The Effects of Visual Object Size in The Depth Perception in KM Mouse

Whether mice perceive the depth of space dependent on the visual size of object targets was explored when visual cues such as perspective and partial occlusion in space were excluded. A mouse was placed on a platform the height of which is adjustable. The platform located inside a box in which all other walls were dark exception its bottom through that light was projected as a sole visual cue. The visual object cue was composed of 4x4 grids to allow a mouse estimating the distance of the platform relative to the grids. Three sizes of grids reduced in a proportion of 2/3 and seven distances with an equal interval between the platform and the grids at the bottom were applied in the experiments. The duration of a mouse staying on the platform at each height was recorded when the different sizes of the grids were presented randomly to test whether the Judgment of the mouse for the depth of the platform from the bottom was affected by the size information of the visual target. The results from all conditions of three object sizes show that time of mice staying on the platform became longer with the increase in height. In distance of 20 similar to 30 cm, the mice did not use the size information of a target to judge the depth, while mainly used the information of binocular disparity. In distance less than 20 cm or more than 30 cm, however, especially in much higher distance 50 cm, 60 cm and 70 cm, the mice were able to use the size information to do so in order to compensate the lack of binocular disparity information from both eyes. Because the mice have only 1/3 of the visual field that is binocular. This behavioral paradigm established in the current study is a useful model and can be applied to the experiments using transgenic mouse as an animal model to investigate the relationships between behaviors and gene functions.

Biologically Inspired Features for Scene Classification in Video Surveillance

Inspired by human visual cognition mechanism, this paper first presents a scene classification method based on an improved standard model feature. Compared with state-of-the-art efforts in scene classification, the newly proposed method is more robust, more selective, and of lower complexity. These advantages are demonstrated by two sets of experiments on both our own database and standard public ones. Furthermore, occlusion and disorder problems in scene classification in video surveillance are also first studied in this paper.

The experimental research of R-phycoerythrin subunits on cancer treatment: A new photosensitizer in PDT

The mouse tumor cell 5180 and human liver carcinoma cell SMC 7721 cells were first treated with R-PE and its subunits (alpha, beta, gamma subunits), then irradiated with Argon laser (496 nm, 28.8 J/cm(2)). Survival rate was measured by MTT method. In order to compare the phototoxicity in normal cells, the mouse marrow cells were treated with photofrin II and beta-subunit, irradiated with 45 J/cm(2) of light; survival rate was also measured by MTT method. The result showed that R-PE subunits had better PDT effect on s180 cells than R-PE and lower phototoxicity in marrow cells than photofrin II Flow cytometric analysis showed that PDT results in a growth inhibition and a G(0)-G(1) cell cycle arrest in SMC 7721 cells. The tumor cells inhibited by PDT in vivo were morphologically observed by TEM, the tumor cell death was daze to the occlusion of tumor blood vessels and inducement of cell programmed death in nuclei. Therefore, with the advantage in special fluorescence activity, loth molecular weight, good light absorbent character and weak phototoxicity, R-PE subunit is art attractive option for improving the selectivity of PDT.

目标局部或全部遮挡条件下跟踪方法研究

运动目标跟踪技术是计算机视觉领域的核心课题之一，具有十分重要的意义和广泛的实用价值。它融合了图像处理、模式识别、人工智能、自动控制以及计算机应用等相关领域的先进技术和研究成果。运动目标跟踪的实质是一种从图像信号中实时自动识别目标，提取目标的位置信息，自动跟踪目标的技术。它的难点在于图像信号采集过程中的信息损失和复杂的应用环境。特别是跟踪过程中的遮挡问题越来越成为限制跟踪算法实用性的关键因素。如何解决运动目标跟踪过程中的遮挡问题是本文的研究重点。 针对运动目标的遮挡问题，本文提出了一种将基于灰度特征的多子区域相关跟踪算法与二维目标运动估计理论相结合的跟踪策略。在目标未被遮挡或仅被部分遮挡时，采用多子区域相关跟踪算法，利用目标剩余的特征对目标进行定位。当遮挡比较严重，目标剩余特征不足以准确定位目标位置或目标已被完全遮挡时，多子区域相关跟踪算法已经不能识别目标，这时根据目标运动的历史信息，运用二维目标运动估计理论预测出目标的位置。经过大量仿真试验，结果表明：该算法已经达到了良好的抗遮挡效果，在遮挡和半遮挡情况下，跟踪结果达到了较高的精度。 为了验证算法在实际工程中的效果，笔者模拟某红外观瞄指示装置，为算法搭建了一个基于虚拟现实技术的硬件验证平台，并在该平台上验证了部分算法。

Elliptic Object Detection Based on Shape Preserving and Active Contour

本文通过形状约束方程(组)与一般主动轮廓模型结合,将目标形状与主动轮廓模型融合到统一能量泛函模型中,提出了一种形状保持主动轮廓模型即曲线在演化过程中保持为某一类特定形状。模型通过参数化水平集函数的零水平集控制演化曲线形状,不仅达到了分割即目标的目的,而且能够给出特定目标的定量描述。根据形状保持主动轮廓模型,建立了一个用于椭圆状目标检测的统一能量泛函模型,导出了相应的Euler-Lagrange常微分方程并用水平集方法实现了椭圆状目标检测。此模型可以应用于眼底乳头分割,虹膜检测及相机标定。实验结果表明,此模型不仅能够准确的检测出给定图像中的椭圆状目标,而且有很强的抗噪、抗变形及遮挡性能。

基于形状保持主动轮廓模型长直条的检测

通过形状约束方程(组)与一般主动轮廓模型结合,将目标形状与主动轮廓模型融合到统一能量泛函模型中,提出一种形状保持主动轮廓模型。模型通过参数化水平集函数的零水平集表示某一类特定形状,不仅达到了分割即目标的目的,而且能够给出特定目标的定量描述。根据形状保持主动轮廓模型,建立一个用于长直条状目标检测的统一能量泛函模型,导出相应的Euler-Lagrange常微分方程并用水平集方法实现了长直条状区域的检测。此形状保持模型的一种特殊情况可以用于直线状地平(海天)线提取。实验结果表明,该模型不仅能够准确地检测出给定图像中的长直条状区域而且有很强的抗噪、抗变形及遮挡性能。

基于外极线分区的动态立体匹配算法

基于区域的立体匹配算法仅针对支撑窗内的灰度信息定义匹配代价函数,导致在弱(无)纹理区域采用WTA优化出现歧义性。该文在外极线分区的基础上,改用区域作为匹配基元,针对歧义性区域,在代价函数中引入遮挡项和平滑项,并按照区域优先级的高低,动态匹配相应区域,获得可靠的视差信息。实验证明,该算法在保持实时性的同时对弱纹理区域处理具有有效性。

遮挡范式下碰撞时间估计的特性研究

The estimation of Time to Collision (TTC) related to avoiding collisions or making interceptions is an important cognitive ability for individuals. A number of studies have been carried out on this topic and related theories are developed. One of the most famous theories is the τ Theory. Based on the τ Theory, researches have found that visual information and physical information of moving objects would influent the TTC estimation. Are there any other factors that can affect people’s TTC estimation? A mixed design was used to examine the TTC estimation by different types of participants (professional drivers / people can not drive), with different moving objects (car/tricycle), under different speed (slow/fast) and direction (left to right / right to left) in Occlusion Paradigm. There were 21 professional drivers and 20 individuals who cannot drive participated in the experiment which was displayed on a computer. Participants were asked to click the button when he/she believed that the moving object had just contacted the red line. E-prime was used to establish the whole experimental environment and the RT was recorded at the same time. The results revealed that: (1) there is significant different TTC estimation between car and tricycle; (2) the professional drivers have more accurate TTC estimation than people do not drive. We can come to conclusion that conceptual information and driving skill could affect TTC estimation.