the global object shape and uses the properties of the two-dimensional fractional integration. The efficiency of FLM
was illustrated by the comparison to other six competing methods recently published and it was tested on real-world problem.
In this paper, we present an original method to evaluate the deformations in the third cerebral ventricle on a brain cine-
MR imaging. First, a segmentation process, based on a fractional differentiation method, is directly applied on a 2D+t
dataset to detect the contours of the region of interest (i.e. <i>lamina terminalis</i>). Then, the successive segmented contours
are matched using a procedure of global alignment, followed by a morphing process, based on the Covariance Matrix
Adaptation Evolution Strategy (CMAES). Finally, local measurements of deformations are derived from the previously
determined matched contours. The validation step is realized by comparing our results with the measurements achieved
on the same patients by an expert.