请哪位英语高手帮我翻译下这文章
KadewTraKuPong and Bowden [4] presented a method that improves this adaptive background mixture model [1], reinvestigating the update equations. They use different equations at different phases. This allows the system to learn faster and more accurately, as well as to adapt to changing environments effectively. This approach updates the equations derived from sufficient statistics [5] and Lrecent window formula over other approaches of McKenna et al apud [4]. The model begins to estimate the Gaussian mixture model using expected sufficient statistics update equations, then switches to L-recent window version when the first L samples are processed. The expected sufficient statistics update equations provide a good estimate at the beginning before all L samples are collected. The L-recent window update equations prioritize recent data, therefore the tracker can adapt to changes in the environment.
Power and Schoones [3] suggest approximations and modifications from the standard algorithm [1] to improve performance. The modification consists on replacing the probability density function value by the weight of distribution in the update equations. Other approaches different from the one mentioned previously, have been developed as an alternative to statistical models, such as Weiner filter [12], Kalman filter [14], Bayesian decision theory and principal components analysis (PCA) [13], and Hidden Markov Models (HMM) [15]. Most of these solutions present high computational costs when compared to the GMM and/or worst results for robust background estimation.
3. Background Model Estimation
The basic idea is to define a segmented region, delimiting the pixels of interest. To do so, it is necessary to model the color attribute of each pixel of an image sequence (pixel process) through an adaptive mixture of Gaussian distributions. The mixture of Gaussian distribution model is updated for each new captured observation, reducing the influence from the past observations and allowing the model adaptation according to a gradual variation on illumination. However, the Gaussian distributions represent both foreground and background. It would be necessary to define the distribution of the subsets to describe the background model. The subset definition happens at each observation,according to the associated weights of every distribution indicating the frequency that the distribution better represented the pixel. After the pixel process, the foreground pixels are submitted to skin color segmentation. The next task is to find the separated objects. First, the speckle noise is removed by morphology (opening and closing) [11], then connected regions are determined and grouped regions
are separated into objects.
3.1. Pixel-Processing
A mixture of K Gaussian distributions models each pixel in the scene. The history of pixels can be defined as a temporal series from these pixel values that are vectors
KadewTraKuPong和Bowden [4]提出了改进这个能适应的背景混合物模型的一个方法[1]，再调查更新等式。 他们使用不同的等式在不同的阶段。 这允许系统快速地和更加准确地学会，并且有效地适应变化的环境。 这种方法更新从在其他方法的充足的统计[5]和Lrecent窗口惯例获得的等式McKenna等apud [4]。
当第一L样品被处理时，使用期望的足够的统计更新等式，模型开始估计高斯混合物模型，然后换成L最近窗口版本。 在所有L样品收集之前，期望的足够的统计更新等式首先提供一个好估计。 L最近窗口更新等式给予最近数据优先，因此跟踪仪能适应在环境上的变化。
Power和Schoones [3]建议从标准算法[1的]略计和修改改进表现。 修改在替换包括可能性密度函数价值由重量在更新等式的发行。 其他接近与以前被提及的那个不同，被开发了作为对统计模型的一个选择，例如Weiner过滤器[12]， Kalman过滤器[14]，贝叶斯决策理论和主要成分分析(PCA) [13]和暗藏的马尔可夫模型(HMM) [15]。
大多这些解答提出高计算费用，当与健壮背景估计的GMM和最坏的结果比较。
3。背景式样估计
The基本思想是定义一个被分割的区域，划定映象点利益。 要做如此，通过高斯发行一个能适应的混合物塑造图象序列(映象点过程)的每个映象点颜色属性是必要的。
高斯发行模型混合物为每新的被夺取的观察是更新，减少从过去观察的影响和允许式样适应根据逐渐变异对照明。 然而，高斯发行代表前景和背景。 定义子集的发行描述背景模型是必要的。 子集定义发生在每观察，根据表明伴生的重量每发行那的频率发行更好代表了映象点。
在映象点过程以后，前景映象点递交给肤色分割。 下项任务是发现被分离的对象。 首先，形态学取消斑点噪声(开头和结束) [11]，然后被连接的地区是坚定和被编组的地区
are分离了入对象。
3。
1。 映象点处理
K高斯发行A混合物塑造在场面的每个映象点。 映象点的历史可以被定义，当从是的这些映象点价值的一个世俗系列。
答：它讨论的是教育中的爱的价值详情>>
问：在填美签160表格，想问一下，精整、精加工是属于哪一类行业？,表格列举了以下行业...
答：职业都跟精有关系,确实没这类。详情>>
答：这个问题很难回答，就算是问老师或者是专家他们也很难给你一个确切的答案：） 考试通不通过完全看个人在准备考试上付出的努力够不够，只要用心学用功学习我相信是肯定可以...详情>>
