The ability to detect movement in vivo and build a safe and reliable payment level "face brushing" experience

Face recognition At present, it has been widely used in mobile phone unlocking 、 Brush your face to pay 、 Life scenes such as gate authentication , However , Although the ability of face recognition brings great convenience , But can't identify whether the face is real , For example, use high simulation pictures 、 Precision gypsum or 3D Modeling mask , You can easily break through the face recognition algorithm , Using this capability alone has great potential safety hazards .

Huawei Machine learning services The ability to detect movement in vivo , Live body detection is carried out by means of command action coordination , Blinking 、 open one 's mouth 、 Shake your head left 、 Shake your head right 、 gaze 、 Randomly choose three of the six actions of nodding , Let the user complete the action according to the instructions , Use face key points and face tracking technology , Through continuous pictures , Calculate the ratio of changing distance to constant distance , Compare the image of the previous frame with the image of the next frame , So as to verify whether the user is a real living person , To the photo 、 Video and mask attacks have a good defense effect , It is a prerequisite for the effective application of face recognition .

besides , In the use of Motion vivisection detection ability In the process , Detect scenes such as occlusion and poor light , Support boot detection , Such as timely display “ Dark light hint ”、“ The portrait is blurred ”“ sunglasses 、 Mask blocking ”、“ Face too close 、 Too far ” Etc , Achieve a more friendly interactive experience , Create a safe and reliable payment level living detection capability .

Compared with silent vivisection, which does not require the user to make cooperative actions , The interactive motion detection capability is more suitable for banking and finance 、 Medical and other scenes requiring human-computer interaction . such as , Use this technology in the financial field , Users do not have to go to the bank site in person , You can open a financial account remotely 、 Insurance financing and other operations ; In self-service payment scenarios such as offline supermarkets , The user needs to complete the payment through action vivisection , Ensure the security of personal funds ; In social security 、 Health care 、 In the individual income tax and other handling operation scenarios , It is also necessary to accurately verify whether the operator is a living person through action living body detection , In order to improve operational safety .

So how to integrate motion detection capabilities ？ Steps are as follows .

1 Development steps

Before development , You need to complete the necessary development preparations , At the same time, please make sure that your project has been configured with HMS Core SDK Of Maven Warehouse address , And completed the service of SDK Integrate .

Mode one ：fullSDK Way integration

dependencies{
    //  Introduce the action vivisection collection package .
    implementation 'com.huawei.hms:ml-computer-vision-interactive-livenessdetection
: 3.2.0.122'
}

Mode two ： Basics SDK Way integration

dependencies{
    //  Introduce in vivo detection plugin package .
    implementation 'com.huawei.hms:ml-computer-vision-interactive-livenessdetection-plugin:3.2.0.122'
}

Motion vivisection provides two call modes , You can choose the corresponding calling method to build the living body detection service according to your needs .

1.1     Default scanning interface 
1. Create a live test result callback , Used to obtain test results .
private MLInteractiveLivenessCapture.Callback callback = new MLInteractiveLivenessCapture.Callback() {
    @Override
    public void onSuccess(MLInteractiveLivenessCaptureResult result) {
        //  Detect successful processing logic , The test results may be living or non living .
        swich(result.getStateCode()) {
            case InteractiveLivenessStateCode.ALL_ACTION_CORRECT:
            // After the verification is passed, the corresponding specific operations 

            case InteractiveLivenessStateCode.IN_PROGRESS:
            // Corresponding specific operation when detecting 
            …
    }

    @Override
    public void onFailure(int errorCode) {
        //  The test is not complete , If the camera is abnormal CAMERA_ERROR, Add failed processing logic .
    }
};
2. Create an in vivo detection instance , Start detection .
MLInteractiveLivenessConfig interactiveLivenessConfig = new MLInteractiveLivenessConfig.Builder().build();

        MLInteractiveLivenessCaptureConfig captureConfig = new MLInteractiveLivenessCaptureConfig.Builder()
                .setOptions(MLInteractiveLivenessCaptureConfig.DETECT_MASK)
                .setActionConfig(interactiveLivenessConfig)
                .setDetectionTimeOut(TIME_OUT_THRESHOLD)
                .build();
MLInteractiveLivenessCapture capture = MLInteractiveLivenessCapture.getInstance();
capture.startDetect(activity, callback);
1.2     Customize the scanning interface 
1. establish MLInteractiveLivenessDetectView, And load to Activity Layout .
/**
* I. Bind camera preview interface , Set the living body recognition area .
* In the camera preview stream , In vivo detection will judge whether the face is in the face frame of the preview video stream , In order to improve living * Body passing rate , It is recommended that the face frame be placed in the middle of the screen , And the living body recognition area is slightly larger than the drawn face frame .
* II. Set whether to detect mask .
* III. Set the result callback .
* IV. take MLInteractiveLivenessDetectView Load into Activity.
*/
@Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);
        setContentView(R.layout.activity_liveness_custom_detection);
        mPreviewContainer = findViewById(R.id.surface_layout);
        MLInteractiveLivenessConfig interactiveLivenessConfig = new MLInteractiveLivenessConfig.Builder().build();
mlInteractiveLivenessDetectView = new MLInteractiveLivenessDetectView.Builder()
                .setContext(this)
                // Set whether to detect mask 
                .setOptions(MLInteractiveLivenessCaptureConfig.DETECT_MASK)
                // Set the detection action , Silence is 0, Action for 1.
                .setType(1)
                // Set the preview position of camera video stream （ The upper left and lower right pixel values are based on the preview view）
                   .setFrameRect(new Rect(0, 0, 1080, 1440))
                // Set the live action call 
                .setActionConfig(interactiveLivenessConfig)
                // Set the face frame relative to the preview view The location of （ Top left and bottom right are based on 640*480 Image coordinates , It is suggested that the aspect ratio should conform to the actual face proportion ）, The function of the face frame is to detect the distance and offset of the face 
                .setFaceRect(new Rect(84, 122, 396, 518))
                // Set detection timeout , Suggest 10000 Millisecond or so .
                .setDetectionTimeOut(10000)
                // Set the result callback 
                .setDetectCallback(new OnMLInteractiveLivenessDetectCallback() {
                    @Override
                    public void onCompleted(MLInteractiveLivenessCaptureResult result) {
                    //  The result callback when the vivisection is completed 
                    swich(result.getStateCode()) {
                        case InteractiveLivenessStateCode.ALL_ACTION_CORRECT:
                        // After the verification is passed, the corresponding specific operations 

                        case InteractiveLivenessStateCode.IN_PROGRESS:
                        // Corresponding specific operation when detecting 
                        …
                        }
                    }

                    @Override
                    public void onError(int error) {
                    //  Error code callback when an error occurs in vivo detection 
                    }
                }).build();

        mPreviewContainer.addView(mlInteractiveLivenessDetectView);
        mlInteractiveLivenessDetectView.onCreate(savedInstanceState);
}

2. Yes MLInteractiveLivenessDetectView Set the life process monitor .
@Override
protected void onDestroy() {
    super.onDestroy();
    MLInteractiveLivenessDetectView.onDestroy();
}

@Override
protected void onPause() {
    super.onPause();
    MLInteractiveLivenessDetectView.onPause();
}

@Override
protected void onResume() {
    super.onResume();
    MLInteractiveLivenessDetectView.onResume();
}

@Override
protected void onStart() {
    super.onStart();
    MLInteractiveLivenessDetectView.onStart();
}

@Override
protected void onStop() {
    super.onStop();
    MLInteractiveLivenessDetectView.onStop();
}

Learn more >>

visit The official website of machine learning service

visit Official website of Huawei developer Alliance
obtain Development guidance document
Huawei mobile service open source warehouse address ：GitHub、Gitee

Pay attention to our , The first time to understand HMS Core Latest technical information ~