1260556 五、發明說明(1) 【發明所屬之技術領域】 本發明係有關數位相機, 識別認證功能的方法。 其是有關使數位;j 【先前技術】 目前有些數位相機且 者自行設定個人偏好的 用模式之設計,供;^ 用者可透過選單選擇自行嗖定、Π 4攝影數值。3 但使用者模式常以選單=二^的使用模式,以進f 擇的項次受到使用者界選擇所要的模式’ 用上的不便。 7限制’若過多的選項脅 f數位相機或靜態影像播放器,另具 必須輪入密碼後,才可依權限的不 ,的,像。但在家庭上使用之數位相機 呆被模式,則有使用者易遺忘密碼或 問喊。 13月1日公佈的台灣專利公告第4 7 7 9 5 9 ,臉部影像辨識方法及系統,係應用正 夕2析度分解方法,將影像分解成至少 子衫像’將這些經分解的子影像經過自 ^進行無監督式歸類學習。測試階段由 右於低解析度無法辨識,再將可能人選 度進行辨識。 月2 1日公佈的台灣專利公告第々Μι 〇ί 谷辨硪之門禁系統,係預先取得的個人 機具有 目刖 式者,使 同安全層 像播放器 夠周全之 20 02 一種多層 波器,以 解析度的 神經網路 度開始, 一層解析 20 0 2 示一種面 、同使用 、同的使 ~攝影。 其可選 造成使 保密模 瀏覽不 數位影 保密不 ,揭示 鏡像濾 種不同 構圖類 低解析 給更高 ,揭 官長 1260556___ 五、發明說明(2) 度,估計其多變數機率分布。在辨識時,首先形成人面五 官結構的多變數機率分布,此一機率分布對於每一個人而 言均不相同,因而切分資料機率空間為每人一區。再對線 上偵得的人面萃取五官長度與每個人面五官結構之多變數 分布,計算其協方差。然後,計算此數據落入每一個機率 分布的機率加以識別,藉以決定來人是否擁有通過門禁的 權限。 目前的數位相機,並未見有利用上述專利案所揭露人 體臉部的辨識技術,使數位相機具有識別認證功能的相關 技術。 【發明内容】 為了使數位相機能以其拍攝的影像資料,作為識別依 據,而具有識別認證的功能,以解決上述使用者模式及保 密模式之設定所造成之問題,而提出本發明。 本發明的主要目的,在提供一種數位相機識別認證的 方法,使數位相機拍攝的影像,可作為識別認證的資訊, 並自動進行識別認證,以簡化使用者進行認證之操作,及 解決因遺忘密碼而無法認證之問題。 本發明的另一目的,在提供一種數位相機識別認證的 方法’使數位相機具識別認證的功能,確認使用者之後才 月b使數位相機具有攝影的功能,以增加產品之科技感。 本發明的又一目的,在提供一種數位相機影像識別認 證的方法,使數位相機具識別認證的功能,而不增加額外 的硬體設置成本。1260556 V. DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to a digital camera that recognizes an authentication function. It is related to making digits; j [Prior Art] At present, some digital cameras have their own design patterns for personal preference, and users can choose to set their own photographic values through the menu. 3 However, the user mode is often in the mode of use of the menu = two ^, and the order of the selection is subject to the inconvenience of the mode selected by the user community. 7 Limit 'If too many options threat f digital camera or still video player, the other must be entered after the password, can not rely on the rights, like. However, when the digital camera used in the home stays in the mode, there is a user who forgets the password or asks. The Taiwan Patent Publication No. 4 7 7 9 5 9 published on the 1st of November, the facial image recognition method and system, is applied to the Epoch 2 resolution decomposition method, and the image is decomposed into at least a sub-shirt like 'these decomposed sub- The image is subjected to unsupervised classification learning. The test phase is not recognized by right to low resolution, and the possible candidates are identified. The Taiwan Patent Notice No. 々Μι 〇ί 谷 硪 硪 硪 硪 硪 门 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷 谷Starting with the degree of neural network of resolution, a layer of resolution 20 0 2 shows a kind of face, the same use, the same make ~ photography. It can be used to make the security model browse the digital shadows. The composition of the mirror filter is different. The low-resolution is higher, and the official is 1260556___. The invention is (2) degrees, and the probability distribution of multiple variables is estimated. In the identification, the multi-variable probability distribution of the facial features of the human face is first formed. This probability distribution is different for each individual, so the probability of dividing the data space is one zone per person. The covariance is calculated by extracting the facial features of the human face and the multivariate distribution of each facial features. Then, calculate the probability that this data falls into each probability distribution to identify whether the person has the right to pass the access control. At present, digital cameras have not seen the identification technology of the human face disclosed by the above patents, so that the digital camera has the related technology of identifying and certifying functions. SUMMARY OF THE INVENTION The present invention has been made in order to enable a digital camera to use the image data captured thereon as a basis for identification and to have a function of identification authentication to solve the problems caused by the setting of the user mode and the security mode. The main object of the present invention is to provide a method for digital camera recognition and authentication, which enables an image captured by a digital camera to be used as information for identifying authentication, and automatically performs identification and authentication, so as to simplify the operation of the user for authentication, and solve the problem of forgetting the password. And the problem of not being able to authenticate. Another object of the present invention is to provide a digital camera identification and authentication method for enabling a digital camera to recognize and authenticate, and to confirm that the user has the function of photographing the digital camera to increase the sense of technology of the product. It is still another object of the present invention to provide a method of digital camera image recognition authentication that enables a digital camera to recognize authentication functions without adding additional hardware setup costs.
1260556 五、 發明說明(3)1260556 V. Description of invention (3)
本發明的其他目的、功效,請參閱圖式及實施例,詳 細說明如下。 【實施方式】 本發明主要是使數位相機具有包括影像分析軟體、類 砷經網路軟體,以執行本發明的各項步驟。 數位相機拍攝人臉的影像,可利用如前述專利案所揭 露者,計算出影像内容的構成特徵,作為類神經網路的輸 入值,使類神經網路進行學習,產生影像識別認證學習結 果之數據值,及計算出與該人臉的影像相對應的輸出值, 作為後續影像的識別認證之用。 類神經網路是指利用電腦來模仿生物神經網路的處理 系統,即是一種計算系統,使用大量簡單的相連人工神經 元來模仿生物神經網路的能力。類神經網路的架構包括輸 入層、隱藏層及輸出層。利用類神經網路之前,必須使其 從學習設定的輸入資料與設定的輸出資料的關係中,建立 内在的對映規則。 本發明有關的類神經網路,所應用的學習方法為監督 式學習。係使類神經網路從訓練認定人臉的範例上,學習 輸入值和目標輸出值的内在對應規則。在學習的過程中, 當輪入某一人臉的影像構成特徵值資料後,網路會以目前 的權重計算出相對應的推論值,即用以認定某一人臉的輸 出值,以及推論值和目標輸出值的誤差值,而誤差值再回 饋到網路中以調整網路内部的權重。經由不斷的讀入同一 人臉數張影像的訓練範例輸入值,以及重複的訓練把例For other purposes and functions of the present invention, please refer to the drawings and the embodiments, which are described in detail below. [Embodiment] The present invention mainly provides a digital camera with an image analysis software and an arsenic-based network software to perform the steps of the present invention. The digital camera captures the image of the human face, and can use the method disclosed in the aforementioned patent to calculate the constituent features of the image content, and as an input value of the neural network, the neural network is learned, and the image recognition authentication learning result is generated. The data value and the output value corresponding to the image of the face are calculated as the identification and authentication of the subsequent image. A neural network is a processing system that uses a computer to mimic a biological neural network. It is a computing system that uses a large number of simple connected artificial neurons to mimic the capabilities of a biological neural network. The architecture of a neural network includes an input layer, a hidden layer, and an output layer. Before using a neural network, it is necessary to establish an inherent mapping rule from the relationship between the learned input data and the set output data. In the neural network of the present invention, the learning method applied is supervised learning. The neural network is used to learn the intrinsic correspondence rules between the input value and the target output value from the example of training the face. In the process of learning, when the image of a face is formed into the feature value data, the network calculates the corresponding inference value with the current weight, that is, the output value used to identify a face, and the inference value and The error value of the target output value, and the error value is fed back to the network to adjust the weight inside the network. Through the continuous input of training examples input values of several images of the same face, and repeated training examples
1260556___ I、發明說明(4) ' "^ ~—-- 後,網路會漸漸的修正内部的權重,而使得網路推論值漸 漸的逼近目標輸出值。當目標輸出值和推論值接近到某^ 範圍内時,表示網路已經從訓練範例中學到範例資料之間 的規則,則可停止學習而不再改變權重,而獲得一人臉影 像識別認證學習結果之數據值。 〜 本發明數位相機影像識別認證的方法,包括建立影像 識別認證依據的方法,及進行影像識別認證依據的方法兩 部分。 請參閱圖1所示。本發明數位相機建立影像識別認證 依據的方法,包括如下步驟: (1) 使數位相機攝取一張人臉的相片; (2) 使數位相機的c PU依據影像分析軟體,計算該相片的影 像資訊,以獲得該影像的構成特徵值; (3) 使CPU控制類神經網路,以該構成特徵值做為輸入值, 進行學習以獲得影像識別認證學習結果之數據值,及一與 該人臉相對應的輸出值; (4) 使C P U將該影像識別§忍證學習結果之數據值及輸出值’ 儲存記憶體’俾作為影像識別認證之依據。 上述步驟(1)中人臉相片’其拍攝人臉的大小、遠近 及環境須受到限制。 本發明可以不同的人臉影像,重複上述方法,使類神 經網路一再學習,以獲得更新的影像識別認證學習結果之 數據值,益獲得與該不同的人臉相對應不同的輸出值。該 更新的影像識別認證學習結果之數據值,及所有不同的輸1260556___ I, invention description (4) ' "^ ~—-, the network will gradually correct the internal weight, and make the network inference value gradually approach the target output value. When the target output value and the inference value are close to a certain range, indicating that the network has learned the rules between the sample data from the training paradigm, the learning can be stopped without changing the weight, and a face image recognition authentication learning result is obtained. The data value. ~ The method for image recognition and authentication of the digital camera of the present invention comprises two methods of establishing an image recognition authentication basis and a method for performing image recognition authentication. Please refer to Figure 1. The method for establishing the image recognition authentication basis of the digital camera of the present invention comprises the following steps: (1) causing the digital camera to take a photo of a human face; (2) causing the cPU of the digital camera to calculate the image information of the photo according to the image analysis software. Obtaining a constituent characteristic value of the image; (3) causing the CPU to control the neural network, using the constituent feature value as an input value, learning to obtain a data value of the image recognition authentication learning result, and a face Corresponding output value; (4) Let the CPU identify the data value of the § forcible learning result and the output value 'storage memory' as the basis for image recognition authentication. In the above step (1), the size, proximity and environment of the face of the human face must be limited. The present invention can repeat the above method for different facial images, so that the neural network can learn again and again to obtain the updated data identification authentication learning result data value, and obtain different output values corresponding to the different human faces. The updated image identifies the data values of the certified learning results and all the different inputs.
1260556 五、發明說明(5) 出值均儲存於數位相機的 措轉内 像的識別認證之依據^以後攝取人臉影 本發明數位相機影像識別認證的方 請參閱圖2所示 法,包括如下步驟: (1)使數相機攝取使用者的人臉相片; f'r ί ^ ^ ^ # ^ ^ ^ iMa ^ # # 貝afl,以獲付該影像的構成特徵值; (3)使CPU控制類神經網路,以該構成 :::::認證學習結果之數據值1算』人臉相對 =::::一值值相比較,若: 者通過認證。 』出值被一问時’即認定該使用 > ΪΓ::設Ϊ數位相機,當使用者要使用數位相機之 二=‘!: Γ者的人臉相片。之後,⑯位相機即 m :證,若其輸出值與存播的某-輸 出值相同,或兩者之差異在某一被認為相同的範圍内時, 即7認定該使用者通過認證,使CPU控制數位相機進入符 合该使用者原先設定的使用模式及/或保密模式,甚至使 CPU控制數位相機進入可被使用拍照操作的 才能利用數位相機進行拍照。 、Λ 使用者 j 2明使數位相機拍攝的影像,可作為識別認證的資 Λ,並自動進灯識別H以簡化使用者使數位像機 認證之操作,能增加產品之科技感…增加額外的硬體1260556 V. INSTRUCTIONS (5) The basis for the identification and authentication of the image stored in the digital camera. The image of the digital camera of the present invention is shown in the figure below, including the following steps. : (1) Let the camera capture the user's face photo; f'r ί ^ ^ ^ # ^ ^ ^ iMa ^ # #贝afl, to obtain the constituent eigenvalues of the image; (3) Make the CPU control class The neural network, with the composition::::: the data value of the authentication learning result is calculated as "the face is relatively =:::: the value of one value is compared, if: the person passes the authentication. When the value is asked, it is determined that the use > ΪΓ:: is set to a digital camera, when the user wants to use the digital camera 2 = ‘!: The face photo of the 。. After that, the 16-bit camera is m: the card, if the output value is the same as the stored-output value, or the difference between the two is within a certain range that is considered to be the same, that is, 7 the user is authenticated, so that The CPU controls the digital camera to enter a usage mode and/or a privacy mode that is originally set by the user, and even allows the CPU to control the digital camera to enter a photo-taking operation to take a picture with the digital camera. Λ User j 2 can make the image captured by the digital camera as the identification of the authentication, and automatically enter the light recognition H to simplify the user's operation of digital camera authentication, which can increase the technological sense of the product...Additional additional Hardware
$ 9頁 1260556 五、發明說明(6) 設置成本。 以上所記載,僅為利用本發明技術内容之實施例,任 何熟悉本項技藝者運用本創作所為之修飾、變化,皆屬本 創作主張之專利範圍,而不限於實施例所揭示者。$ 9 pages 1260556 V. Description of invention (6) Setting costs. The above description is only for the embodiments of the present invention, and any modifications and variations made by those skilled in the art using the present invention are the scope of patents of the present invention, and are not limited to those disclosed in the embodiments.
第10頁 1260556 圖式簡單說明 【圖示簡單說明】 圖1為實施本發明數位相機建立影像識別認證依據的方法 的流程圖。 圖2為實施本發明數位相機影像識別認證的方法的流程 圖。 【主要元件符號說明】 1 步驟(1) 2 步驟(2 ) 3 步驟(3 )Page 10 1260556 Brief Description of the Drawings [Simple Description of the Drawings] Fig. 1 is a flow chart showing a method for establishing the basis for image recognition authentication by the digital camera of the present invention. Fig. 2 is a flow chart showing a method of implementing image recognition and authentication of a digital camera of the present invention. [Main component symbol description] 1 Step (1) 2 Step (2) 3 Step (3)
4 步驟(4 )4 steps (4)
第11頁Page 11