The robustness and reliability of vision algorithms is, nowadays, the key issue in robotic research and industrial applications. To control a robot in a closed-loop fashion, different tracking systems have been reported in the literature. A common approach to increased robustness of a tracking system is the use of different models (CAD model of the object, motion model) known a priori. Our hypothesis is that fusion of multiple features facilitates robust detection and tracking of objects in scenes of realistic complexity. A particular application is the estimation of a robot's end-effector position in a sequence of images. The research investigates the following two different approaches to cue integration: 1) voting and 2) fuzzy logic-based fusion. The two approaches have been tested in association with scenes of varying complexity. Experimental results clearly demonstrate that fusion of cues results in a tracking system with a robust performance. The robustness is in particular evident for scenes with multiple moving objects and partial occlusion of the tracked object.

/pdf/cue-integration-for-visual-servoing-1dxesuputp.pdf

Cue integration for visual servoing

Road traffic movement is a very important source of information in traffic management. Although systems exist which can detect the presence of a vehicle and its speed under certain conditions, there is generally a lack of effective means to measure both the speed and direction of traffic movement. This is particularly true for road junctions, where conflicting traffic shares the same space and where some control strategy could be more effectively applied with the help of speed and direction estimates. The increasing use of closed circuit television (CCTV) systems has provided the opportunity to apply image processing techniques to extract such information. However, such techniques are computationally intensive in general, and the application of parallel processing methods is one of the best choices which could bring the desired acquisition of movement information into practical reality. This paper describes a parallel video-based image analysis system which is capable of extracting movement information, including direction and speed, of road vehicular traffic over any part of a road surface. The prototype has been implemented on an array of 36 transputers and an image grabber with a SUN SPARC IPC as the host machine. The software mainly consists of median filtering, feature extraction, spatio-temporal analysis, matching of image features in successive images by neural networks and aggregation of matched results. This algorithm has been tested using data for a signal-controlled junction aiming to capture an opposed turning traffic movement with promising results. It has also been shown that a real-time system based on the described algorithm is feasible.

A parallel algorithm to extract information about the motion of road traffic using image analysis

Robot Soccer is a rich domain for the study in artificial intelligence. Teams of players must work together in order to put the ball in the opposing goal. Learning is essential in this task since the dynamics of the system can change as the opponents' behaviours change. The players must be able to adapt to new situations. In this paper, we create a passing, avoiding obstacle and shooting strategy for robot soccer coordination. Based on a scenario in robot soccer, we stimulate a mini case study which involves two robots and ball.This research proposes coordination algorithm for robots collaboration in soccer game. The method is based on role, act, and behaviour of the robots. The actions of each robot depend on the created situation. The simulation result shows its probability to be applied in the real robot soccer game.

/pdf/proceeding-of-the-international-conference-on-advanced-4fnyl5chc8.pdf

Proceeding of the International Conference on Advanced Science, Engineering and Information Technology 2011

Many vision tasks are based on edge/corner detection. We propose some rules and utilize them to find suitable window for edge/corner detection by using quadratic programming. Experimental results over synthetic images and real images show satisfactory outcomes.

/pdf/edge-corner-programming-wpv4fu7qee.pdf

Edge/Corner Programming

Nowadays, logo and arm detection with growing variety and number of arms in companies and countries is one of the significant topics in image processing. For logo detection there are many image processing algorithms which can be used for this purpose regard to features of the logo. In the most of recent works, images are exactly the logos, but in general, the logo can be one portion of another image or even can transform (rotate, skew, shift, …) and have more complexity. In this paper we will select Iran logo as a sample, because it has not any regular geometric form and has special complexity. In a bank of images we will try to find this logo and its position with its transformation. Images bank includes images that has not Iran logo or has Iran logo with affine transform within another pictures. Approaches that will be discussed here are pose clustering and momentum clustering. Simulation results show that this approach can be used as a suitable way for finding the existence and position of arm in this field.

/pdf/logo-detection-using-pose-clustering-and-momentums-2yruo0cnm9.pdf

Logo Detection Using Pose Clustering and Momentums

Corners have become now one of the most commonly used features in image analysis, in particular, in motion analysis and stereo vision. There are many corner detection algorithms which can be classified into two groups. One is based on the extraction of object contours and followed by the analysis of their structures. The other works directly on the local grey values. In this paper, a moving corner detector for analysis of traffic movements is introduced which belongs to the second category. It comprises a corner response calculation component and a spatial temporal analysis part. The corner response calculation uses only the first order image derivatives, which is detected from Harris and Stephens' combined corner edge detector. Spatio-temporal analysis combines both spatial and temporal information to suppress noise and is achieved through more than two images. Apart from some random noises digitization, there are some slight shifts in space from frame to frame. These cause some stationary corners with high contrast being picked up as moving corners. Results showed that this moving corner detector is reasonably successful in separating these stationary corners from moving ones. >

https://ieeexplore.ieee.org/iel2/1132/7852/00333127.pdf

Detection of moving corners in dynamic images

The impact of fire on natural resources and the ecological environment is reflected in the fact that it alters the global carbon cycle, vegetation renewal process and soil properties through a combination of natural and human factors, thereby aggravating climate change, and also seriously endangers public life and property and causes huge economic losses. In the early stage of fire, smoke is more obvious than flame, so accurate and fast smoke detection algorithm is of great significance for preventing fire. In order to meet the requirements of smoke detection accuracy and detection speed, an improved real-time smoke detection algorithm is proposed on the basis of YOLOv5. Based on the YOLOv5 network structure, the GhostNet lightweight network is used to replace the backbone network, and the prediction header is replaced by C3Ghost. The C3 module only uses a small amount of traditional convolution to generate some feature maps, which greatly reduces the network parameters and speeds up the network inference speed; the CA attention mechanism is added to the backbone network, which makes the mobile network by embedding position information into the channel attention. Get information on a larger area and avoid introducing large overhead. Compared with the original YOLOv5 network model, the average detection accuracy of the improved method is increased by 2.2%, the FPS is changed from 181.82 to 222.22 frames per second, and the number of model parameters is reduced by 51.3%. The experimental results show that this method can effectively improve the performance of the smoke detection model.

YOLOv5 Improved Smoke Detection Algorithm *

To address the problems of low detection performance and large memory consumption of traditional smoke and fire detection methods in complex scenes such as grasslands. Based on the YOLOv5 model, a YOLOv5-GDE optimization model is proposed. The C3 module in YOLOv5 is replaced by GhostC3 with a smaller number of parameters, and some standard convolution blocks are replaced by depth-separable convolutions to make the model more lightweight. Finally, to solve the problem of unstable target regression frame, the EIoU loss function is introduced, which effectively improves the convergence speed and detection accuracy of the model. Experimental results on the homemade grassland smoke dataset show that the optimized model reduces the number of parameters and computational effort by 65.4% and 65.8%, respectively, compared with the original model, and the model size is only 36.8% of the original model, which is more suitable for smoke target detection in grassland scenes and more suitable for deployment in embedded devices with limited computational power, under the premise of ensuring detection accuracy.

Xu Zhang

Papers

Detection of moving corners in dynamic images

YOLOv5 Improved Smoke Detection Algorithm *

Improved YOLOv5 lightweight grassland smoke detection algorithm