Adaptive Recognition of Color-Coded Objects in Indoor and Outdoor Environments

Abstract

To achieve robust color perception under varying light conditions in indoor and outdoor environments, we propose a three-step method consisting of adaptive camera parameter control, image segmentation and color classification. A controller for the intrinsic camera parameters is used to improve color stability in the YUV space. Segmentation is done to detect spatially coherent regions of uniform color belonging to objects in the image. Then, a probabilistic classification method is applied to label the colors by use of a Gaussian color distribution model. Experiments under combination of artificial and natural illuminations indoors and outdoors have been carried out. The results show the feasibility of this approach as well as the problems that occur under these highly diverse light situations. In particular we investigate the application in a RoboCup soccer scenario pointing toward future outdoor use.

Finding near optimum colour classifiers: genetic algorithm-assisted fuzzy colour contrast fusion using variable colour depth

“…In addition, hardware-assisted adaptive illumination invariant techniques were introduced. Takahashi et al [14] in 2008 employed a mechanical PID control to automatically adjust camera parameters, such as the iris and the gain, to adapt to illumination changes in the target environment. A reference red ring around the lens was used to determine when and how much adjustments for the iris and gain parameters need to be performed.…”

A Survey on Visual Perception for RoboCup MSL Soccer Robots

“…The theoretical colour values were used as reference values, so the effect from illumination could be eliminated, but special image regions must be selected manually by users in this method. Takahashi et al used a set of PID controllers to modify the camera parameters like gain, iris and two white-balance channels according to the changes of a white reference colour, which is always visible in the omnidirectional vision system [21]. Lunenburg and Ven adjusted the shutter time by designing a PI controller to modify the colour values of the referenced green field to the desired values [22].…”

“…In image analysis and understanding, the dynamic lighting conditions bring a great challenge to traditional object recognition methods by segmenting the image first and then detecting the colour blobs. Therefore, several object recognition algorithms that do not depend on colour segmentation have been proposed [21, 32, 33]. In [21], Markov Random Fields was used to segment the panoramic image and then, based on the assumption that the distribution of the object colour is Gaussian, each pixel of the image was classified to be an object colour according to its Mahalanobis distances to the Gaussian distribution of all of the reference object colours.…”

“…Therefore, several object recognition algorithms that do not depend on colour segmentation have been proposed [21, 32, 33]. In [21], Markov Random Fields was used to segment the panoramic image and then, based on the assumption that the distribution of the object colour is Gaussian, each pixel of the image was classified to be an object colour according to its Mahalanobis distances to the Gaussian distribution of all of the reference object colours. The experimental results in the indoor and outdoor environments validated the effectiveness of this method.…”

Robust and real-time self-localization based on omnidirectional vision for soccer robots

“…The theoretic color values were used as referenced values, so the effect from illumination could be eliminated, but special image regions needed to be selected manually by users as required by this method. Takahashi et al used a set of PID controllers to modify the camera parameters like gain, iris, and two white balance channels according to the changes of a white referenced color which is always visible in the omnidirectional vision system [18]. Lunenburg et al adjusted the shutter time by designing a PI controller to modify the color of the referenced green field to be the desired values [19].…”